Datasets:

blesspearl

/

arxiv-summarization

like 0

quant-ph9810047

the phenomenon of localization manifests itself in many quantum mechanical systems ranging from the localization of light in a random medium @xcite , to anderson localization of an electronic wave @xcite and to atoms in time - dependent laser fields @xcite . in all these cases the underlying classical system is chaotic and shows diffusion as a function of time .in contrast , the quantum mechanical counterpart has a localized wave function whose width is governed by the classical diffusion and planck s constant @xcite . in the present paperwe show that there exists an additional quantum parameter that controls the localization length . in the system of a two - level ion stored in a paul trap @xcite and interacting with a standing wave it is the detuning between the transition frequency of the ion and the laser frequency .a recent paper @xcite has shown that a stored ion moving along a far - detuned standing wave shows localization in position and momentum variables . in this workwe have neglected the internal structure of the ion .we have therefore only considered the quantum dynamics of a particle moving in a one - dimensional time - dependent potential . in the present paperwe extend this analysis and consider the motion of an ion taking into account its internal dynamics .our paper is organized as follows : in sec . [ 2 ] wefirst summarize the essential ingredients of the problem .in particular , we introduce the relevant equations and describe the methods of solution for the classical and quantum mechanical equations of motion .we also define the quantities which we calculate such as the position and momentum distributions and their moments . in sec .[ 3 ] we discuss the dependence of the classical and quantum mechanical position and momentum distributions on the detuning .we find characteristic oscillations in the widths of these distributions as a function of the detuning .these oscillations are absent in the corresponding classical curves .we explain these structures by transforming the hamiltonian into an interaction picture .we emphasize that this is different to the case of atoms moving in a phase modulated standing wave . therethe corresponding oscillations appear in the widths of the quantum as well as of the classical momentum distributions @xcite .hence , the detuning is an additional parameter controlling the width and the shape of the quantum distributions .moreover , in the quantum systems discussed so far in the context of dynamical localization the quantum diffusion is always slower than the classical one .however , in the present model the new control parameter detuning can create situations in which the quantum diffusion temporarily exceeds the classical one .quantum interference effects such as dynamical localization are extremely sensitive to decoherence arising for example from noise or spontaneous emission @xcite .we are therefore forced to investigate in sec . [ 4 ] the influence of spontaneous emission .we find that the drastic difference between quantum and classical behavior is preserved in the limit of a far off resonance situation .we conclude in sec . [ 5 ] by summarizing our main results .we consider the standard paul trap set - up realized experimentally in many labs @xcite : a standing electromagnetic wave of frequency @xmath0 and wave number @xmath1 aligned along the @xmath2-axis couples the internal states @xmath3 and @xmath4 of a single two - level ion of mass @xmath5 to the center - of - mass motion .the resulting dynamics of the ion follows from the time - dependent schrdinger equation with the hamiltonian & = & + ^2 + _ 0 _ z + & + & _ 0 _ x ( k ) ( _ l ) .[ schr ] here the parameters @xmath6 and @xmath7 are proportional @xcite to the dc and ac voltages applied to the trap and @xmath8 are the standard pauli matrices .moreover , we denote the frequencies of the ac field , the atomic transition , the rabi frequency by @xmath9 , and @xmath10 , respectively .we introduce the dimensionless position @xmath11 , time @xmath12 and momentum @xmath13 . when we transform into an interaction picture with the unitary transformation @xmath14 the dimensionless hamiltonian in rotating wave approximationreads @xmath15 \hat{x}^2 - { k\hspace{-0.45em}\raisebox{0.7ex}{-}\hspace{0.2em}}\delta\hat{\sigma}_z \nonumber \\ & + & { k\hspace{-0.45em}\raisebox{0.7ex}{-}\hspace{0.2em}}\omega_0 \hat{\sigma}_x \cos \hat{x}. \label{ham } \end{aligned}\ ] ] here we have defined the dimensionless rabi frequency @xmath16 and detuning @xmath17 .the dynamics of the ion follows from the time dependent schrdinger equation @xmath18 for the state vector @xmath19 |x\rangle .\label{t_ev } \end{aligned}\ ] ] here the effective planck constant @xmath20 is consistent with the commutation relation @xmath21 = 2 k^2/(m \omega)[\hat{\tilde{x}},\hat{\tilde{p } } ] = 2 k^2/(m \omega ) i \hbar \equiv i { k\hspace{-0.45em}\raisebox{0.7ex}{-}\hspace{0.2em}}$ ] .the state vector @xmath22 provides the probabilities p_g(x , t ) |_g(x , t)|^2 or p_e(x , t ) |_e(x , t)|^2 to find the ion at the time @xmath23 at position @xmath2 given it is in the internal state @xmath24 or @xmath25 , respectively .when we are not interested in the internal states the position probability distribution @xmath26 reads p(x , t ) |_g(x , t)|^2 + we simulate the effect of spontaneous emission by the quantum monte carlo method @xcite using the effective non hermitian hamiltonian @xmath27 here we introduced the atomic operators @xmath28 and @xmath29 , and @xmath30 is the spontaneous decay rate scaled by @xmath31 .the moments of time when a spontaneous emission event takes place are chosen at random .then the wave function is projected onto the ground state and renormalized , that is @xmath32 .the recoil @xmath33 $ ] is chosen randomly according to the probability distribution @xcite @xmath34 of dipole radiation .when the results of single runs are averaged , we obtain the same result as predicted by a master equation @xcite . as an initial condition for the internal states we use the superposition @xmath35 andthe the center - of - mass wave function is a gaussian of width @xmath36 . in order to investigate localizationwe start with this wave packet at the origin where the classical phase space is a stochastic sea @xcite .we calculate the time evolution using the split - operator method @xcite with a grid of 8192 points .we control numerical errors using an adaptive time step size algorithm @xcite .this numerical integration allows us to find the time dependence of the position probability distributions @xmath37 and @xmath26 .moreover , we calculate the corresponding momentum distributions where the wave functions @xmath38 and @xmath39 in position space are replaced by the wave functions in momentum space . we compare and contrast these quantum results to the dynamics @xcite resulting from the classical equations of motion @xcite & = & p + & = & -(a+2q(2 t ) ) x + k_0 ( x ) r_1 [ bl1 ] for the center - of - mass motion .these equations are driven by the bloch equations @xcite _ 1 & = & - 2 r_2 + _ 2 & = & 2 r_1 + 2 _ 0 ( x ) r_3 [ bl2 ] + _ 3 & = & - 2 _ 0 ( x ) r_2 describing the internal dynamics .here @xmath40 , @xmath41 , and @xmath42 denote the in and out of phase quadratures of the dipole moment , and the atomic inversion , respectively . for the comparison with the quantum mechanical results we calculate 4096 trajectories starting from a classical gaussian ensemble centered initially at the origin and having the same widths in position and momentum as the quantum wave packet . in this waywe obtain the classical phase space distribution from which we find by integration over @xmath2 or @xmath43 the classical momentum or position distributions @xmath44 or @xmath45 .in the present section we study the influence of the detuning of the laser field with respect to the atomic transition on the dynamics of the system . herewe consider the following quantities of interest : the classical and quantum distributions , and their width as a function of time and detuning . in fig . 1 we show the influence of the internal structure of the ion on the classical and on the quantum diffusion . herewe analyze for various detunings @xmath46 the time dependence of the widths @xmath47 ( left column ) of the classical ( thin lines ) and quantum mechanical ( thick lines ) position distributions @xmath48 and @xmath26 ( right column ) . whereas the left column emphasizes the time dependence of the widths , the right column shows position distributions averaged over the time interval @xmath49 $ ] . in order to bring out most clearly the influence of the detuning we have neglected spontaneous emission in this figure . for a very small detuningthe quantum width @xmath47 lies well below the classical one .the latter increases as a function of time .in contrast , the quantum result displays oscillations around a steady state value .this suppression of the classical diffusion is due to dynamical localization as discussed in ref.@xcite .however , for slightly larger detunings the quantum curve is still oscillating but reaches partially above the classical one .for even larger detunings the quantum curve falls again below the classical one . eventually for larger detuningsit again goes partially above .hence , for a fixed detuning , there exist time regimes in which the quantum diffusion is stronger than the classical one .this detuning dependence of the quantum diffusion manifests itself in the shape of the quantum mechanical position distribution .the latter consists of a sharp peak and a broad background .the detuning essentially controls the shape of the background : depending on @xmath46 we find distributions either of negative or positive curvature .in contrast , the detuning hardly influences the classical distributions shown on the right column by thin curves .we emphasize that the momentum distributions not shown here display almost identical behavior . to discuss the influence of the detuning on the localization length we show in fig .2 the widths @xmath47 of the classical and the quantum distributions averaged from @xmath50 to @xmath51 .we note characteristic oscillations in the quantum mechanical curve as a function of the detuning @xmath46 .these oscillations do not appear in the classical curve which decreases monotonously for increasing @xmath46 .this feature becomes clear when we recall that the width of the classical distribution at a given time is determined by the classical diffusion constant .the latter is proportional to the perturbing potential caused by the laser field . in the limiting case of large detuning ,the effective potential becomes @xcite proportional to the effective coupling constant @xmath52 . as @xmath46 increases, the perturbation decreases and hence the diffusion is slower .an interesting domain is a small region around @xmath53 . herethe classical diffusion rate sharply decreases .the explanation of this feature follows from the set of eqs .( [ bl1 ] ) and ( [ bl2 ] ) : the effective potential provided by the laser field in eq .( [ bl1 ] ) is proportional to the factor @xmath54 .this quantity can take the maximum value of unity .for the resonant case , @xmath55 , we have @xmath56 and hence @xmath40 is independent of time and determined by the initial condition .the bloch vector coordinates corresponding to the initial state @xmath57 are @xmath58 and @xmath59 . hence @xmath60 results in the maximum driving strength of the center - of - mass motion .the curve for the quantum widths shows a distinctly different behaviour with a lot of structure .in particular , for @xmath61 we note a well developed resonance structure and two sharp minima and maxima .the minimum around @xmath62 is related to the separable quantum dynamics in the two diabatic potentials @xmath63 x^2 \pm \hbar \omega_0 \cos x$ ] . to bring this out most clearly we transform from the @xmath64 basis states to their superpositions @xmath65 , and @xmath66 . in this basisthe hamiltonian eq .( [ ham ] ) reads & = & ^2+^2 + & & + k_0 _ z + k_x , [ ham_tr ] and the coupling is now proportional to the detuning @xmath46 .hence we indeed find the potentials @xmath67 .when we now increase the detuning , transitions between the states @xmath68 and @xmath69 contribute to the quantum diffusion .therefore , the quantum diffusion rate quickly rises to the maximal classical one ; this corresponds to the formation of the first maximum .the next minimum fits the condition @xmath70 , where @xmath71 is the secular frequency of the trap .the origin of this resonance becomes clear , when we write the hamiltonian , eq .( [ ham ] ) , in the interaction picture where the dynamics of the trap as well as the internal state energy are transformed away by unitary transformations @xcite . in this picturethe hamiltonian reads h_(t ) = _ n=0^ _ k=-[n/2]^ _ l=-^ k & & _ l^(n , n+2k ) e^2i ( l- k_s+)t + & & _ + + [ hamflo ] with the lamb - dicke parameter @xmath72^{1/2 } } $ ] and & & _ l^(n , n+2k ) _ 0 [ i]^2k + & & _ -/2^/2 dt [ ^*(t)]^2k e^-^2|(t)|^2 l^2k_n(^2|(t)|^2 ) e^-2ilt . here @xmath73 denotes the @xmath74th energy eigenstate of the time independent reference oscillator @xcite with frequency @xmath75 .moreover , we have made use of the floquet solution @xmath76 , where @xmath77 is a periodic function .the solution @xmath78 obeys @xcite the differential equation + & = & 0 [ mathier ] with @xmath79 and @xmath80 . from the term @xmath81 in the hamiltonian ( [ hamflo ] ) we expect resonance effects , when @xmath46 satisfies the condition @xmath82 .for @xmath83 we have @xmath84 , which corresponds to two phonon transitions .this resonance suggests an enhancement of the quantum diffusion rate on resonance .instead , we have a deep minimum .in order to understand this counter - intuitive behavior we calculate the characteristic frequencies @xmath85 for @xmath86 and @xmath87 at different vibrational quantum numbers @xmath74 . from the inset of fig .2 we recognize that @xmath88 has a deep minimum around @xmath89 , which explains suppression of the diffusion over the vibrational states . as shown by the curve in the inset for @xmath90 there is no minimum just a decay . this causes the quantum diffusion to take on the classical value , which explains the formation of the second maximum @xcite .we conclude this section by noting that the phenomenon of oscillations in the width of distributions also appears in another quantum system showing classical chaos and quantum localization : an atom moving in a phase - modulated standing wave @xcite .however in this case the oscillations are of classical origin and appear both in the classical and quantum widths .we now consider the effect of spontaneous emission using the quantum monte carlo technique @xcite .the purpose of our investigations is twofold .first , we want to show that indeed the phenomena discussed so far are quantum interference effects .they are therefore sensitive to decoherence such as spontaneous emission .the latter predominantely occures when the laser is tuned close to resonance with the atomic transition .second , we want to show that in the far - detuned limit the effect of decoherence is negligible and the phenomenon of dynamical localization survives . in fig .3 we show the results of our simulations with a realistic rate of spontaneous emission corresponding to the decay rate of the @xmath91 transition at @xmath92 mhz in @xmath93 @xcite . we compare them to the classical and to the quantum result without noise .the loss of the coherence causes destruction of localization . in this casethe widths of the quantum mechanical position and momentum distributions are larger than the classical ones .this additional diffusion is caused by the random recoil kicks following each spontaneous emission event .when we neglect the recoil which of course is not realistic the quantum curve follows the classical one .the position and momentum distributions are of classical type , that is on the logarithmic scale they are polynomial curves . note that the pattern of the standing wave appears on top of the position distributions . in fig .4 we show the results for @xmath94 , corresponding to a detuning of @xmath95 ghz , a value which was mentioned in ref .the small oscillations in the quantum widths @xmath47 and @xmath96 are destroyed , but the main phenomenon , the substantial quantum suppression of classical diffusion is still visible . it is interesting to note that in a single realization of the dynamics the coherence of the two level superposition is completely destroyed by a single spontaneous emission .however , it affects only slightly the motional coherence because the population of the excited state is very low .furthermore , the small difference between the results with and without spontaneous emission in fig .4 is of the same order of magnitude that was found for the problem of an atom in a phase modulated standing wave @xcite . in the position distribution of the ground state, there is more probability in the classical - like background than in the case of no spontaneous emission. we have found that this background is very slowly growing , and no real steady state can be reached . however ,if we increase the detuning , the real steady state is approached asymptotically .since dynamical localization appears for a large range of parameters @xcite , there is a lot of room for optimizing the parameters .thus the phenomenon of dynamical localization in a paul trap could be observed experimentally . in order to observe not only dynamical localization but also the oscillations in the localization length discussed in the previous section, one should consider a configuration where decoherence is weaker even for small detunings , for example a different set of parameters in the present system , or a different system such as a raman transition between two ground states .there are basically two ways of controlling dynamical localization : ( i ) through the classical diffusion and ( ii ) through the quantized nature of the variable showing localization .this stands out most clearly in the estimate @xmath97 for the localization length . here@xmath98 is the classical diffusion coefficient determined by the perturbing potential .the coupling to this potential is called the control parameter , since it is a direct way to control the localization length .the scaled planck constant @xmath99 describes how important the quantization of the system is with respect to the perturbation . in this paperwe have shown that , when the perturbing potential has a quantum character as well , the relation @xmath97 is not exactly true anymore ; we observe oscillations in the localization length which do not appear in the classical diffusion rate .the parameter determining the oscillations could be called the quantum control parameter .we thank b. kneer and m. el ghafar for many fruitful discussions .t. and v. s. acknowledge the support of the deutsche forschungsgemeinschaft .we thank the rechenzentrum ulm and the rechenzentrum karlsruhe for their technical support .a. lagendijk and b. a. van tiggelen , phys. rep . * 270 * , 143 ( 1996 ) .p. w. anderson , phys .109 * , 1492 ( 1958 ) . for a review see , for example p. m. koch and k. a. h. van leeuwen , phys. rep .* 255 * , 289 ( 1995 ) ; g. casati , phys .a * 45 * , 7670 ( 1992 ) .f. l. moore , j. c. robinson , c. bharucha , p. e. williams , and m. g. raizen , phys .* 73 * , 2974 ( 1994 ) ; theoretical proposal in r. graham , m. schlautmann , and p. zoller , phys .a * 45 * , r19 ( 1992 ) .f. haake , _ quantum signatures of chaos _ ( springer - verlag , berlin , 1992 ) . ,eds . g. casati and b. chirikov ( cambridge university press , 1995 ) .w. paul , rev .phys . * 62 * , 531 ( 1990 ) .m. el ghafar , p. trm , v. savichev , e. mayr , a. zeiler , and w. p. schleich , phys .lett . * 78 * , 4181 ( 1997 ) .j. c. robinson , c. bharucha , f. l. moore , r. jahnke , g. a. georgaki , m. g. raizen , and b. sundaram , phys . rev . lett . * 74 * , 3963 ( 1995 ) .j. bardroff , i. bialynicki - birula , d. s. krhmer , g. kurizki , e. mayr , p. stifter , and w. p. schleich , phys .lett . * 74 * , 3959 ( 1995 ) . in the work of r.graham and s. miyazaki , phys .a * 53 * , 2683 ( 1996 ) the problem of atoms in a phase modulated standing wave was considered without the adiabatic elimination of the upper state .however , the emphasis of that paper was on spontaneous emission , not on the effect of an additional quantum degree of freedom . for experimental studies of the effect of noise and dissipation on dynamical localizationsee m. arndt , a. buchleitner , r. n. mantegna , and h. walther , phys .lett . * 67 * , 2435 ( 1991 ) ; f. l. moore , j. c. robinson , c. bharucha , b. sundaram , and m. g. raizen , phys .lett . * 75 * , 4598 ( 1995 ) ; r. blmel , a. buchleitner , r. graham , l. sirko , u. smilansky , and h. walther , phys .a * 44 * , 4521 ( 1991 ) , h. ammann , r. gray , i. shvarchuck , and n. christensen , phys .lett . * 80 * , 4111 ( 1998 ) ; b. g. klappauf , w. h. oskay , d. a. steck , and m. g. raizen , phys . rev .lett . * 81 * , 1203 ( 1998 ) .d. m. meekhof , c. monroe , b. e. king , w. m. itano , and d. j. wineland , phys .76 * , 1796 ( 1996 ) ; c. monroe , d. m. meekhof , b. e. king , and d. j. wineland , science * 272 * , 1131 ( 1996 ) ; d. leibfried , d. m. meekhof , b. e. king , c. monroe , w. m.itano , and d. j. wineland , phys .* 77 * , 4281 ( 1996 ) .g. birkl , j. a. yeazell , r. rckerl , and h. walther , europhys .* 27 * , 197 ( 1994 ) ; h. katori , s. schlipf , and h. walther , phys .* 79 * , 2221 ( 1997 ) .b. appasamy , y. stalgies , and p. e. toschek , phys .* 80 * , 2805 ( 1998 ) .r. dum , a. s. parkins , p. zoller , and c. w. gardiner , phys .a * 46 * , 4382 ( 1992 ) ; k. mlmer , y. castin , and j. dalibard , j. opt .b * 10 * , 524 ( 1993 ) .h. carmichael , _ an open system approach to quantum optics _ ( springer verlag , berlin , 1991 ) j. javanainen and s. stenholm , appl .* 21 * , 35 ( 1980 ) . for more information about the classical dynamics of this systemsee r. chacn and j. i. cirac , phys .a * 51 * , 4900 ( 1994 ) ; m. el ghafar , e. mayr , v. savichev , p. trm , a. zeiler , and w. p. schleich , j. mod . opt .* 44 * , 1985 ( 1997 ) .m. d. feit , j. a. fleck , and a. steiger , j. of comput .* 47 , * 412 ( 1982 ) .k. riedel , _ quanteneffekte in paul fallen , dynamische lokalisierung und dekohrenz _ , diplomarbeit , universitt ulm , 1997 , unpublished . g. m. zaslavsky , _ chaos in dynamic systems _ ( harwood academic publishers , chur , 1985 ) . in deriving these equations , we have assumed that the quantum mechanical position distribution is very narrow and localized near the classical trajectory @xmath100 .this property allows us to take the slowly varying cosine potential out of the integrals .l. mandel and e. wolf , _ optical coherence and quantum optics _( cambridge university press , 1995 ) a. p. kazantsev , g. i. surdutovich , and v. p. yakovlev , _ mechanical action of light on atoms _ ( world scientific , singapore , 1990 ) .j. bardroff , c. leichtle , g. schrade , and w. p. , phys .lett . * 77 * , 2198 ( 1996 ) ; ibid acta physica slovaka * 46 * , 231 ( 1996 ) .r. j. glauber , _ laser manipulation of atoms and ions _school of physics enricofermi course 118 , edited by e. arimondo , w. d. phillips , and f. strumia ( north holland , amsterdam , 1992 ) ; see also g. schrade , p. j. bardroff , r. j. glauber , c. leichtle , v. yakovlev , and w. p. schleich , appl .b * 64 * , 181 ( 1997 ) .there is a close similarity between such type behavior and the so called fano resonance : u. fano , phys .rev . * 124 * , 1866 ( 1961 ) .s. r. jefferts , c. monroe , e. w. bell , and d. j. wineland , phys .a * 51 * , 3112 ( 1995 ) .p. goetsch and r. graham , phys .a * 54 * , 5345 ( 1996 ) .

<S> we identify a new parameter that controls the localization length in a driven quantum system . </S><S> this parameter results from an additional quantum degree of freedom . </S><S> the center - of - mass motion of a two - level ion stored in a paul trap and interacting with a standing wave laser field exhibits this phenomenon . </S><S> we also discuss the influence of spontaneous emission . </S><S> # 1 # 1^ # 1 # 1^ # 1| # 1 # 1#1 | </S>

0810.4966

graphene has attracted intensive interest recently because of its novel properties , such as the room - temperature quantum hall effect @xcite @xcite and the dirac - equations - governed electrons novoselov2005@xcite .moreover the properties of the graphene nanoribbons@xcite change notably due to various edges @xcite , different symmetries @xcite , constrictions@xcite and even controllable defects by irradiating the material with electrons and ions @xcite@xcite@xcite .any wanted structures could be achieved by engineering the graphene with the help of state - of - art fabrication technologies , which supports the graphene as a prospective candidate for significant applications in electronics- and magnetics - related devices .defects have been revealed to be an influential factor in determining the transport properties of structures in nanoscale .for example , it has been known that defects can reduce or totally block the quantum conductance in carbon nanotube @xcite@xcite@xcite .graphene could be regarded as an unrolled carbon nanotube , and hence the transport of graphene in the presence of defects is also an interesting problem .actually , several kinds of defects in a single layer of graphene have already been paid attention to , such as pentagon / heptagon defects @xcite@xcite , vacancies @xcite , adatoms @xcite@xcite , substitution@xcite , disorder@xcite and even combinations of some of them @xcite .although the electronic energy band structures and magnetic properties in the presence of defect have been investigated in both graphene and graphene ribbons , how these properties manifest in transport is still lacking in attention .recently , it has been found that the interior pentagon / heptagon defects@xcite or the defects such as vacancies or disorder at the edges@xcite would lead to total quantum blockade in graphene nanoribbons at certain fermi energies .however , how the quantum blockade is connected with the severity or the positions ( not limited to the edges ) of the defect , and the size or the symmetry of the graphene nanoribbons is not so clear . as the adatom defect would not change the topological structure of graphene lattice to some extent, the connection between these factors and the defect can be unambiguously understood by comparing with the clean graphene system .as we will show , all these factors are really critical on the influence of the defect .therefore , a careful study of the influence of the adatom defect on the transport properties of graphene nanoribbons is deserved . in this paper, we investigate the electronic transport in narrow graphene ribbons in the presence of a single defect , which could be induced by the absorption of a hydrogen atom .the model of the defect is based on the ab - initio calculation .we find that the conductance profile is greatly reshaped by the defect . compared with the integer steps of perfect graphene nanoribbons , the conductance in the presence of a single defect drops even down to zero in the low energy regime , leading to a transformation from metal to semiconductor .accompanied by this global conductance blockade , a loop current builds up around the defect .the conductances of the defective graphene nanoribbons with different widths , edges , and positions of the defect are studied .we find the conductance blockade and the coexistence of the loop current sensitively depend on the severity and relative position of the defect .for metallic armchair - edges graphene nanoribbons , whether the conductance blockade and the coexistence of the loop current develop or not , they could be controlled by the position of the defect .the transport properties in graphene ribbons under low bias voltages are governed by the @xmath0 bonded electrons , which are modeled by the tight - binding spinless hamiltonian @xmath1 where @xmath2 ( @xmath3 ) creates ( annihilates ) an electron on the site @xmath4 , @xmath5 is the nearest - neighbor hopping energy and is set as 2.7 @xmath6 in this paper .the distance between the nearest neighbor carbon atoms is @xmath7 ( @xmath81.42 ) .the conductance can be calculated by the landauer - buttiker formula based on the green function method@xcite@xcite@xcite@xcite .the formalism separates the entire system into three parts , the central region and two leads . to mimic the transport through infinitely long ribbon with specific edge ,two semi - infinite graphene leads are connected to the central region , see fig.[fig1 ] ( a ) .the zero temperature conductance @xmath9 of the graphene nanoribbons is given by @xmath10 $ ] , where @xmath11 is the retarded ( advanced ) green - function matrix of the central part , @xmath12 is the retarded ( advanced ) self - energy due to the source ( drain ) [ labeled as s(d ) in the fig.[fig1 ] ( a ) ] , and @xmath13 $ ] . here@xmath14 represents the unitary matrix .the local current density at the fermi level @xmath15 between two neighboring sites @xmath4 and @xmath16 can be expressed as @xmath17 , \label{current}\]]where @xmath18 is the electron correlation function and @xmath19 is the corresponding matrix element of the hamiltonian ..,width=340 ] in the absence of the defect , the well - known conductance of the perfect graphene nanoribbon with zigzag edges is shown by the dot line in fig .due to the metallic nature of the system , there exists no energy gap .the quantum conductance begins with @xmath20 ( defined as @xmath21 ) and displays a typical step - increasing feature as a function of the fermi energy .we also show in fig .[ fig1 ] ( a ) the spatial distribution of the local current density at a selective energy value of @xmath15=0.32@xmath5 .one can see that the current throughout the nanoribbon is uniform along the transport direction and forms unambiguous paths . at this energy ,the current density amplitude at the central region of the nanoribbon is bigger than that at the edge regions .0.05e / @xmath22 ) .charge flows from the blue into the red regions.,width=226 ] now , let us focus on the effect of a single defect site on the quantum conductance and local electric current distribution in the graphene nanoribbons .to physically illustrate how to model the presence of a single defect site by coupling a single carbon atom in the graphene with an adatom atop , we have simulated the atomic hydrogen adsorption on the graphene [ see fig .[ adatom](a ) ] at a low coverage of 0.02 monolayer ( namely , one hydrogen atom presents in a 5@xmath235 graphene supercell ) , by using density - functional theory within the local density approximation and a supercell approach .the total energy calculation shows that the most stable adsorption site for h adatom is the on - top site and the h - c bond is characterized by a hybridization of h 1@xmath24 and c 2@xmath25 states .prominently , it is found that the c atom beneath the h adatom is pulled a little bit out of the garphene plane by h - c chemical bonding , which obviously alters the coupling between this c atom and its three neighboring c atoms and thus results in a single lattice defect site .for further clarification , fig .[ adatom](b ) plots the calculated electron density difference @xmath26(@xmath27 ) , which is obtained by subtracting the electron densities of noninteracting component systems , @xmath28 ( @xmath27 ) + @xmath29(@xmath27 ) , from the density @xmath30(@xmath27 ) of the h / graphene system , while retaining the atomic positions of the component systems at the same location as in the h / graphene system .one can see that the charge redistribution only occurs around the h - c bond and influences the three neighboring c - c bonds , while the charge densities of other c atoms are not affected at all .based on the physical picture revealed in fig .[ adatom ] , we model the presence of a single defect lattice site by modifying the interactions between this lattice site and its nearest neighboring sites while keeping other hopping elements unchanged. therefore we change the hopping amplitude between the single defect site [ hollow circle in fig .[ fig1](c ) ] and its three nearest neighbors from @xmath5 to , for example , @xmath31=0.3@xmath5 .the conductance of the system with the defect is calculated and plotted with the solid line in fig .[ fig1](b ) .one can see that the conductance is prominently reshaped by the presence of the single defect site .the explicit changes include : ( i ) compared to the perfect case , the quantum step feature disappears ; ( ii ) the conductance drops for most fermi energies , ( iii ) zero - conductance dips appear at certain fermi energies .these defect - induced phenomena are understandable because the defect is a kind of destructive factor for the electron transmission in otherwise perfect graphene lattices .we now concentrate on the conductance dip at @xmath15=0.32@xmath5 .since the conductance drops to zero at this point , it means that the transmission is totally blockaded by the defect .the dip also implies the appearance of energy gap and consequently a transformation from metal to semiconductor happens . for further explanation to the conductance dip, we calculate the local current density at the fermi energy ( @xmath15=0.32@xmath5 ) and plot the result in fig .[ fig1 ] ( c ) .compared to the ideal graphene nanoribbon system [ fig .[ fig1](a ) , plotted at the same energy @xmath15=0.32@xmath5 ] , it shows in fig.[fig1 ] ( c ) that the pattern of local current density distribution is prominently changed in the defect - included system and there appears loop current circling around the defect site .the magnitude of this loop current around the defect can be larger than the maximal current in fig.[fig1](a ) .note that the current is normalized to the maximum value for each case .the loop current would produce an effective magnetic field , which is consistent with the previous experimental and theoretical predication.@xcite@xcite@xcite the formation of the loop current could be explained as the interference of all transport channels . in the two dimensional graphene lattices ,an electron could jump from source to drain via many different transport channels .the local electric current actually demonstrates the interference results of these channels . a defect would cause the blockade of relative transport channels to some extent and therefore change the local electric current .a discussion using a simple one - dimensional model is given in reference@xcite .as the different transport channels are numerous in the two dimensional graphene lattices , whether the loop current appears and what kind of patterns it takes depend on the whole environment of the defect , such as the edges and width of the graphene ribbons , the position and severity of the defect , etc . therefore , we would dwell on the influence of these parameters on the transport properties in the defective graphene ribbons . decreasing from top to bottom .the system is the same one as in fig.[fig1 ] .right : conductances of the graphene nanoribbons with different widths . in this case, the position of the defect and length of the ribbon is the same as the one in fig.[fig1 ] .the width changes by cutting or adding chains of c atoms at the bottom of the graphene nanoribbons .l is number of the atoms connecting the source lead .for every plot , the conductance of the perfect graphene nanoribbon is shown for reference with the dotted lines.,width=302 ] we can define the severity of the defect site by the ratio @xmath33 .the vacancy defect is actually the defect with @xmath33=@xmath34 .a question thereby arises : how effective could the defect cause the conductance to drop to zero ? in the following , we will show that it depends on the competition between the width of the graphene ribbon and the severity of the defect . for a narrow graphene ribbon ,the zero conductance exists in most cases when for small ratios of @xmath33 . in the left panel of fig.[td ] , we present the conductance of the same defective graphene ribbon as in fig.[fig1 ] but with different @xmath33 .we can see that as the ratio @xmath33 decreases , the dip begins to appear in the lowest conductance step and conductance reaches zero at @xmath33=@xmath35 .this critical value at which zero conductance is obtained is also dependent on the width of the graphene ribbon . to show the influence of the width, we modify the width of the defective graphene nanoribbon shown in fig.[fig1 ] by cutting or adding chains of c atoms on the lower edge of the ribbon . the width @xmath36 is denoted by the number of chains . in the right panel of fig.[td ], the conductances of the graphene ribbons with four different widths are plotted .it is obvious that as the width of the graphene ribbon gets wider , the dips become shallow and narrow and finally almost disappear .for the case of @xmath37 , the values of conductances almost get equal between @xmath38 and , for example , @xmath39 . however , the loop current only exists near @xmath38 , not @xmath39 , although the net current can not be totally blocked by the defect . , shown with solid line .other parameters are the same as fig.[fig1](b ) . for reference , the case of a perfect lattice ( dot line ) and the one with a single defect but assuming on site energy unchanged ( dashed line ) are also shown.,width=302 ] besides the change of inter - site hopping matrix elements by the coupling of a single lattice site with the adsorbed atom , the on - site energy of this lattice sitemay also be changed due to the bonding and anti - bonding hybridizations .the conductance by changing the on - site energy of the defect site by 0.1@xmath5 is shown in fig.[onsite ] .we see that the conductance dip still appears but lies at a shifted energy not far away .the distance depends on the changed amplitude of on - site energy . at the conductance dip, the loop current also exists without apparent difference from the one shown in fig.[fig1 ] ( c ) .so we conclude that the on - site energy is not an influential factor to the occurrence of conductance dip and formation of loop current induced by the single defect in this model .we have also studied the influence of the position of the defect site on the conductance of the graphene nanoribbons .it is found that the horizontal shift of a single defect produces no changes to the conductance , while the different defect positions in vertical direction leads to different conductances and hence the center of loop current patterns . in fig .[ fig2](a ) we choose respectively four sites ( labeled by a , b , c and d ) as a single defect site to make a comparison .the conductance corresponding to these four choices for the defect site are plotted in fig .[ fig2](b ) .one can see that for each case the conductance drops to zero with a dip . the width of the dip becomes smaller with the site going closer to the boundary , see the inset in fig.[fig2 ] ( b ) . by comparing the current flowing through these sites in perfect lattice ,we see that the defect locating on the position with bigger current in the perfect lattices would cause a conductance dip with a broader width .the role of the presence of a single defect lattice site becomes more important in the graphene nanoribbons with armchair edges . as we know, a graphene nanoribbon with armchair edges has the zero energy gap only when the number of atoms on a zigzag edge connecting the source ( or lead ) is @xmath40 ( @xmath41 is an integer ) .this kind of graphene nanoribbons hence has a quantum conductance beginning from @xmath20 with increasing the electron fermi energy .we plot the local current distribution of the system without the defect in fig .[ fig3](a ) , which clearly displays a series of well - separated pathes for current flow .each path has the same weight in current amplitude .it is the very result of interference between the transport channels in the metallic armchair - edges graphene ribbons.@xcite according to this stripe current distribution , the lattice sites in the graphene nanoribbon can be therefore classified into two parts , i.e. , sites like a [ see fig . [ fig3](a ) ] carrying current flow and sites like b without carrying current flow .we set the single defect site to be on site a or on site b , and then calculate the corresponding quantum conductance respectively .the results are shown in fig .[ hamiltonian](b ) and [ fig3](c ) , respectively .remarkably , it is shown in fig .[ fig3](b ) that when a single defect is placed on site a , no conductance dip occurs at all and the first quantum conductance step is invariant upon the presence of the defect .whereas , when the defect is placed on site b , two dips develop in the conductance spectrum , as shown in fig .[ fig3](c ) .again , at the energy position @xmath15=0.32@xmath5 of the conductance dip , the spatial distribution of the internal current ( not shown here ) is featured by a loop current around the defect site .thus , the lattice sites b are more important than the lattice sites a for the transport in this kind of armchair - edges graphene ribbons .this is an important ingredient for practical control of the graphene transport by manipulating the position of a single defect site . from the point of the interference among different transport channels mentioned above , the defect in site b gives a kind of destructive interference .the local current density distribution pattern of the perfect metallic armchair - graphene ribbons indicates that the total conductance is mainly contributed by several parallel and independent channels . at the sitea , the local density is zero . in other words ,the site a is decoupled and isolated from these channels .that is why the graphene ribbon is immune to the defect on site a.in summary , we have studied the influence of a single defect on the electron transport properties in narrow graphene nanoribbons with different edges , widths and positions of the defect .it has been found that the conductances may drop even down to zero at certain fermi energies in the presence of a single defect site for metallic graphene nanoribbons .the conductance dip is accompanied by the formation of a remarkable loop current around the defect .the width of the conductance dip is sensitive to the position as well as to the severity of the defect site . for the graphene nanoribbons with armchair edges, it has been found that the transport properties , including the low - energy conductance step , the conductance dip and the loop current , can be controlled by choosing the position of the single defect .this result enlightens us on controlling the conductance of the graphene nanoribbons by simply manipulating the defects due to the absorbed atom .

<S> we investigate theoretically the electronic transport properties in narrow graphene ribbons with an adatom - induced defect . </S><S> it is found that the lowest conductance step of a metallic graphene nanoribbon may develop a dip even down to zero at certain values of the fermi energy due to the defect . accompanying the occurrence of the conductance dip , a loop current develops around the defect . </S><S> we show how the properties of the conductance dip depend on the parameters of the defect , such as the relative position and severity of the defect as well as the width and edges of the graphene ribbons . in particular , for metallic armchair - edges graphene nanoribbons , whether the conductance dip appears or not , they can be controlled by choosing the position of the single defect . </S>

hep-th0001065

anti - de sitter ( ads ) gravity has attracted much attention due to the conjectured correspondence to a conformal field theory ( cft ) on the boundary of the ads spacetime @xcite leading to the so - called ads / cft correspondence .( for a review , see @xcite ) . the ads / cft correspondence has been extended to a dw / qft correspondence @xcite for d@xmath0branes in ten dimensions . in this talkwe extend the discussion of @xcite to general two - block @xmath0branes in various dimensions . in section 1we first discuss some general facts on domain - wall and anti - de sitter spacetimes . in section 2we calculate the near - horizon geometries of a generic @xmath0brane . in section 3we discuss the field theory limit for the general case .the formulae we give in this section are applied to the special case of 0branes we discuss in the section 4 . in this final sectionwe discuss the quantum mechanics of 0branes , or extreme black holes , in various dimensions .domain - wall ( dw ) spaces , i.e. spaces of co - dimension 1 , occur as solutions to the equations of motion of a ( super-)gravity action with a dilaton scalar @xmath1 and a ( d-1)form gauge potential ( for a review , see @xcite ) .they are @xmath0branes with worldvolume dimension @xmath2 which is one less than the dimension @xmath3 of their target spacetime , i.e. @xmath4 .the part of the supergravity action needed to describe the dw solution is given by ( we use the einstein frame and mostly plus signature ) : @xmath5\ , , \ ] ] where @xmath6 is the gravitational coupling constant with ( ignoring constants ) @xmath7 and @xmath8 is the dilaton coupling parameter .we also introduced a parameter @xmath9 defined by @xmath10 solving the equations of motions following from ( [ einstcosm ] ) one finds : @xmath11 with @xmath12 a parameter and where @xmath13 and @xmath14 ( which is invariant under dimensional reduction ) are defined by @xmath15 the function @xmath16 is harmonic on the 1-dimensional transverse space with coordinate @xmath17 : @xmath18 with @xmath19 constant .the domain - wall is positioned at the discontinuity @xmath20 .different choices of @xmath12 correspond to different choices of coordinates and lead to different expressions for the metric @xcite .for instance , one can choose @xmath12 such that the powers of the harmonic function become equal or opposite .shifting the position of the domain wall to infinity , e.g. @xmath21 allows us to discard the constant @xmath22 in the harmonic function and , furthermore there is only one side of the domain wall . we can eliminate @xmath12 if we define a mass parameter @xmath23 by : @xmath24 with @xmath25 . making a co - ordinate transformation and by going to the so - called `` dual frame '' metric @xmath26 , indicated by a subscript star , @xmath27 we find for the solution ( [ domainwall ] ) @xmath28 in the dual frame ( [ reg ] ) the domain - wall solution ( [ dwr ] ) describes an @xmath29 spacetime with a linear dilaton , the latter breaking the full conformal structure of the ads spacetime. this observation will be the key to generalizing the ads / cft duality to a dw / qft duality .our starting point is the @xmath3-dimensional action @xmath30\ , , \ ] ] which contains three independent parameters : the target spacetime dimension @xmath3 , the dilaton coupling parameter @xmath31 and a parameter @xmath0 specifying the rank @xmath32 of the field strength @xmath33 . the parameter @xmath9 is a generalization of ( [ park ] ) and is given by @xmath34 we have furthermore introduced two useful dependent parameters @xmath35 and @xmath36 which are defined by @xmath37 we next consider the following class of diagonal `` two - block '' @xmath0brane solutions ( using the einstein frame ) .the p - brane solution is _ electrically _ charged with respect to the p+1form potential . ] : @xmath38 where @xmath39 we use a constant , i.e. metric independent , levi - civita tensor .furthermore , @xmath40 and @xmath41 is a generalization of the @xmath14 in the previous section defined by @xcite @xmath42 which is invariant under reductions and oxidations ( in the einstein frame ) .the function @xmath16 is harmonic over the @xmath43 transverse coordinates and , assuming that @xmath44 ( i.e. no constant or logarithmic harmonic ) this harmonic function is given by @xmath45 here @xmath46 is an integration constant with the dimensions of length .it is related to the mass and charge of the @xmath0brane a follows .the mass @xmath47 per unit p volume is given by the adm formula : @xmath48 on the other hand , the charge @xmath49 per unit p - volume is given , in terms of the same integration constant @xmath46 , by the gauss - law formula @xmath50 hence , the p brane solution satifies the bps bound @xmath51 to derive an expression for @xmath46 in terms of the string parameters @xmath13 and @xmath52 , which fixes the scaling of @xmath16 , one must add a source term to the supergravity bulk action . using the no - force condition and the fact that in the string frame the electric ( p+1)form potential @xmath53 is proportional to @xmath54 , as follows from the action ( [ generalaction ] ) , we find that @xmath55 comparing with ( [ taup ] ) and using ( [ kappad ] ) we deduce that , for a single brane , @xmath56 the two - block solutions ( [ 2block ] ) include the ( supersymmetric ) domain - wall spaces of the previous section .they correspond to the case @xmath57 , @xmath58 and @xmath59 .the solutions also include the known branes in ten and eleven dimensions ( m2 , m5 , d@xmath0 , f1 , ns5 etc . ) as well as branes in lower dimensions .if the branes under consideration preserve any supersymmetries we can set @xcite @xmath60 where generically @xmath61 is the number of unbroken supersymmetries .we now consider the limit for which the constant part in the harmonic function is negligible . as in the previous sectionwe make a co - ordinate transformation and go to the dual frame @xmath62 after these manipulations we can write the near - horizon metric as @xmath63 which has an @xmath64 geometry and a linear dilaton . reducing over the @xmath65 angular variables of the sphere we end up with a gauged supergravity in @xmath66 dimensions of the form ( [ einstcosm ] ) supporting a domain - wall solution of the form ( [ domainwall ] ) . the precise relation between the parameters of the action( [ einstcosm ] ) and its solution ( [ domainwall ] ) in terms of those of the action ( [ generalaction ] ) and its solution ( [ nhgeometry ] ) can be found in section 2.3 of @xcite .summarizing , in this section we showed that in the dual frame , defined by ( [ dualframe ] ) , all @xmath0branes solutions ( [ 2block ] ) have a near horizon geometry @xmath67 .the domain - wall metric has all the isometries of an ads space .these isometries are broken in the full background because of the presence of a linear dilaton .in this section we will set up the framework for the dw / qft duality similar to the analysis of @xcite but for arbitrary dimensions .the analysis is similar to the ads / cft duality .there one finds ( in the low - energy limit ) that supergravity in the near - horizon geometry of a large number of non - dilatonic branes is dual to the conformal field theory living on these branes ( which are located at the boundary of the near - horizon geometry ) .this so - called holographic principle lies at the heart of the ads / cft conjecture .as we showed in the previous sections , the near - horizon geometry of a general @xmath0-brane in the dual - frame is equivalent to that of a non - dilatonic @xmath0-brane .it is therefore natural to assume that the duality might be extended .the presence of the dilaton makes the ads background into a domain - wall background and the conformal field theory into a general quantum field theory , hence the name dw / qft duality . in the followingwe will only consider d - branes and their intersections in lower dimensions .one starts with a d@xmath0-brane configuration in string theory and takes the low - energy limit in which only the massless modes survive and in which the theories in the bulk and on the brane decouple .next , one introduces a new energy - scale and a dimensionless coupling constant which are the free parameters of the decoupled theories .the energy - scale can be given in several ways : [ probes ] if one probes the stacked branes by another d@xmath0-brane , a natural energy - scale is given by the mass of the endpoint of a stretched string , which acts like a w boson ( see figure 1 ) @xmath68 this was the approach taken by @xcite , all branes have the same energy - scale but the disadvantage is that the near - horizon geometry when written in the @xmath69 co - ordinates does not take the same form for all branes .one can also probe the branes by a supergravity field @xmath70 ( suppressing possible quantum numbers ) . by solving the wave equation for this field @xciteone finds the following energy scale @xmath71 in @xcite it was shown that this energy - scale corresponds to the so - called holographic energy - scale used in entropy calculations .although the energy depends on the parameters of the brane solution , one finds that the near - horizon geometry ( [ dwr ] ) in these holographic @xmath72 co - ordinates is particularly simple : @xmath73\ , .\ ] ] for all branes this metric has the form @xmath74 differing only in the ( relative ) radii of the sphere and anti - de - sitter space .the quantum field theory on the brane is in general not explicitly known , but dimensional analysis enables us to extract some information .if we assume that the theory is given by some @xmath75-form potential on the @xmath76 dimensional worldvolume @xmath77 then we have for the coupling constant scaling behavior @xmath78 for 10-dimensional d@xmath0-branes we have a vector - multiplet ( i.e. @xmath79 ) on the worldvolume reproducing the familiar @xmath80 scaling behavior for the yang - mills theory on the d@xmath0-brane worldvolume .the effective dimensionless coupling constant @xmath81 is given by @xmath82 since we have two different probes , we can in principle construct two different dimensionless couplings , constructed by either @xmath83 or @xmath84 . since both probes have a sensible interpretation , it must be possible to use them both , independently of the low - energy limit .this leads to the following constraint on the scaling behavior in terms of the brane - solution parameters @xcite : @xmath85 combining this with the definitions of the probe - scales @xmath86 and @xmath84 we find the following relations between them and their corresponding dimensionless couplings @xmath87 and @xmath88 @xmath89 the decoupling limit is now given by @xmath90 the dimensionless coupling made from newton s constant in this limit becomes @xmath91 so that if @xmath92 becomes too large one has to take @xmath93 to infinity to ensure the decoupling of gravity . in the ads/ cft correspondence one has in the field theory side the parameters @xmath94 which correspond to the curvature and string coupling on the supergravity side .they generalize to the effective tension of a string in the dual frame and the dilaton , respectively : @xmath95 we now have the following ranges of validity on the two sides of the duality * perturbative sym : @xmath96 * classical sugra : @xmath97 using the above formulae one can easily see that classical supergravity describes strongly coupled , large n field theory .however , the conformal invariance which in the ads / cft duality facilitates computations in the strongly coupled field theory is now broken so that any direct check of a dw / qft duality is ruled out . for specific examples ,we refer to @xcite . as long as @xmath98 is positive , the supergravity and field theory are valid in different regimes ( i.e. @xmath99 and @xmath100 ) and one can find consistent phase diagrams as in @xcite .however , for negative @xmath98 , e.g. in case of the d6-brane , this seems no longer true at first sight : @xmath101 and @xmath100 .as one can see by using the relation ( [ proberel ] ) between the two couplings , this corresponds to @xmath102 and @xmath100 .this matches nicely with the explanation of @xcite , namely that the low - energy hilbert space of the field theory has two separate sectors : one describing nearby ( in u co - ordinates ) brane probes and one describing supergravity probes far away ( also in u co - ordinates ) from the brane .finally we note that when the dilaton becomes larger one can perform an s - duality transformation .if the curvature is small with respect to the s - dual string scale , then one can still trust the supergravity approximation .in this section we consider the special case of 0branes in various dimensions .the ( bosonic ) lagrangian for a particle with mass @xmath23 and charge @xmath75 , moving in the string frame near - horizon background of @xmath93 stacked 0branes reads @xmath103 where the dot represents derivatives with respect to the worldline time .the d - dimensional 0brane solution in the dual frame @xmath104 is given by the expression ( [ limitsol ] ) , taken for @xmath105 .introducing the canonical momentum @xmath106 one can write down the mass - shell constraint in the string frame for the probe particle as @xmath107 we would like to solve this equation for @xmath108 .for this purpose , we transform the mass - shell equation ( [ mass - shell ] ) to the dual frame and substitute the metric and gauge field of the solution ( [ limitsol ] ) . for the gauge field ( being in the temporal gauge ) we can write : @xmath109 and we find for the hamiltonian @xmath110 we can write this hamiltonian in a rather suggestive form by making the following transformation @xmath111 after which the hamiltonian takes on the following form @xmath112 the dynamics of the system is now generated by the poisson brackets with respect to the canonical co - ordinate and momentum @xmath113 . the transition to quantum mechanicsis the usual one . defining two other operators@xmath114 we find that the system possesses a classical @xmath115-symmetry @xmath116 this is only true when the effective coupling constant @xmath88 of the model under consideration is fixed under scale transformations .whenever we have a non - trivial dilaton the model will not be conformally invariant by itself .only when introducing a transformation to keep @xmath88 fixed under conformal transformations , will the model be invariant under what are called generalized conformal transformations @xcite . taking the limit @xmath117 and at the same time @xmath118 @xcitewe find a `` non - relativistic '' hamiltonian of the form : @xmath119 this model , as first considered in @xcite , is in fact an integrable model so that it might offer a concrete test of the @xmath120 duality in this particular limit .the work reported here is based upon hep - th/9907006 .we thank our collaborators klaus behrndt and jan pieter van der schaar for numerous discussions .this work is supported by the european commission tmr programme erbfmrx - ct96 - 0045 , in which e.b . and r.h .are associated to the university of utrecht .j. maldacena , _ the large n limit of superconformal field theories and supergravity _ _ adv .phys . _ * 2 * ( 1998 ) 231 , hep - th/9711200 ; + s.s .gubser , i.r . klebanov and a.m.polyakov , _ gauge theory correlators form non - critical string theory _ , phys* b248 * ( 1998 ) 105 , hep - th/9802109 ; + e. witten , _ anti - de sitter space and holography _ , adv .* 2 * ( 1998 ) 253 , hep - th/9802150 .petersen , _ introduction to the maldacena conjecture on ads / cft _ , int .j. mod .a14 ( 1999 ) 3597 , hep - th/9902131 ; + o. aharony , s. s. gubser , j. maldacena , h. ooguri and y. oz , _ large n field theories , string theory and gravity_ , hep - th/9905111 .h. l , c.n .pope , e. sezgin and k.s .dilatonic @xmath0brane solitons _ , phys . lett .* b371 * ( 1996 ) 46 , hep - th/9511203 ; _ stainless super @xmath0branes _ , nucl . phys .* b456 * ( 1995 ) 669 , hep - th/9508042 .p. claus , m. derix , r. kallosh , j. kumar , p.k .townsend and a. van proeyen , _ black holes and superconformal mechanics _ , phys . rev* 81 * ( 1998 ) 4553 , hep - th/9804177 ; + r. kallosh , _ black holes and quantum mechanics _ , hep - th/9902007 .

<S> we investigate domain - wall / quantum field theory correspondences in various dimensions . we give particular emphasis to the special case of the quantum mechanics of 0branes . </S>

astro-ph0208178

an evaluation of gaia s contribution to the extragalactic distance scale requires as a first step a brief description of the present situation ( section [ sec:2 ] ) .inadequacies of the present situation of the distance scale are discussed in section [ sec:3 ] .important improvements expected from gaia are outlined in section [ sec:4 ] .some conclusions are given in section [ sec:5 ] .the two pillars of the extragalactic distance scale are cepheids and supernovae of type ia , which are discussed in turn , as well as other distance indicators .the fundamental rle of classical cepheids for extragalactic distances rests on their period - luminosity ( p - l ) relation , whose shape and zero point must be known .most published cepheid distances are based on a linear , wavelength - dependent p - l relation as derived by madore & freedman ( 1991 ) from lmc cepheids : @xmath10 ( as to the zero point see 2.1.2 . ) .these equations yield apparent moduli @xmath11 , @xmath12 , and @xmath13 , still affected by galactic , and in the case of extragalactic cepheids by internal absorption . the true distance modulus @xmath14 can be obtained by a combination of the apparent moduli if an absorption law is adopted ( here : @xmath15 , @xmath16 , and @xmath17 ) , i.e. @xmath18 almost all galaxies with cepheid distances outside the local group have been observed with hst in @xmath19 and @xmath20 . if one assumes ( optimistically ) that the apparent moduli @xmath12 and @xmath13 have errors of @xmath21 the typical error of a cepheid modulus @xmath22 becomes @xmath23 ( 7% in linear distance ) from equation ( [ eq:5 ] ) . if the cepheids in a galaxy suffer variable absorption they can be reduced one by one with equations ( [ eq:4 ] ) and ( [ eq:5 ] ) ; their true distances can then be averaged .the ogle programme has provided new @xmath24 , @xmath19 , and @xmath20 photometry for hundreds of fundamental - mode lmc cepheids ( udalski et al .1999 ) which requires a significant change of slope of the p - l relations at periods of @xmath25days ( cf .[ fig:1 ] ) @xmath26 -@xmath27 relations in @xmath24 , @xmath19 , and @xmath20 of the fundamental - mode lmc cepheids adopted by udalski et al .small symbols : @xmath28 ; large symbols : @xmath29 . ]a corresponding change of slope is also seen in the period - color ( p - c ) relations ( cf .[ fig:2 ] ) . and @xmath30 vs. @xmath31 diagrams of the fundamental - mode lmc cepheids adopted by udalski et al . ( 1999 ) .color excesses from that source .symbols like in fig . 1 .] most of the available extra - local group cepheid distances rely on variables with @xmath29 .their distances derived from equation ( [ eq:10 ] ) and ( [ eq:11 ] ) are smaller by @xmath32 as those from equations ( [ eq:2 ] ) and ( [ eq:3 ] ) .all following distances are here still based on the conventional p - l relations in equations ( [ eq:2 ] ) and ( [ eq:3 ] ) . the zero point of the p - l relations in equations ( [ eq:1])@xmath33([eq:3 ] ) ( as well as the one in equations [ eq:6]@xmath33[eq:11 ] ) is based on an adopted lmc distance modulus of @xmath34 ( cf .table [ tab:1 ] ) .this value is the straight mean of various methods to determine the lmc distance .a straight mean is preferred here because the systematic errors are quite uncertain ..the distance modulus of lmc [ cols="<,<,<",options="header " , ] @xmath35 group / cluster membership as defined by kraan - korteweg ( 1986 ) @xmath36 two separate agglomerations can be distinguished in the uma cluster ; the nearer , presumably more nearly complete agglomeration with 18 tf distances is considered here ( federspiel 1999 ) @xmath37 cluster membership as defined by binggeli et al .( 1993 ) sources : ( a ) for a list of individual cepheid distances see tammann et al .( 2001 ) ( the number of cepheid distances is shown in parentheses ) ; ( b ) sneia calibration as in section 2.3 ( the number of sneia is shown in parentheses ) ; ( c ) federspiel ( 1999 ) ; ( d ) luminosity classes ( sandage 2001 ) ; ( e ) globular clusters ( tammann & sandage 1999 ) ; ( f ) kennicutt et al .( 1998 ) ; ( g ) tammann & sandage ( 1999 ) the seven groups / clusters within @xmath0 in table [ tab:2 ] define a hubble diagram with @xmath38 ( giving half weight to the nearby south polar group ) with a scatter of only @xmath39 . if also the coma cluster is included one obtains @xmath40 , @xmath41 .the distances discussed in sections 2.1 to 2.4 are plotted in a distance calibrated hubble diagram ( fig .[ fig:3 ] ) .they are best fit by a local value of @xmath42 and a large - scale value out to @xmath43 of @xmath44 .the difference between the local and cosmic value of @xmath45 is insignificant .this demonstrates that the expansion rate changes very little with scale .based on cepheids ( large dots ) , tf distances ( small dots ) , and groups / clusters ( squares ) is independent of the right - hand value of @xmath45 based on sneia ( diamonds ) , except that all distances rest ultimately on cepheids .the full drawn line corresponds to @xmath46 , @xmath47 . ]one could gain the impression from section 2 that the problem of extragalactic distances was solved .nothing is further from the truth .only about 200 galaxies with random individual distance errors between @xmath48 and @xmath49 were involved . for the vast majority of galaxiesone has still no distances .of course , a dynamical distance can be obtained from the recession velocity and an adopted value of @xmath45 , but the accuracy is limited by peculiar motions . if one assumes @xmath50 as a typical large - scale peculiar motion , as judged by the local motion with respect to the microwave background , the dynamical distances of galaxies beyond @xmath9 are affected by only @xmath51 from peculiar motions . for galaxies at @xmath52 the dynamical distance error is roughly tripled , but stays at this value for still nearer galaxies because the peculiar motions tend to decrease over shorter scales .the `` hubble '' scatter of the cepheid distances suggests that @xmath53 ( cf .section 2.1.3 ) . of course , the dynamical distances of very nearby galaxies ( @xmath54 or @xmath55mpc ) are highly unreliable , and they require `` honest '' distance determinations .this picture is over - optimistic for galaxies in high - density regions ( like the virgo region ) because the peculiar motions are larger here .some investigations depend just on a comparison of dynamical distances and `` honest '' distance determinations , e.g. if the infall into the virgo cluster should be studied , or generally if the local and not so local expansion field should be mapped , which would result , among other things , in a delineation of the co - moving volume which partakes in the local motion towards the microwave background . just for this purpose the data in fig .[ fig:3 ] are insufficient because there are only four sneia in the most relevant range @xmath56 and not a single sneia in the range @xmath57 ! for the velocity mapping _relative _ tf and d@xmath58 distances of clusters can close the gap .giovanelli et al .( 1997 ) and dale et al .( 1999 ) have derived tf distances of 71 clusters out to @xmath59 .they define a tight hubble line with a scatter ( @xmath60 ) rivalling sneia .but they do not yield an independent value of @xmath45 ; this is because the individual cluster distances are the mean of about ten subjectively selected cluster members .they are far from constituting a complete sample , which is the _ conditio sine qua non _ for any application of equation ( [ eq:14 ] ) .also the relative d@xmath61 distances of ten clusters by kelson et al .( 2000 ) can not independently be put on an absolute scale for the same reason ( cf .tammann 2001 ) . in the preceding pagesonly random statistical errors were quoted , although systematic errors are probably prevailing .most serious are of course systematic errors of the p - l relation of cepheids because the local expansion field on the left side of fig .[ fig:3 ] as well as the large - scale expansion field on the right side depend on the goodness of the calibrating cepheid distances .so far all cepheid distances are based on equations ( [ eq:2 ] ) and ( [ eq:3 ] ) .if equations ( [ eq:10 ] ) and ( [ eq:11 ] ) were used instead they would be reduced by @xmath62 or 6% ( equation [ eq:12 ] ) on average , assuming a median period of @xmath63 for the cepheids involved . yetthe new p - l relations are only defined up to @xmath63 , the ogle images being saturated for lmc cepheids with longer periods . identification and photometry of the rare cepheids with long periods are therefore still needed. the distance modulus @xmath34 ( table [ tab:1 ] ) , which defines the zero point of the p - l relations , has a small statistical error , but a systematic error of about @xmath64 can not be excluded .the whole extragalactic distance scale would benefit from a firm lmc modulus .no corrections for the dependence of cepheid luminosities on metallicity have been applied , because even the sign of these corrections is not well known ( cf .tammann et al .model calculations ( chiosi et al . 1993 ; saio & gautschy 1998 ; sandage et al . 1999 ; alibert & baraffe 2000 ) as well as observations ( kennicutt et al . 1998 ; tammann et al . 2001 ) indicate that the effect is smaller than @xmath65[fe / h ] , yet a dependence of this size would still decrease / increase the known cepheid distances by @xmath66 ( 4% ) on average .independent distance information of cepheids of different metallicity are therefore important .the finite width of the instability strip in the color - magnitude diagram causes an _ intrinsic _ scatter of the p - l relation which decreases with increasing wavelength ( see the scatter in fig .[ fig:1 ] ) .a p - l relation gives therefore the absolute magnitude and distance of a single cepheid to only @xmath67 .a p - l - color ( p - l - c ) relation of the form @xmath68 has repeatedly been proposed to position a cepheid within the instability strip .however , the pulsation models of saio & gautschy ( 1998 ) have shown that the constant - period lines in the instability strip change slope in function of mass , and hence the coefficient @xmath69 becomes a function of period .it has also been proposed to use the amplitude to locate a cepheid within the strip ( sandage & tammann 1971 ) .lack of sufficient data has brought these attempts to a preliminary halt . as statedbefore the calibration of the terms @xmath70 and @xmath69 in equation ( [ eq:13 ] ) is difficult , if not to say controversial .particularly statistical parallaxes of galactic rr lyrae stars , which strongly depend on the sample definition , have confused the issue .but there remains a tendency of rrlyr s although sometimes exaggerated to yield a smaller lmc modulus than cepheids ( cf .table [ tab:1 ] ) .there is still the possibility that an additional parameter influences the rrlyr luminosities .the luminosity calibration of the @xmath20-band trgb rests so far only on a few adopted globular cluster distances , or otherwise has to rely on the cepheid distances of external galaxies .consequently the insensitivity of the trgb against metallicity is not well tested yet .a definitive calibration would make the trgb a powerful tool requiring relatively little observing time .the large - scale value of @xmath45 depends entirely on the absolute magnitudes of sneia ( equation [ eq:17 ] ) , whose goodness in turn depends on nine calibrating cepheid distances .any improvement of the p - l relation of cepheids reflects therefore directly on @xmath45(cosmic ) . additional sneia at intermediate distances are needed to tighten the correlation of their absolute magnitudes @xmath71 with their decline rate @xmath7 and intrinsic color @xmath8 .allowance for this correlation only develops the full power of sneia as standard candles , i.e. it reduces the observed scatter of their hubble diagram from @xmath72 to @xmath73 .but the correction of @xmath71(observed ) for second parameters has also a systematic effect on @xmath45 .this is because the nine calibrating sneia lie necessarily in late - type spirals ( to contain cepheids ) , and sneia in spirals tend to be brighter and bluer and to have smaller decline rates than average ( cf .parodi et al .gaia will measure the trigonometric parallaxes of countless stars in lmc and smc as well as in umi and in at least seven other nearby dwarf galaxies .this will not only provide fundamental distances of these galaxies but also fix their orientation in space .gaia will in addition provide rotational parallaxes for some nearby galaxies like m31 and m33 and marginally even for galaxies outside the local group like ngc55 , ngc247 , ngc253 , ngc300 , m81 and ngc2403 , by comparing their rotation curves observed from radial velocities ( probably best observed from the ground ) and from proper motions .the direct distance measurements of nearby galaxies important for themselves will offer a control of the distance indicators discussed in the next section .gaia will determine trigonometric parallaxes of thousands of galactic cepheids , rr lyrae stars , and of stars at the trgb .it will also provide highly reliable zams positions in function of metallicity , and the position of the cooling sequence of white dwarfs ( wd ) . moreover, gaia will measure the parallaxes of most of the 140 galactic globular clusters ( gc ) to better than 1% .their luminosity function ( gclf ) may in turn become valuable for an extragalactic distance scale based on only old objects ( cf .finally gaia will decide to what extent the p - l relation of mira variables can be made useful for distance determinations .the above distance indicators are interlocked in a multiple way ( fig .[ fig:4 ] ) .rr lyrae and trgb stars as well as zams and wd fitting will yield distances of gcs whose trigonometric parallaxes will be known .distances to lmc , smc , and other nearby galaxies with known trigonometric and some rotational parallaxes will be measured again by cepheids and rr lyrae and trgb stars .this network of independent distance determinations will clarify the metallicity dependence of each method .any discrepancies which may occur will lead to new astrophysical insights .( , ) ( 0,0 ) ( 0,0)(,0 ) ( 0,0)(0 , ) ( 0,)(,0 ) ( , 0)(0 , ) ( , ) ( , ) ( 0,0)[c ] ( , ) ( , ) ( 0,0)[c ] ( , ) ( , ) ( 0,0)[c ] ( , ) ( , ) ( 0,0)[c ] ( , ) ( , ) ( 0,0)[c ] ( , ) ( , ) ( 0,0)[c ] ( , ) ( , ) ( 0,0)[c ] ( 50,50)(0,0)23 ( 50,50)(0,-1)23 ( 80,50)(0,0)3 ( 80,50)(0,-1)3 ( 28,28.5)(4,1)45 ( 28,28.5)(-4,-1)1 ( 28,51.5)(4,-1)45 ( 28,51.5)(-4,1)1 ( 24.5,64.5)(2,-3)25 ( 24.5,64.5)(-2,3)0 ( 27,71.7)(3,-2)46 ( 27,71.7)(-3,2)0 ( 73,59.5)(-2,-3)21.7 ( 73,59.5)(2,3)0 gaia offers a unique opportunity to discover `` nearby '' sneia ( @xmath74 ) .each snia with good photometry yields the chance of the century to determine the distance of its host galaxy to within @xmath73 ( 5% ) .no other distance indicator can give the same accuracy , except galaxies with @xmath75 for which the best distance is obtained from its recession velocity and by assuming a ( correct ) value of @xmath45 ( cf .section 3 ) . therefore sneia with @xmath76are of particular interest . out to this limit their number can be estimated from their frequency per unit luminosity and from the luminosity density summed over all galaxies ( both scaled to [ for instance ] @xmath77 ) .the former is @xmath78 sneia per @xmath79 solar luminosities ( @xmath80 ) and per 100 years , independent of the galaxy type ( tammann et al .1994 ) ; the total galaxian luminosity density is @xmath81mpc@xmath82 ( tammann 1982 ) .the volume @xmath83 being @xmath84mpc@xmath85 yields then 555 sneia per year with an error of probably less than a factor 2 .all of them have @xmath86 , where @xmath87 for sneia _ before _ correction for decline rate and color .half of these sneia will be caught by gaia , which covers every point in the sky once per month , near maximum , i.e. earlier than 10 days after maximum . at this timea snia has faded to @xmath88 assuming the standard light curve of leibundgut ( 1991 ) , and if it is subsequently followed for 30 days , even the faintest snia of the sample will have @xmath89 .this follow - up is sufficient to extrapolate @xmath90 and to determine the decline rate @xmath7 and @xmath91 with sufficient accuracy even for sneia whose maxima have been missed by 10 days .the resulting @xmath90 corrected for @xmath7 and @xmath91 will yield , if combined with @xmath92 in equation ( [ eq:17 ] ) , unrivalled distances for 1100 galaxies during the 4-year operation time of gaia .@xmath93 this number is overoptimistic because of galactic absorption .roughly one third of the sneia will be at low galactic latitudes , i.e. @xmath94 and be dimmed by @xmath95. they are either missed , or their photometry is impaired by large absorption corrections .sneia at higher @xmath96 will suffer an absorption of @xmath97 on average ( cf.schlegel et al .1998 ) . a corresponding margin of the detection and follow - up limits should therefore be allowed for . moreover, many day time sneia will not be suitable for follow - up from the ground .nevertheless the benefit of this programme is fourfold : \(1 ) the precision distances of these galaxies will map the not so local velocity field and explain the origin of the cmb dipole in great detail ; \(2 ) the residuals of the new sneia from the mean hubble line will fix the dependencies of @xmath90 on @xmath7 and @xmath91 with any desiderable accuracy .these dependencies rest now only on 35 sneia and their errors can still cause systematic errors of all sneia distances of 3 - 4% ( cf .parodi et al .2000 ) .\(3 ) the new sneia will richly populate the gap of the hubble diagram ( right side of fig .[ fig:3 ] ) and fix the position of the hubble line to within a vanishingly small error , reducing still further the random error of the constant term @xmath98 in equation ( [ eq:16 ] ) and correspondingly also of @xmath45 .moreover , the position of the hubble line at @xmath99 is essential when the cosmological constant @xmath100 shall be derived from the hubble diagram of very distant sneia .\(4 ) the new sneia , for which one will have reliable survey times , are ideally suited for the determination of their frequency with unprecedented accuracy .gaia can not provide the follow - up photometry of the newly discovered sneia .the above sneia program is therefore only meaningful if it will be possible to organize a dedicated ( earth - bound ) program for the photometry in @xmath24 , @xmath19 ( and @xmath20 ) of the new sneia down to @xmath101 . if one wants to carry the gaia search for sneia to still fainter limits their total number per year increases as @xmath102 ( cf .hg et al .. the sn sample provided by gaia will be little `` contaminated '' by sneii / ib because they are fainter by @xmath103 on average than sneia . therefore all magnitude - limited sn searches are strongly biased against sneii / ib in spite of their considerably higher frequency per unit volume .their value for cosmology is much smaller because their luminosity function is broad ( @xmath104 ; tammann & schrder 1990 ) and even skewed towards fainter sneii ( woltjer 1998 ) .they can hence not be used as standard candles .the determination of their _ individual _ distances from the expanding photosphere method ( epm ) poses formidable problems due to the compton scattering in _ moving _ atmospheres .available results may therefore still be affected by systematic effects ( cf .hamuy et al.2001 and references therein ) .gaia s high resolution power will reveal a large number of gravitationally lensed double or multiple quasars . in principlethey offer themselves for an independent distance determination , but the solution is typically degenerate as to the distance _ and _ the lensing mass .thus a single lensed quasar leaves a wide margin for @xmath105 ( saha & williams 2001 ) .a _ combined _solution for six quasars gives @xmath106 ( saha 2002 ) .this value can be much tightened when additional cases will be discovered .reliable distance indicators ( cepheids , tf distances of complete samples , and nearby cluster distances ) require a _ local _ value of @xmath107 .cepheid - calibrated sneia give a _large - scale _ value of @xmath108 .the two values are statistically indistinguishable and one may assume @xmath109 everywhere .the quoted error is only the statistical error .but the solution for @xmath45 is dominated by _ systematic _errors which could amount to as much as 10% .the largest systematic errors are introduced by cepheids , which form the basis of the local as well as of the large - scale expansion rate .equation ( [ eq:12 ] ) suggests , in case of a non - linear form of the p - l relation of cepheids , that @xmath45 is increased by 6% for cepheids with a median period of 30 days , i.e. a typical period for extra - local group galaxies .a metallicity dependence of the cepheid p - l relation could affect @xmath45 by @xmath110 ; if the metallicity correction of kennicutt et al .( 1998 ) is taken at face value the effect would go in the direction of decreasing @xmath45 .the zero point of the p - l relation seems to be well determined ( table [ tab:1 ] ) , but a systematic error of 3 - 4% can not be excluded . the next important source of a systematic error at least for the large - scale value of @xmath45 comes from the homogenization of sneia as to decline rate @xmath7 and color @xmath8 .in fact errors of the slope of the @xmath111 and @xmath112 relations could introduce systematic errors of the sneia distances of @xmath113 ( parodi et al .2000 ) . this problem can entirely be solved if good photometry will be obtained for the many hundreds of sneia to be discovered within @xmath9 by gaia .this at the same time will yield irreplaceable distances to an equal number of field galaxies ( to within random errors of @xmath114 ) , which will outline any concerted deviations from pure hubble flow out to @xmath9 .the one remaining systematic error of @xmath115 comes from the difficult photometry of extra - local group cepheids with the wide - field camera ( wfpc-2 ) of hst .gaia can not offer a handle on this problem , and it has to await future photometry from space .gaia will not only much improve the value of @xmath45 , but also the age of the oldest objects .definitive distances to globular clusters will reduce the error of their ages , and the improvement of the distances of local group galaxies _ and _ the measurement of their proper motions are very important for the determination of the dynamical age of the local group ( lynden - bell 1999 ) .thus @xmath45 and a minimum age of the universe can be used as strong priors for the cmb fluctuation spectrum ( cf .netterfield et al .2002 ; pryke 2001 ) which will narrow down the possible range of other cosmological parameters like the baryon density @xmath116 , the matter density @xmath117 , and the cosmological constant @xmath100 ._ note added in proof ( sep . 5 , 2002 ) ._ it has become clear now that the p - l relation of galactic fundamental - mode cepheids is distinctly different from that in lmc .the galactic relation is quite steep , the short - period cepheids being fainter than in lmc and the long - period ones being brighter ( tammann , sandage , & reindl , in preparation ) .the non - uniqueness of the p - l relations which is probably due , at least in part , to metallicity differences greatly increases the need for fundamental distances of nearby galaxies from gaia .alibert , y. , & baraffe , i. 2000 , in : the impact of large - scale surveys on pulsating star research , eds .l. szabados and d. kurtz , asp conf .* 203 * , p. 250 .baraffe , i. , & alibert , y. 2001 , _ a&a _ * 371 * , 592 . branch , d. , fisher , a. , & nugent , p. 1993 , _ aj _ * 106 * , 2383 .carroll , s.m . , press , w.h . ,& turner , e.l .ara&a _ * 30 * , 499 .chaboyer , b. , demarque , p. , kernan , p.j . , & krauss , l.m .1998 , _ apj _ * 494 * , 96 .chiosi , c. , wood , p.r . , & capitanio , n. 1993 , _ apjs _ * 86 * , 541 .cioni , m .- r.l . , van der marel , r.p . , loup , c. & habing , h.j .a&a _ * 359 * , 601 .da costa , g.s . , & armandroff , t.e . 1990 , _aj _ * 100 * , 162 .dale , d.a . , giovanelli , r. , haynes , m.p . ,camusano , l.e . , & hardy , e. 1999 , _ aj _ * 118 * , 1489 .feast , m.w .1984 , _ mnras _ * 211 * , 51 .feast , m.w . 1999 , _pasp _ * 111 * , 775 . feast , m.w . , & catchpole , r.m .1997 , _ mnras _ * 286 * , l1 .federspiel , m. , tammann , g.a . , & sandage , a. 1998 , _ apj _ * 495 * , 115 .federspiel , m. 1999 , ph.d .thesis , univ . of basel .freedman , w.l ., et al . 2001 , _apj _ * 553 * , 47 .giovanelli , r. , et al .1997 , _ aj _ * 113 * , 22 .girardi , l. , & salaris , m. 2001 , _ mnras _ * 323 * , 109 .groenewegen , m.a.t .& oudmaijer , r.d .a&a _ * 356 * , 849 .groenewegen , m.a.t . & salaris , r.d .a&a _ * 366 * , 752 .hamuy , m. , et al .2001 , _ apj _ * 558 * , 615 .hendry , m. 2001 , in : a new era in cosmology , eds . t. shanks & n. metcalfe asp conf ., in press .hg , e. , fabricius , c. , & makarov , v.v . 1999 ,_ baltic astronomy _ * 8 * , 233 .kelson , d.d ., et al .2000 , _ apj _ * 529 * , 768 .kennicutt , r.c ., et al . 1998 , _ apj _ * 498 * , 181 .kovcs , g. 2000 , _ a&a _ * 363 * , 1 . kraan - korteweg , r.c .a&as _ * 66 * , 255 .leibundgut , b. 1991 , private communication .lynden - bell , d. 1999 , in : the stellar content of local group galaxies , eds .p. whitelock & r. cannon , iau symp .192 , p. 39 .madore , b. , & freedman , w.l .1991 , _ pasp _ * 103 * , 933 .madore , b. , & freedman , w.l .1998 , _ apj _ * 492 * , 110 .netterfield , c.b ., et al . 2002 , _ apj _ * 571 * , 604 .pont , f. 1999 , in : harmonizing cosmic distance scales in a post - hipparcos era , eds .d. egret & a. heck , asp conf . ser . *167 * , p. 113 .panagia , n. 1999 , in : new views of the magellanic clouds , eds . y .- h . chu , n. suntzeff , j. hesser , & d. bohlender , iau symp . * 190 * , p. 549 .parodi , b.r ., saha , a. , sandage , a. , & tammann , g.a .2000 , _ apj _ * 540 * , 634 .pryke , c. 2001 , in : a new era in cosmology , eds . t. shanks & n. metcalfe asp conf .ser . , in press .romaniello , m. , salaris , m. , cassisi , s. , & panagia , n. 2000 , _ apj _ * 530 * , 738 .saha , a. , sandage , a. , tammann , g.a . , dolphin , a.e . , christensen , j. , panagia , n. , & macchetto , f.d .2001 , _ apj _ * 562 * , 314 .saha , p. 2002 , talk presented at a colloquium , basel , 26.06.2002 .saha , p. , & williams , l.l.r .2001 , _ aj _ * 122 * , 585 .saio , h. , & gautschy , a. 1998 , _ apj _ * 498 * , 360 .sakai , s. , zwitsky , d. , & kennicutt , r.c .2000 , _ aj _ * 119 * , 1197 .sandage , a. 1993 , _ aj _ * 106 * , 719 .sandage , a. , bell , r.a . , & tripicco , m.j . 1999 , _ apj _ * 522 * , 250 .sandage , a. , & tammann , g.a .1971 , _ apj _ * 167 * , 293 .schlegel , d. , finkbeiner , d. , & davis , m. 1998 , _ apj _ * 500 * , 525 .tammann , g. a. 1982 , in : landolt - brnstein , astronomy & astrophysics , vol .w. schaifers & h.h .voigt ( berlin : springer ) , p. 346 .tammann , g.a .2001 , in : the century of space science , eds .j. blecker , j. gei , & m.c.e .huber , ( dordrecht : kluwer ) , in press .tammann , g.a . ,lffler , w. , & schrder , a. 1994 , _ apjs _ * 92 * , 487 .tammann , g.a . ,reindl , b. , thim , f. , saha , a. , & sandage , a. 2001 , in : a new era in cosmology , eds . t. shanks & n. metcalfe asp conf ., in press .tammann , g.a . , & sandage , a. 1999 , in : harmonizing cosmic distance scales in a post - hipparcos era , eds .d. egret & a. heck , asp conf . ser . * 167 * , p. 204 .tammann , g.a . , & schrder , a. 1990 , _ a&a _ * 236 * , 149 .theureau , g. , rauzy , s. , bottinelli , l. , & gouguenheim , l. 1998 , _ a&a _ * 340 * , 21 .udalski , a. , et al .1999 , _ aca _ * 49 * , 201 .van leeuwen , f. , et al .1997 , _ mnras _ * 287 * , 955 .walker , a.r .1992 , _ apj _ * 390 * , l81 .walker , a. 1999 , in : post - hipparcos cosmic candles , eds .a. heck and f. caputo , ( dordrecht : kluwer ) , p. 125 .woltjer , l. 1998 , in : supernovae and cosmology , eds . l. labhardt , b. binggeli and r. buser ( basel : univ .basel ) , p. 87

<S> the local expansion field ( @xmath0 ) _ and _ the cosmic expansion field out to @xmath1 are characterized by @xmath2[kms@xmath3mpc@xmath3 ] . while the random error of this determination is small ( @xmath5 units ) , it may still be affected by systematic errors as large as @xmath6 . </S><S> the local expansion is outlined by cepheids and by cepheid - calibrated tf distances of a complete sample of field galaxies and by nearby groups and clusters ; the cosmic expansion is defined by cepheid - calibrated sneia . </S><S> the main source of systematic errors are therefore the shape and the zero point of the p - l relation of cepheids and its possible dependence on metallicity . </S><S> gaia will essentially eliminate these systematic error sources . </S><S> another source of systematic error is due to the homogenization of sneia as to decline rate @xmath7 and color @xmath8 . </S><S> gaia will discover about half of the 2200 sneia which will occur during a four - year lifetime within @xmath9 . for many of them </S><S> ground - based follow - up will provide useful photometric parameters ( and spectra ) , which will allow to fix the dependence of the sneia luminosity on @xmath7 and @xmath8 with high accuracy . at the same time they will yield exquisite distances to a corresponding number of field galaxies . </S><S> gaia will also revolutionize the very local distance scale by determining fundamental distances of the companion galaxies of the milky way and even of some spirals in- and possibly outside the local group from their rotation curves seen in radial velocities and proper motions . moreover , gaia will obtain trigonometric parallaxes of rr lyrae stars , of red giants defining the trgb , of stars on the zams , of white dwarf defining their cooling sequence , and of globular clusters , and determine the metallicity dependence of these distance indicators . </S><S> it will thus establish a self - controlling network of distance indicators within the local group and beyond . </S>

0706.2627

within the last few years , it has become more and more obvious that the milky way stellar halo is dominated by substructure , particularly dwarf galaxies , clouds , and tidal tails .data from the sloan digital sky survey ( sdss ; * ? ? ?* ) have revealed abundant examples of streams and substructure .for example , @xcite used a simple colour cut @xmath2 to map out the distribution of stars in sdss data release 5 ( dr5 ) .the `` field of streams '' , an rgb - composite image composed of magnitude slices of the stellar density of these stars , showed the overlap of the leading and trailing arm of the well - known sagittarius stream and the monoceros ring very clearly .also prominent was a new stream , which did not have an identified progenitor , and was called the `` orphan stream '' by @xcite .the observational data on the orphan stream @xcite was used by @xcite to argue that its progenitor may be the newly - discovered disrupting dwarf galaxy uma ii @xcite .tidal tails have proved to be an important diagnostic of the galactic potential .especially the tails of the dissolving sagittarius dwarf galaxy ( see e.g. * ? ? ?* ; * ? ? ?* ; * ? ? ?* ; * ? ? ?* ) , which wrap around the milky way , are an excellent tracer of the strength and shape of the potential .@xcite have shown with their numerical models that the bifurcation of the sagittarius stream as seen in the `` field of streams '' is composed of two wraps of the tidal tails and can only be reproduced if the orbital precession is small , i.e. if the milky way dark matter halo is close to spherical .extra - tidal extensions and onsets of tidal tails have been claimed around a number of galactic globular clusters in recent years ( see e.g. * ? ? ?the most spectacular and convincing discovery remains the long and thin tail from the disrupting globular cluster pal 5 @xcite .the tails extend at least @xmath3 kpc from the cluster in the leading and trailing direction and contain more mass than the remaining cluster .recently , two different groups @xcite claim to have detected tidal tails of various extents around the disrupting globular cluster ngc 5466 .this is an old , metal - poor ( [ fe / h]@xmath4 ) cluster , lying at galactic coordinates @xmath5 , @xmath6 . in @xcite ,the observed tails of ngc 5466 are not as long as those of pal 5 , stretching about @xmath7 or @xmath8 pc in either direction .@xcite reported afterwards that they found evidence for a much larger extension of the tidal tails of ngc 5466 .they claimed that the leading arm extends over @xmath9 degrees and the trailing arm extends at least @xmath10 degrees , before it leaves the area covered by sdss .this finding makes the tails of ngc 5466 even longer , but much fainter , than the tails of pal 5 .the aim of our paper is to confront these claims with theoretical expectation , as well as to study the survival of the tails .the following data for ngc 5466 are taken from various sources in the literature @xcite .the central surface brightness is @xmath11 magarcsec@xmath12 .the total luminosity is @xmath13 mag and the mass - to - light ratio as given by @xcite is @xmath14 .using these values , we derive a total mass of about @xmath15 m@xmath16 .the core radius has values ranging from @xmath17 to @xmath18 pc ; the half - mass radius ranges from @xmath19 to @xmath20 pc .the most substantial differences in the literature occur for the tidal radius . here, values are spread between @xmath21 and @xmath22 pc .we use this data as constraints on our numerical simulations .our aim is to access possible initial models of this globular cluster and analyse its stability and evolution in different sets of milky way potentials . in the next section, we describe the setup of our simulations namely , the choice of galactic potential models , the orbit of ngc 5466 and finally the initial model of the cluster itself .this is followed by a study of the relative importance of two - body relaxation and disc shocking in section [ sec : sbox ] , justifying our use of particle - mesh simulations in this paper .then , in section [ sec : res ] , we present simulations that reproduce the shape , extent and surface density of the tidal tails detected by @xcite and @xcite .the properties of the remnant and shown to correspond to the present - day ngc 5466 . finally , we examine how the tidal tails change as a function of the proper motion and hence orbit .-axis ) of the ml ( dotted , green ) and db ( long - dashed , blue ) models .both potentials have the same circular velocity ( @xmath23 kms@xmath24 ) at the solar radius .right : gravitational potential in the disc plane for the ml ( dashed ) and db ( solid ) models .also shown is the gravitational potential along the @xmath25-axis at the solar radius for the ml ( dotted ) and db ( long - dashed ) models .even though the potentials agree well in the innermost part , the ml potential is much steeper in the outer parts.,title="fig:",height=151 ] -axis ) of the ml ( dotted , green ) and db ( long - dashed , blue ) models .both potentials have the same circular velocity ( @xmath23 kms@xmath24 ) at the solar radius .right : gravitational potential in the disc plane for the ml ( dashed ) and db ( solid ) models .also shown is the gravitational potential along the @xmath25-axis at the solar radius for the ml ( dotted ) and db ( long - dashed ) models .even though the potentials agree well in the innermost part , the ml potential is much steeper in the outer parts.,title="fig:",height=151 ] dynamical friction does not play a significant role in the evolution of a low - mass star cluster .so , we are able to model the galactic tidal field as a smooth and analytic background potential . for the galactic potential ,we use one of two standard models .the first ( hereafter ml from ( m)iamoto - nagai + ( l)ogarithmic halo ) is a superposition of three components .the halo is represented by a spherical logarithmic potential of the form @xmath26 with @xmath27 kms@xmath24 and @xmath28 kpc ( and @xmath29 is the spherical radius ) .the galactic disc is modelled by a miyamoto - nagai potential : @xmath30 with @xmath31 m@xmath16 , @xmath32 kpc and @xmath33 kpc ( where @xmath34 and @xmath25 are cylindrical coordinates ) .the bulge is represented by a hernquist potential @xmath35 using @xmath36 m@xmath16 and @xmath37 kpc . for comparison, we also use the galactic potential suggested by @xcite and hereafter denoted by db .it consists of three disc components , namely the ism , the thin and the thick disc , each of the form @xmath38 with @xmath39 , eq .( [ eq : dehndisc ] ) describes a standard double exponential disc with scale - length @xmath40 , scale - height @xmath41 and central surface - density @xmath42 . for the stellar discs , @xmath43is set to be zero , while for the ism - disc , we allow for a central depression by setting @xmath44 kpc @xcite .in addition to the the disc potential , we use the analytic potential corresponding to two spheroidal density distributions for the halo and the bulge in the form @xmath45 where @xmath46 we choose the parameters of our db model according to the best - fit model 4 in the paper of @xcite . in fig .[ fig : potcomp ] , we compare the two potentials . for both, the circular velocity at the solar radius is @xmath23 kms@xmath24 .however , the ml model contains more mass within a given radius than the db model . as a initial model for the star cluster ,we choose a @xcite sphere : @xmath47 with @xmath48 being the scale - length of the plummer sphere , which is identical to the half - light radius , and @xmath49 the total mass .this is a fairly good representation of a star cluster , especially a young one .due to tidal shaping and internal evolution at later stages , a @xcite model usually fits the photometric data better .the advantage of a plummer model is that all physical quantities are analytically accessible .the initial plummer model has a half - light radius of @xmath50 pc , an initial mass of @xmath51 m@xmath16 and is represented by @xmath52 particles .the numerical set - up of the particles is performed using the algorithm of @xcite .we checked that our initial model is able to survive for a hubble time by comparing our initial configuration with the dissolution times given in @xcite ( see their fig .as the orbit of ngc 5466 is most of the time located far out in the halo , it is well represented by the uppermost lines in @xcite , giving us a dissolution time of a few hubble times .kpc.,height=302 ] to determine the orbit of ngc 5466 , we use the positions and proper motion from the literature @xcite , namely : @xmath53 its galactocentric distance is @xmath54 kpc .we transform the positions and velocities into a galactocentric cartesian coordinate system and integrate a test particle back in time for @xmath50 gyr .this endpoint of the backward - integration is then the starting position of our initial model .even though the two potentials are quite similar in the innermost parts , the orbits differ in terms of perigalacticon , apogalacticon and number of disc crossings . in the db potential ,the galaxy is less massive in the outer parts , so the cluster can reach an apogalacticon of @xmath55 kpc , while in the ml case , it only reaches @xmath56 kpc .the perigalactica are @xmath57 and @xmath58 kpc , respectively . in fig .[ fig : orbit ] , the shape of the orbits in the @xmath34,@xmath25-plane is plotted ( for the db model , we flipped the radial coordinate onto the negative side to aid visibility ) . and@xmath59 times @xmath60 m@xmath16 , respectively for clarity .the middle curve ( green ) shows the result of the nbody 4 simulation neglecting stellar evolution and lowest curve ( blue ) shows the result of the nbody 4 run with stellar evolution .shown are also the linear fitting lines to the nbody 4 results ( dashed lines ) . ] ) .the ( green ) curve together with the right ordinate shows the evolution of the bound mass .it is visible in this figure that the mass - loss ceases after perigalacticon and turns into a steep mass - loss at the disc crossing .but as shown in the in - set ( vertical lines are the same as in the main figure ) the total mass - loss at this combination of perigalacticon and disc passage is @xmath58 times larger than the steep mass - loss caused by the actual disc passage . ]to simulate the evolution of the tails of ngc 5466 , we use the particle - mesh _ superbox _ package @xcite .a particle - mesh code has the great advantage that we can use millions of particles ( which represent equal - mass phase - space elements rather than single stars ) and trace the faint tails very accurately .however , such a code is often not suitable for simulations of globular clusters , because it neglects the internal evolution due to two - body relaxation completely .the reason why _ superbox _ is nonetheless a valid method for the modelling of ngc 5466 is understood on examining the @xmath61 parameter @xcite : @xmath62 here , @xmath63 denotes the relaxation time - scale , which amounts to @xmath64 gyr for our initial model and to @xmath65 gyr for the present state of the globular cluster .additionally , @xmath66 denotes the disc shock time - scale , which is the time - scale on which the cluster is destroyed by disc shocks . using the formula from @xcite, we have @xmath67 where @xmath68 is the period of the disc crossings , @xmath69 is the velocity with which the object crosses the disc , @xmath70 denotes the ratio of velocity dispersion to half - mass radius @xmath71 of the object and finally @xmath72 is the acceleration perpendicular to the disc . using our simulation data , we derive a disc shock time - scale of about @xmath73 gyr .this gives a @xmath74 , which holds for both galaxy models within the errors .the concentration @xmath75 of our initial model and the star cluster today is in the order of unity .if we now place our initial model in fig .13 of @xcite , we see that it falls in the regime where shocks are more important than internal evolution , but also in the regime where the star cluster survives for a hubble time . still , the location of our model is close to the border - line ( at @xmath76 it is @xmath77 ) where internal evolution becomes dominant .it is interesting to compare ngc 5466 with the well - studied case of pal 5 , which has @xmath78 and @xmath79 .pal 5 will most likely be destroyed at its next disc crossing @xcite .by contrast , ngc 5466 has a good chance of surviving even for the next hubble time ! to demonstrate that the internal evolution has no major effect , we perform two n - body simulations with @xmath80 particles in the ml potential using nbody4 @xcite and compare to the _ superbox _ results . in the first simulation , we use an equal mass for all particles and neglected stellar evolution . in the second simulation, we adopt a mass function which is present after the initial phase of violent mass - loss caused by the evolution of high mass stars ( first few tens of myr ) . in practice , this means another @xmath81 % has to be added to the initial mass to account for the mass - loss due to supernovae , and stellar winds , as well as the stars which become unbound due to this mass - loss . for the remaining stars , stellar evolution in nbody 4is switched on .figure [ fig : nbody ] shows , by extrapolating the mass - loss in the direct n - body simulations linearly , that the additional mass - loss due to two - body relaxation and stellar evolution amounts at the very most to about one - third of the total mass - loss .the linear extrapolation of the mass - loss in this mass regime is justified by appeal to the work of @xcite . in other words , disregarding the initial mass - loss when the star cluster blows away its gaseous envelope ( which depends mainly on the star formation efficiency ) and the violent stellar evolution in the first few tens of myr , the dominant cause of mass - loss during the long - term evolution of ngc 5466 is the tidal field of the mw .having established that internal disruptive processes ( e.g. two - body relaxation ) are of minor importance , we also have to prove if disc shocks are the major external destruction process .we therefore performed a particle - mesh simulation with a halo - only potential .the mass - loss in this case amounts to @xmath82 % of the initial mass only ( see fig .[ fig : mass ] right panel ) .this is a factor of @xmath58 less than in the combined potential case .but this is not yet a genuine proof of the importance of disc shocks .as one can see in figs .[ fig : nbody ] and [ fig : mass ] the mass - loss happens in short time - intervals like a step function . in fig .[ fig : perigal ] we blow up one of these short time intervals . the first ( black ) vertical line shows the time of perigalacticon while the second ( red ) line shows the time when the cluster crosses the disc ( @xmath83 ) .while the mass - loss due to the tidal field ceases after perigalacticon there is an additional steep mass - loss starting when the cluster passes the centre of the disc . but shown in the small in - set in fig .[ fig : perigal ] this mass - loss is about @xmath84 of the total mass - loss at the combined perigalacticon and disc passage .the general conclusion is therefore that it is definitely the tidal field of the disc ( the perigalacticon is well outside the bulge region ) which causes the major contribution of the mass - loss , the actual disc shock when the cluster passes through the centre of the disc may be not that important .this finding explains the rather large time - scale for disc shocks ( @xmath73 gyr ) of the previous section .one of the advantages of _ superbox _ is that it has high resolution sub - grids , which stay focused on the simulated objects and travel with them through the simulation area .this is important in studying the morphology of the tenuous and diffuse tidal tails . within the innermost grid, we resolve the globular cluster at a resolution of @xmath85 pc .the grid with medium resolution is chosen to resolve the tidal tails close to ngc 5466 with a resolution of @xmath86 pc . in fig .[ fig : mass ] ( first two panels ) , we plot the bound mass of our models in two galactic potentials ( db and ml ) . in both cases ,the mass - loss is strong in the first @xmath87@xmath82 gyr and then tends to level off during the later stages of the evolution .the mass - loss is mainly related to each disc crossing near perigalacticon and the mass stays almost constant during the rest of the orbit .this is the major reason why in the db potential the mass - loss over @xmath50 gyr of evolution is less than in the ml potential the star cluster has had fewer disc crossings . in the ml potential, the cluster loses about @xmath88 % of its initial mass , whilst the cluster in the db potential only suffers a mass - loss of @xmath89 % ( with a fluctuation of only a few particles out of 1,000,000 ) .the mass - loss of the ml simulation is in very good agreement with the results found in previous studies by @xcite and @xcite .however , these mass - losses are only lower limits , as there will be a smaller , but not negligible , contribution from internal relaxation effects . in the top left panel of fig .[ fig : cont ] , we show the data on the tails of ngc 5466 reproduced from @xcite , who used neural networks are used to reconstruct the probability density distribution . the contours correspond to level curves of equal neural network output and therefore trace the star density .the tails are clearly visible once the extragalactic contaminants ( predominantly galaxy clusters ) have been eliminated .the tails extend @xmath90 on the sky , corresponding to @xmath14 kpc in projected length . [cols="^,^,^,^,^,^,^,^,^ " , ] in the next two panels of fig .[ fig : cont ] , we show the projection on the sky of our models in the ml and db potentials after @xmath50 gyr of evolution . both models show faint tidal tails , which match the general shape of the contours well .however , one important feature of the data is not reproduced the leading and trailing tails are well - aligned with the proper motion vector .this contrasts with the data , in which the inner parts of the leading tails are slightly below the proper motion vector , whilst the inner parts of the trailing tails are slightly above .however , the observed proper motions are not well determined and have large error - bars ,so one possibility is that the proper motion of ngc 5466 should be either larger in right ascension or smaller in declination than the values given in the literature ( e.g. * ? ? ?* ) to match the observed misalignment .we confirm this result by running another simulation with slightly changed proper motions , namely @xmath91 this orbit gives a perigalacticon of @xmath92 kpc and an apogalacticon of @xmath93 kpc .although the change in proper motion does not make a significant difference to the mass - loss rate , as shown in the right panel of fig .[ fig : mass ] , it does improve the match with the location of the observed tidal streams much better , including the misalignment . in the lower right panels of fig .[ fig : cont ] , we see that the inner parts of the leading tails are now slightly below the old proper motion vector , whilst the inner parts of the trailing tails are now slightly above .[ fig : allsky ] shows all - sky views of the tidal tails , together with density profiles obtained by counting particles .the surface density of the tidal tails falls off along the innermost tails very steeply and stays at a very low density of @xmath81@xmath94 m@xmath16deg@xmath12 throughout the tails .these low densities are very hard to detect , even in surveys like sdss .@xcite found long , almost linear and very tenuous tidal extensions to ngc 5466 using a matched filter .although these extensions are hard to see in the sdss data , they do receive some support from the simulations presented here .the tails of our model with the revised proper motion extend over @xmath95 on the sky .@xcite claim that the average density of the tails is about @xmath50@xmath81 stars deg@xmath12 , which is also in good agreement with out estimate .interestingly , at the point where @xcite start to loose track of the leading arm , our model is close to its apogalacticon and the tails are spread out much wider than is the case close to the cluster .although our simulation with the revised proper motion provides a good representation of the data , it is clearly not unique . in particular , it is interesting to understand the variety of tidal tail morphology for ngc 5466 , especially as forthcoming deeper photometry will provide stronger constraints on the modelling .accordingly , we perform a suite of _ superbox _ simulations to investigate how the choice of proper motion influences the mass - loss and hence the properties of the tidal tails . as a constraint ,we only used proper motions which are within the @xmath96 error range of the observed value @xcite and also require that the orbital path near the cluster aligns with the tails found by @xcite , i.e. , have the same projected orbital path as our refined set of proper motion .all - sky views of selected simulations are shown in fig .[ fig : pmdens ] and show significant differences in the morphology and the properties of the tails .table [ tab : pm ] gives the parameters and the results for the entire suite of simulations .the number of degrees in right ascension over which the tail is detectable represents a measure for the length of the tails .the mean density of the tails is calculated in the following way .we examine one degree in right ascension @xmath97 and search for the highest surface density in the tails for each degree in declination @xmath98 . from these values, we compute the average surface density over the range of right ascension for which the tails are present . the maximum density given in the tableis computed from the square degree of the tails with the highest surface density .effects of varying distances are not taken into account .table [ tab : pm ] shows clearly that the closer the orbit is to the galactic centre the more severe is the mass - loss and the higher is the density in the tails .pc@xmath12 , right ordinate shows magarcsec@xmath12 using the m / l - ratio from literature of @xmath99 .the inner part is still well fitted by the initial plummer profile , while the outer parts are better fitted by a steeper power - law with index @xmath100 .the deviation is visible in the region from tidal radius at perigalacticon to tidal radius at apogalacticon .vertical lines denote the tidal radius now ( solid ) and at the last perigalacticon ( dashed ) , while dotted lines mark the grid boundaries with changes in resolution .right : velocity dispersion profile .green open squares denote the 3d velocity dispersion measured in concentric shells around the cluster centre , crosses are the line - of - sight velocity dispersion measured in concentric rings around the cluster centre .curves show the profile of the initial model .vertical lines are as in the left panel.,title="fig:",height=154 ] pc@xmath12 , right ordinate shows magarcsec@xmath12 using the m / l - ratio from literature of @xmath99 . the inner part is still well fitted by the initial plummer profile , while the outer parts are better fitted by a steeper power - law with index @xmath100 .the deviation is visible in the region from tidal radius at perigalacticon to tidal radius at apogalacticon .vertical lines denote the tidal radius now ( solid ) and at the last perigalacticon ( dashed ) , while dotted lines mark the grid boundaries with changes in resolution .right : velocity dispersion profile .green open squares denote the 3d velocity dispersion measured in concentric shells around the cluster centre , crosses are the line - of - sight velocity dispersion measured in concentric rings around the cluster centre .curves show the profile of the initial model .vertical lines are as in the left panel.,title="fig:",height=154 ] let us consider the internal properties of our remnant cluster in a representative simulation .we choose the one which uses the ml potential and the revised proper motions , shown in the lower right panels of fig .[ fig : cont ] . for this simulation ,[ fig : internal ] shows the surface density and velocity dispersion profiles of the final cluster . adopting the data from @xcite ( updated values from 2003 ) , the cluster has a central surface brightness of @xmath11 magarcsec@xmath12 .this is in good agreement with our simulation , for which the central surface brightness is @xmath101magarcsec@xmath12 , especially taking into account that our particle - mesh code neglects internal evolution , which would lead to higher densities in the core .also , the half - light radius in our simulation is @xmath50 pc and corresponds well with the observed values of @xmath19 pc @xcite and @xmath20 pc @xcite .the actual tidal radius in our model is @xmath102 pc ( using the jacobi limit as given in * ? ? ?* ) and is less than the @xmath22 pc stated in @xcite ( @xmath22 pc ) or @xmath103 pc in @xcite , but only slightly larger than the radius of @xmath21 pc found by @xcite .note that , observationally speaking , @xmath104 is determined by fitting a king ( 1966 ) profile to the surface brightness distribution , which does not correspond exactly to the theoretical definition .according to our simulations , the tidal radius at the last perigalacticon was about @xmath105 pc .while the surface density in the inner parts is not much affected by the mass - loss , the central velocity dispersion is reduced by approximately @xmath50 % . also visibleis a rise in the line - of - sight velocity dispersion in the outer parts , which starts already within the actual tidal radius .this is due to the fact that all line - of - sight measurements are contaminated by unbound stars streaming in front or behind the star cluster . while they do not affect the central values because of their low number , their effect is easily measurable in the outer parts where the densities of the bound stars are much lowerwe have presented numerical simulations of the formation and evolution of the tidal tails of the globular cluster ngc 5466 .we used direct n - body codes to argue that the evolution of the cluster is dominated by external effects rather than internal relaxation , and then grid - based codes to trace the faint tidal tails .this novel , hybrid approach is well - suited to map out the detailed morphology of the low - density tails of ngc 5466 .naively , we might expect that a low mass cluster with observed and very lengthy tails on a disc crossing orbit would not be able to survive for too much longer .however , simulations by @xcite have already shown that the disrupting globular cluster pal 5 has survived for at least many gyr in a tidally - dominated and out - of - equilibrium state , although pal 5 probably will be destroyed at the next disc crossing . here, we have demonstrated that a progenitor cluster of ngc 5466 , which is quite similar to the present cluster , could survive substantially longer , for at least a few hubble times , with its extensive but tenuous tidal tails gradually wrapping around the whole galaxy .the evolution of ngc 5466 is mainly driven by tidal shocks at each perigalacticon / disc crossing combination .although not entirely negligible , internal effects ( two - body relaxation and evaporation of stars driven by post - core collapsed processes @xcite ) play a much less important role in the mass - loss .it is this property which allows us to study the tidal tails using grid - based codes rather than the more cumbersome direct n - body codes .if the observationally determined mass - to - light ratio of @xmath14 is correct , then the initial mass of ngc 5466 is @xmath106 m@xmath16 . by initial mass , we do not mean the embedded mass of the star cluster at its formation inside a gas - cloud .if the star formation efficiency is low , a star cluster can lose about @xmath107 per cent of its initial mass in stars when the gas gets blown out by high velocity winds or supernovae explosions .the rapid stellar evolution of high mass stars then adds another extreme mass - loss of @xmath108 per cent in the first few tens of myr . after this initial phase of rapid evolution , the cluster reaches a quasi - equilibrium .this is the starting point of our simulations and therefore our initial mass refers to this point in time .our numerical simulations reproduce the observational results of both groups who have recently studied the tidal tails ngc 5466 with sdss data .mapping out the tails close to the globular cluster , @xcite found that the leading tail emerges from the side pointing towards the galactic centre and returns to the orbital path from outside , while the trailing tail emerges from the side opposite to the galactic centre and returns to the orbital path from within . with our simulations , we showed that the proper motion of the globular cluster has to be smaller in declination and/or larger in right ascension than reported by @xcite to account for the position of the tidal tails .we propose a new set of proper motions , @xmath109 for which the tail morphology is correctly reproduced .this differs from the observationally determined one by @xmath110 and @xmath111 masyr@xmath24 respectively .these changes are within the error margins of the observed proper motion ( @xmath112 masyr@xmath24 ) .the surface density of the tidal tails falls off along the innermost tails very steeply and stays at a very low density of @xmath81@xmath94 m@xmath16deg@xmath12 throughout the tails .these low densities are very hard to detect , even in surveys like sdss .@xcite found long , almost linear and very tenuous tidal extensions to ngc 5466 using a matched filter approach .their work is supported by the simulations in this paper , which show that the very long ( @xmath113 ) , faint tidal tails are expected .the tails in our simulation have roughly the same surface density as found by @xcite . in the future , deeper photometry , radial velocities and thanks to the gaia and sim satellites proper motions of individual stars in the tidal tails may become available . mapping out the structure of the tails of globular clusters and dwarf galaxies will then provide powerful constraints on the galactic potential .this work , together with the observational papers of @xcite and @xcite , has shown that ngc 5466 is a prime target for such studies of cold streams .its tidal tails , though faint , are the longest so far claimed for any milky way globular cluster . + * acknowledgements : * mf and vb are funded by pparc .miw acknowledges support from a royal society university research fellowship .we thank w. dehnen for providing his galactic potential code and r.spurzem and h.m .lee for useful comments .the direct n - body simulations were performed on the grace supercomputer at ari - zah heidelberg funded by volkswagen stiftung i/80 041 - 043 and the state of baden - wrttemberg , using grape hardware .

<S> the study of substructure in the stellar halo of the milky way has made a lot of progress in recent years , especially with the advent of surveys like the sloan digital sky survey . here , we study the newly discovered tidal tails of the galactic globular cluster ngc 5466 . by means of numerical simulations , we reproduce the shape , direction and surface density of the tidal tails , as well as the structural and kinematical properties of the present - day ngc 5466 . although its tails are very extended in sdss data @xmath0 ) , ngc 5466 is only losing mass slowly at the present epoch and </S><S> so can survive for probably a further hubble time . </S><S> the tidal shaping through the milky way potential , especially the potential of the disc , is the dominant process in the slow dissolution of ngc 5466 accounting for @xmath1 of the mass loss over the course of its evolution . </S><S> the morphology of the tails provides a constraint on the proper motion the observationally determined proper motion has to be refined ( within the stated error - margin ) to match the location of the tidal tails . </S><S> [ firstpage ] galaxy : kinematics and dynamics galaxy : halo globular clusters : individual : ngc5466 methods : n - body simulations </S>

math0601658

the stabilization of nonlinear and nonautonomous control systems , and the construction of their lyapunov functions , are challenging problems that are of significant ongoing interest in the context of robustness analysis and controller design ; see @xcite and @xcite . one popular approach to guaranteeing stability of nonautonomous systemsis the _ averaging method _ in which exponential stability of an appropriate _ autonomous _ system implies exponential stability of the original dynamics when its time variation is sufficiently fast .see @xcite for related results .the preceding results were extended to more general rapidly time - varying systems of the form @xmath0 in @xcite , where uniform ( local ) exponential stability of ( [ m1 ] ) was proven for large values of the constant @xmath1 , assuming a suitable limiting dynamics @xmath2 for ( [ m1 ] ) is uniformly exponentially stable .( we specify the choice of @xmath3 in our main theorem and examples below . )this generalized a result from @xcite on a class of systems ( [ m1 ] ) satisfying certain periodicity or almost periodicity conditions .the main arguments of @xcite use ( partial ) averaging but do not lead to explicit lyapunov functions for ( [ m1 ] ) . in this work ,we pursue a very different approach . instead of averaging, we explicitly construct a family of lyapunov functions for ( [ m1 ] ) in terms of more readily available lyapunov functions for ( [ m2 ] ) , which we again assume is asymptotically stable . while @xcite assumes ( [ m2 ] )is uniformly locally exponentially stable , we consider the case where ( [ m2 ] ) is assumed uniformly globally asymptotically stable ( ugas ) , in which case our conclusion is that ( [ m1 ] ) is ugas ( but not necessarily exponentially stable ) when @xmath1 is sufficiently large . while global exponential and global asymptotic stabilities are equivalent for_ autonomous _ systems under a coordinate change in certain dimensions , the coordinate changes are not explicit and so do not lend themselves to explicit lyapunov function constructions ; see @xcite .the importance of the problem we consider lies in the ubiquity of rapidly time - varying systems in practical applications ( e.g. , suspended pendulums subject to vertical vibrations of small amplitude and high frequency , raleigh s and duffing s equations from ( * ? ? ?* chatper 10 ) , and systems arising in identification in @xcite and below ) and the essentialness of lyapunov functions in robustness analysis and controller design for these applications .in particular , we show that assumptions similar to those of ( * ? ? ?* theorem 3 ) imply that ( [ m1 ] ) is uniformly globally exponentially stable ; our lyapunov function constructions are new even in this particular case and our results are complementary to those of @xcite .the lyapunov functions we construct are also input - to - state stable ( iss ) or integral iss lyapunov functions for @xmath4 under appropriate conditions on @xmath5 and @xmath6 ; see remark [ gissue ] . in section [ dl ] , we provide the relevant definitions and lemmas . in section [ thm ] ,we present our main sufficient conditions for uniform global asymptotic and exponential stability of ( [ m1 ] ) , and for the stability of ( [ fastcontrol ] ) , in terms of limiting dynamics ( [ m2 ] ) .this theorem leads to explicit constructions of lyapunov functions for ( [ m1 ] ) and ( [ fastcontrol ] ) in terms of lyapunov functions for ( [ m2 ] ) , in theorem [ cor1 ] . in section [ altthmsec ], we provide an alternative lyapunov function construction theorem for ( [ m1 ] ) not involving any limiting dynamics .we prove our main results in sections [ proof ] and [ altthm ] .we illustrate our theorems in sections [ illu ] and [ altillus ] using a friction control model and other examples .we close in section [ concl ] by summarizing our findings .we study ( [ m1 ] ) ( which includes dynamics ( [ m2 ] ) with no @xmath7 dependence , as special cases ) in which we always assume @xmath5 is continuous in time @xmath8 , continuously differentiable ( @xmath9 ) in @xmath10 , null at @xmath11 meaning @xmath12 and _ forward complete _ , i.e. , for each @xmath1 , @xmath13 , and @xmath14 there is a unique trajectory @xmath15 for ( [ m1 ] ) ( depending in general on the constant @xmath1 ) that satisfies @xmath16 .we set @xmath17 and let @xmath18 denote the set of all integers .we say @xmath19 is _ of class @xmath20 _ and write @xmath21 provided @xmath22 a continuous function @xmath23 is _ positive definite _ provided it is zero only at zero .a positive definite function @xmath24 is _ of class @xmath25 _ ( written @xmath26 ) provided it is strictly increasing ; if in addition @xmath24 is unbounded , then we say that @xmath24 is _ of class @xmath27 _ and write @xmath28 .a continuous function @xmath29 is _ of class @xmath30 _ ( written @xmath31 ) provided ( a ) @xmath32 for all @xmath33 , ( b ) @xmath34 is nonincreasing for all @xmath35 , and ( c ) for each @xmath35 , @xmath36 as @xmath37 .a positive definite function @xmath24 is called @xmath38 provided @xmath39 as @xmath40 .we always assume there exists @xmath41 such that @xmath42 everywhere .we next define our stability properties for ( [ m2 ] ) .the same definitions apply for ( [ m1 ] ) for any choice of the constant @xmath1 .we call ( [ m2 ] ) _ uniformly globally asymptotically stable ( ugas ) _ provided there exists @xmath43 such that @xmath44 where @xmath45 is the usual euclidean norm and @xmath46 is the flow map for ( [ m2 ] ) .we call ( [ m2 ] ) _ uniformly globally exponentially stable ( uges ) _ provided there exist constants @xmath47 and @xmath48 such that ( [ gasdef ] ) is satisfied with the choice @xmath49 the converse lyapunov function theorem implies ( [ m2 ] ) is ugas if and only if it has a _ ( strict ) lyapunov function _ ,i.e. , a @xmath9 @xmath50 that admits @xmath51 and @xmath52 such that for all @xmath53 and @xmath54 , we have both ( l1 ) @xmath55 and ( l2 ) @xmath56 , where the subscripts on @xmath57 indicate partial gradients ; see @xcite . when ( [ m2 ] ) is uges , the proof of ( * ? ? ?* theorem 4.14 ) shows : [ k02 ] assume ( [ m2 ] ) satisfies the uges condition ( [ gasdef])-([betachoice ] ) for some constants @xmath47 and @xmath48 and that there exists @xmath58 such that @xmath59 for all @xmath60 and @xmath53 .then ( [ m2 ] ) admits a lyapunov function @xmath57 and constants @xmath61 such that latexmath:[\[\label{m5}\renewcommand{\arraystretch}{1.25 } \begin{array}{l } c_1|\xi|^2 \ ; \leq \ ; v(\xi , t ) \ ; \leq \ ; c_2|\xi|^2\ ; , \ ; \ ; \left|v_\xi(\xi , t)\right| \ ; \leq \ ; c_3 |\xi|\ ; , \\ { \rm and}\ ; \ ; \ ; v_t(\xi , t ) + v_\xi(\xi , t ) \bar{f}(\xi ,t ) \ ; \leq \ ; - @xmath53 and @xmath63 . [ compatt ] given @xmath26 , the dynamics ( [ m2 ] )is said to be _@xmath24-compatible _ provided it admits a lyapunov function @xmath64 and two constants @xmath65 , @xmath66 such that : * * * @xmath67 . * * @xmath68 . * * @xmath69 . [ compat]note the _ asymmetry _ in the bounds on @xmath70 and @xmath71 in @xmath72 . if ( [ m2 ] ) satisfies the assumptions of lemma [ k02 ] , then it is @xmath24-compatible with @xmath73 .however , by varying @xmath24 ( including cases where @xmath24 is bounded ) , one also finds a rich class of non - uges @xmath24-compatible dynamics ; see e.g. section [ exa1 ] below .we also consider the nonautonomous _ control system _ @xmath74 which we always assume is continuous in all variables and @xmath9 in @xmath75 with @xmath76 , and whose solution for a given control function @xmath77all measurable locally essentially bounded functions @xmath78 ) and given initial condition @xmath16 we denote by @xmath79 .we always assume ( [ nonauto ] ) is _ forward complete _ , i.e. , all trajectories @xmath80 so defined have domain @xmath81 .we next recall the input - to - state stable ( iss ) and integral input - to - state stable ( iiss ) properties from @xcite and @xcite .let @xmath82 denote the essential supremum of @xmath83 restricted to any interval @xmath84 .[ iss ] ( a ) we say that ( [ nonauto ] ) is _ iss _ provided there exist @xmath85 and @xmath43 for which @xmath86}\right)\ ] ] holds when @xmath87 , @xmath13 , and @xmath83 .if in addition @xmath88 has the form ( [ betachoice ] ) , then we say that ( [ nonauto ] ) is _ input - to - state exponentially stable ( ises)_. ( b ) we say that ( [ nonauto ] ) is _ iiss _ provided there exist @xmath89 and @xmath31 such that @xmath90 holds when @xmath87 , @xmath91 , and @xmath92 .a function @xmath93 is called _ uniformly positive definite _ provided @xmath94 is positive definite in which case we write @xmath95 .we call @xmath95 _ uniformly proper and positive definite _ provided there exist @xmath96 such that condition ( l1 ) above is satisfied in which case we write @xmath97 .the following lyapunov function notions agree with the usual iss and iiss lyapunov function definitions when ( [ nonauto ] ) is autonomous , because functions @xmath98 are invertible .note that @xmath99 in ( [ iissd ] ) _ need not _ be of class @xmath25 .[ corr ] let @xmath100 .( a ) we call @xmath57 an _ iss lyapunov function _ for ( [ nonauto ] ) provided there exist @xmath101 such that for all @xmath102 , @xmath60 , and @xmath103 : @xmath104 $ ] .( b ) we call @xmath57 an _ iiss lyapunov function_ for ( [ nonauto ] ) provided there exist @xmath105 and a positive definite function @xmath106 such that @xmath107 holds for all @xmath53 , @xmath63 , and @xmath108 . since ( [ nonauto ] ) has an iss lyapunov function when it is iss ( by the arguments of @xcite ) , the proof of ( * ? ? ?* theorem 1 ) shows that if ( [ nonauto ] ) is iss and autonomous , then it is also iiss , but not conversely , since e.g. @xmath109 is iiss but not iss .the next lemma follows from the arguments used in @xcite , @xcite , and @xcite .[ suff ] if ( [ nonauto ] ) admits an iss ( resp . , iiss ) lyapunov function , then it is iss ( resp . ,we show that hypotheses similar to those of ( * ? ? ?* theorem 3 ) ensure that ( [ m1 ] ) is uges . in fact , we show our conditions imply @xmath110 is iss when @xmath1 is sufficiently large ; see remark [ gissue ] below for results on the more general systems ( [ fastcontrol ] ) .our main assumption will be : there exist @xmath26 , a @xmath24-compatible dynamics ( [ m2 ] ) , and @xmath21 ( cf .( [ m3 ] ) above ) such that for all @xmath111 , all @xmath112 and sufficiently large @xmath113 , @xmath114 which is similar to ( * ? ? ? * property 2 ) in the special case where @xmath115 .two more advantages of our result are ( a ) it applies to cases where ( [ m2 ] ) is ugas but not necessarily uges ( cf . section [ exa1 ] below ) and ( b ) its proof leads to explicit lyapunov functions for ( [ dis ] ) ( cf .theorem [ cor1 ] below ) .see also theorem [ altthm ] below for cases where @xmath116 is not necessarily globally bounded . recall the definition of compatibility ( in definition [ compatt ] ) and the requirement @xmath117 in ( [ m3 ] ) .[ lef2 ] consider a system ( [ m1 ] ) .assume there exist @xmath26 , a @xmath24-compatible ugas system ( [ m2 ] ) , two constants @xmath118 and @xmath119 , and @xmath120 such that ( [ relate ] ) holds whenever @xmath121 , @xmath111 and @xmath112 and such that : @xmath122 then there is a constant @xmath123 such that for all constants @xmath124 , ( [ m1 ] ) is ugas and ( [ dis ] ) is iiss .if in addition @xmath28 , then ( [ dis ] ) is iss for all constants @xmath125 . in the special case where ( [ m2 ] ) is uges , ( [ m1 ] ) is uges for all constants @xmath126 and ( [ dis ] ) is ises for all constants @xmath125 . by ( [ nullness ] ) ,the condition @xmath127 in ( [ m16 ] ) is redundant when @xmath24 has the form @xmath128 for a constant @xmath129 , since @xmath130 can always be enlarged .our proof of theorem [ lef2 ] in section [ proof ] below will also show : [ cor1 ] let the hypotheses of theorem [ lef2 ] hold for some @xmath131 , and @xmath64 be a lyapunov function for ( [ m2 ] ) satisfying the requirements of definition [ compatt ] .then there exists a constant @xmath123 such that for all constants @xmath132 , @xmath133}(\xi , t):=\\ v\left(\xi-\frac{\sqrt{\alpha}}{2}\int _ { t-\frac{2}{\sqrt{\alpha}}}^t\int_s^t\{f(\xi , l,\alpha l)-\bar{f}(\xi , l)\}\ , { \rm d } l\ , { \rm d}s , t\right)\end{array}\ ] ] is a lyapunov function for ( [ m1 ] ) and an iiss lyapunov function for ( [ dis ] ) .if also @xmath134 , then @xmath135}$ ] is also an iss lyapunov function for ( [ dis ] ) for all constants @xmath132 .[ alt ] the proof of theorem [ lef2 ] constructs strict lyapunov functions for ( [ m1 ] ) in terms of strict lyapunov functions for the limiting dynamics ( [ m2 ] ) .it is natural to inquire whether one can instead construct strict lyapunov functions for ( [ m1 ] ) by strictifying nonstrict lyapunov functions for ( [ m1 ] ) ; see @xcite where the strictification approach was applied to nonautonomous systems that are not rapidly time - varying . in this section ,we extend this approach to cover ( [ m1 ] ) .our strictification result has the advantages in certain situations that ( a ) it does not require any knowledge of limiting dynamics , ( b ) it allows the derivative of the nonstrict lyapunov function to be zero or even positive at some points , and ( c ) it does not require ( [ m1 ] ) to be globally lipschitz in the state .in the rest of the section , we focus on systems with no controls , but the extension to the control system ( [ fastcontrol ] ) for appropriate @xmath6 can be done by similar arguments .we assume : * * * there exist @xmath100 , @xmath136 , a @xmath9 function @xmath137 , a bounded continuous function @xmath138 , and constants @xmath139 such that for all @xmath140 , @xmath33 , @xmath1 , and @xmath141 , we have : * * * @xmath142 * * * @xmath143 * * * @xmath144 , @xmath145 of @xmath57 , @xmath146 , @xmath147 , and the partial gradients of @xmath57 and @xmath147 whenever this would not lead to confusion . in section [ altproof ] ,we prove : [ altthm ] if assumption h. holds , then there exists a constant @xmath148 such that for all constants @xmath149 , @xmath150}(x , t)\\ = v(x , t ) - \left(\int_{t - 1}^{t } \left(\int_{s}^{t } p(\alpha l ) dl\right ) ds\right ) \theta(x , t)\end{array}\ ] ] is a lyapunov function for ( [ m1 ] ) .in particular , ( [ m1 ] ) is ugas for all constants @xmath151 .to make our arguments easy to follow , we first outline our method for proving these theorems .first , we give the proof of theorem [ lef2 ] for the special case where ( [ m2 ] ) is ugas and @xmath152 which includes the proof of theorem [ cor1 ] for the @xmath152 case .then we indicate the changes required if @xmath153 is bounded .finally , we specialize to the special case where @xmath154 for some constant @xmath129 which will prove the ises assertion of theorem [ lef2 ] .we first assume that ( [ m2 ] ) is ugas and @xmath155 , and we prove the iss property for ( [ dis ] ) for large constants @xmath1 . in what follows ,we assume all inequalities and equalities hold globally , unless otherwise indicated .let @xmath156 be as in the statement of the theorem , and fix @xmath157 with @xmath121 , @xmath158 , and a trajectory @xmath159 for ( [ dis ] ) and @xmath160 , with arbitrary initial condition .set @xmath161 where @xmath162 this is well - defined since we are assuming our dynamics are forward complete .set @xmath163 with @xmath164 so defined , one easily checks ( as was done e.g. in @xcite ) that for any @xmath165 and @xmath166 , @xmath167 and@xmath168 taking @xmath169 , ( [ fua ] ) multiplied through by @xmath170 , @xmath171 , and ( [ ze1 ] ) give @xmath172 let @xmath57 , @xmath173 , and @xmath174 satisfy the requirements @xmath175-@xmath176 from our compatibility condition ( in definition [ compatt ] ) and ( l1 ) .by @xmath175 with @xmath177 and ( [ along ] ) , the derivative of @xmath178 along the time - varying map @xmath179 defined in ( [ ze1 ] ) ( which we denote simply by @xmath180 in the sequel ) satisfies @xmath181\nonumber\\ \times \left(f(x(t ) , t,\eta^2t)+ { \mathbf u}(t)\right)+v_\xi(z(t),t)\ , \mathbf{u}(t).\nonumber \end{array}\ ] ] we deduce from ( [ relate ] ) , ( [ m16 ] ) , ( [ ze1 ] ) , and @xmath72 that @xmath182 moreover , ( [ m16 ] ) , the definition of @xmath164 , and @xmath72 give @xmath183 combining these inequalities and grouping terms gives @xmath184 on the other hand , ( [ ze1 ] ) , ( [ rest ] ) , and @xmath176 give @xmath185 when @xmath186 .since @xmath131 , this gives @xmath187 setting @xmath188 , it follows from ( [ rew])-([tohere ] ) that @xmath189 setting @xmath190}(x , t):=v(x+r_\alpha(x , t),t)$ ] , we see the derivative @xmath191 of @xmath178 along ( [ dis ] ) satisfies @xmath192}_t(x , t ) + v^{\scriptscriptstyle [ \alpha]}_x(x , t)\ , \{f(x , t,\alpha t)+{\mathbf u}(t)\}.\ ] ] we deduce from ( [ m3 ] ) , ( [ rew ] ) , and ( [ itfollows ] ) that when the constant @xmath7 ( and so also @xmath193 ) is sufficiently large , @xmath194}_t(x , t ) + v^{\scriptscriptstyle [ \alpha]}_x(x , t)\ , [ f(x , t,\alpha t)+u ] \leq -\frac{\bar c}{2}\delta^2(|x|/2)\end{array}\ ] ] and @xmath195}(x , t ) \le \delta_2(|x|+2\delta(|x|/2)/\eta ) $ ] by ( [ rest ] ) .it follows that @xmath135}$ ] is an iss lyapunov function for ( [ dis ] ) , so ( [ dis ] ) is iss for large @xmath7 , by lemma [ suff ] , as claimed .the ugas conclusion is the special case where @xmath196 . to prove the iiss assertion ,we instead follow the preceding argument up through ( [ tohere ] ) ( which is valid since that part did use the unboundedness of @xmath24 ) and then substitute @xmath197 into ( [ tohere ] ) and bound the resulting coefficient of @xmath198 to show that @xmath135}$ ] is an iiss lyapunov function for ( [ dis ] ) for sufficiently large @xmath7 ( by again using ( [ rest ] ) and ( [ rew ] ) and taking @xmath199 and @xmath200 for a suitable @xmath129 ) , which implies that ( [ dis ] ) is iiss for large @xmath7 , by lemma [ suff ] .we turn next to the special case where ( [ m2 ] ) is uges .let @xmath57 satisfy the requirements of lemma [ k02 ] above for ( [ m2 ] ) , and let @xmath159 be any trajectory for ( [ dis ] ) for any control @xmath201 starting at @xmath16 . define @xmath179 by ( [ ze1 ] ) .arguing as before except with this new choice of @xmath57 shows @xmath202 ( by taking @xmath203 in ( [ rest ] ) ) and that ( [ keyformula ] ) satisfies ( by @xmath176 and ( [ relate ] ) ) @xmath204 since @xmath205 for large @xmath193 as before .if we now define @xmath206 by @xmath207 , then we deduce as in the ugas case that if @xmath193 is large enough , and if @xmath208 for all @xmath209 , then we also have @xmath210 for all @xmath209 and @xmath211 .this gives @xmath212 , so @xmath213 so our estimate on @xmath214 and the form of @xmath179 give @xmath215 we conclude as before that if ( [ m2 ] ) is uges , then , when the constant @xmath1 is large enough , ( [ m1 ] ) is also uges and ( e.g. , by the proof of ( * ? ? ?* lemma 2.14 ) ) ( [ dis ] ) is ises , which proves our theorem .[ gissue ] the method we used in the proof of theorem [ lef2 ] can be used to prove the iss property for ( [ fastcontrol ] ) under appropriate growth assumptions on the matrix - valued function @xmath216 . clearly , some growth condition on @xmath6 is needed and linear growth of @xmath6 is not enough , since @xmath217 is not iss .one way to extend our theorem to ( [ fastcontrol ] ) is to add the hypotheses that @xmath6 is @xmath9 and that there is a constant @xmath218 such that for all @xmath219 , @xmath140 , and @xmath1 , @xmath220 , where @xmath221 is the @xmath222-norm on @xmath223 and @xmath134 satisfies @xmath175-@xmath176 for some lyapunov function @xmath57 for ( [ m2 ] ) .applying the first part of the proof of theorem [ lef2 ] except with the new @xmath224 function @xmath225 we then conclude as before that ( [ fastcontrol ] ) is iss for sufficiently large @xmath1 . if instead @xmath226 is bounded , then ( [ fastcontrol ] ) is iiss when @xmath7 is sufficiently large , by our earlier argument .[ newrk ] the decay requirement ( [ m3 ] ) on @xmath21 from theorem [ lef2 ] can be relaxed , as follows .we assume the flow map @xmath46 of ( [ m2 ] ) satisfies the uges conditions ( [ gasdef])-([betachoice ] ) for some @xmath47 and @xmath227 , where @xmath228 satisfies ( [ m16 ] ) , and we let @xmath57 be as in lemma [ k02 ]. we can choose the constant @xmath229 in ( [ m5 ] ) to be @xmath230 by the proof of lemma [ k02 ] .it follows from our argument above that theorem [ lef2 ] remains true for cases where ( [ m2 ] ) is uges if ( [ m3 ] ) is relaxed to @xmath231 a similar relaxation can be made in the more general ugas setting covered by theorem [ lef2 ] .we first set @xmath232}= u^{\scriptscriptstyle [ \alpha]}_t(x , t)+ u^{\scriptscriptstyle [ \alpha]}_x(x , t)f(x , t,\alpha t)$ ] for all @xmath111 , @xmath233 , and @xmath1 .using assumptions h1 . and h3 .and ( [ fua ] ) with @xmath234 independent of @xmath209 gives @xmath235 } & \leq & - w(x , t ) + p(\alpha t ) \theta(x , t ) - p(\alpha t ) \theta(x , t)\\ & & + \left(\int_{t - 1}^{t } p(\alpha l ) dl\right ) \theta(x , t ) \\ & & - \left(\int_{t - 1}^{t } \left(\int_{s}^{t } p(\alpha l ) dl\right ) ds\right ) \left(\frac{\partial \theta}{\partial x } f + \frac{\partial \theta}{\partial t}\right ) \\& \leq & - w(x , t ) + \left|\int_{t - 1}^{t } p(\alpha l ) dl\right| \frac{1}{c } w(x , t ) \\ & & + \left|\int_{t - 1}^{t } \left(\int_{s}^{t } p(\alpha l ) dl\right ) ds\right| \frac{1}{c } w(x , t ) \end{array}\ ] ] along trajectories of ( [ m1 ] ) , where we omit the argument @xmath236 of @xmath5 . for any @xmath1 , @xmath33 , and @xmath237 $ ] , assumption h2 .gives @xmath238 where @xmath239 and @xmath240 .the proof of ( [ zeroing ] ) uses the facts that both endpoints of the interval @xmath241 $ ] are integer multiples of @xmath242 and that , according to assumption h2 , the integral of @xmath164 over any interval whose endpoints are integer multiples of @xmath242 is equal to zero .choosing @xmath243 to be any global bound on @xmath244 over all @xmath245 , ( [ zeroing ] ) gives @xmath246\ ] ] for all @xmath33 and @xmath1 .hence , for all @xmath33 , @xmath247}\left|\int_{s}^{\alpha t } p(l ) dl\right|\nonumber\\ & \le & 2tp_{\max}/\alpha\end{aligned}\ ] ] using these estimates , ( [ jb6 ] ) gives @xmath248}\le -w(x , t)/2 $ ] for all @xmath10 and @xmath33 , as long as @xmath249 . since @xmath250 , this gives the lyapunov decay estimate . by assumption h3 . and ( [ termest ] ) , @xmath251}\in { \rm uppd}$ ] for large enough constants @xmath1 .this proves the theorem .we next illustrate how theorem [ lef2 ] extends the results of @xcite and @xcite . in @xcite ,the limiting dynamics ( [ m2 ] ) are assumed to be uniformly locally exponentially stable . in our first example , ( [ m2 ] ) is ugas but not necessarily uges .we next consider a class of systems ( [ m1 ] ) from @xcite from identification where the limiting dynamics ( [ m2 ] ) is linear and exponentially stable .for these systems , our work complements @xcite by providing formulas for lyapunov functions for ( [ m1 ] ) that are expressed in terms of the quadratic lyapunov functions for the limiting dynamics and that have the additional desirable property that they are also iss lyapunov functions for ( [ fastcontrol ] ) for suitable functions @xmath6 .finally , we apply our results to a friction model for a mass - spring dynamics . in all three examples ,the limiting dynamics has a simple lyapunov function structure so our results give explicit lyapunov functions for the original rapidly time - varying dynamics .consider the following variant of the scalar example on @xcite : @xmath252\{1 + 10\sin(\alpha t)\}\end{array}\ ] ] where @xmath253 are @xmath9 functions such that @xmath254 is odd , @xmath255 , @xmath256 on @xmath257 , and @xmath258 for all @xmath259 .one easily verifies the hypotheses of theorem [ lef2 ] using @xmath260,\\ v(x , t)\ , \equiv\ , \bar v(x)\ , : = \ , \int_0^{x}\sigma_1(s){\rm d}s,\\ \delta(s):=33\sigma_1(2s),\ ; \ ; { \rm and}\ ; \ ; n(\eta):=60/\eta^2 \ ; { \rm for\ large\ } \eta.\end{array}\ ] ] this allows e.g. @xmath261 in which case ( [ m2 ] ) is ugas but not uges because @xmath262 along all of its trajectories @xmath159 .condition @xmath175 follows because @xmath263 for all @xmath264 , which holds because @xmath265 for all @xmath35 .theorem [ cor1 ] then gives the following iiss lyapunov function for ( [ dis ] ) for large @xmath1 : @xmath266 where @xmath267 . in particular, this is a lyapunov function for @xmath268 , and it is also an iss lyapunov function for ( [ dis ] ) if @xmath269 ( e.g. if @xmath270 for @xmath271 and @xmath272 ) . our conditions on the @xmath273 s can not be omitted even if the limiting dynamics ( [ m2 ] ) is uges ; see ( * ? ? ?* @xmath2748.2 ) .for example , if @xmath275 and @xmath276 , then ( [ m2 ] ) is uges , but ( [ nonuges ] ) is only shown to be _ locally _ exponentially stable for large @xmath1 ; see @xcite .this does not contradict our theorem because in that case ( [ m16 ] ) would be violated .consider the following variant of the example in ( * ? ? ?* section 8.1 ) : @xmath277 with state @xmath111 and inputs @xmath108 , where we assume * * * @xmath278 is bounded and continuous and admits a @xmath38 function @xmath279 and a constant @xmath280 for which 1 .@xmath281 2 .@xmath282{\rm d}s\vert\le m(t_2-t_1)$ ] if @xmath283 * * @xmath284 is continuous and satisfies @xmath285 for all @xmath166 , and there exist constants @xmath286 such that for all @xmath287 , @xmath1 , and @xmath111 , we have : 1 .@xmath288 2 .@xmath289 * * @xmath290 is continuous and is @xmath9 in @xmath75 , and there exists a constant @xmath119 such that @xmath291 @xmath292 , @xmath33 , and @xmath1 , and each component @xmath293 of @xmath6 . here@xmath294 denotes the @xmath295 identity matrix .notice that we allow @xmath5 to take both positive and negative values . the special case of ( [ pertid ] ) where @xmath296 was studied in @xcite where it is shown that the corresponding rapidly varying dynamics @xmath297 satisfies the hypotheses of theorem [ lef2 ] with the uges dynamics @xmath298 and with @xmath24 of the form @xmath154 for a constant @xmath129 . the particular case of ( [ pertid ] ) in which @xmath299 has been extensively studied in systems identification ;see @xcite .however , these earlier results do not provide explicit iss lyapunov functions for ( [ pertid ] ) . on the other hand, the following lemma provides an explicit lyapunov function for ( [ limid ] ) : [ plem ] let assumptions i .- iii . hold .if we choose @xmath300 where @xmath301 then @xmath302 is a lyapunov function for ( [ limid ] ) for which @xmath303 satisfies the requirements of lemma [ k02 ] . to prove lemma [ plem ] , we apply ( [ fua])-([fub ] ) with @xmath304 and @xmath305 and group terms to check that the derivative of @xmath57 along trajectories of ( [ limid ] ) satisfies @xmath306 m(t ) m^t(t ) x \\ & & - x^t \left[\int_{t-\tilde c}^t m(l)m^t(l)\ , { \rm d}l \right ] x\\ \dot{v } & \leq & ( 2 f^\star \kappa + \tilde c ) |m^t(t ) x|^2- \alpha ' |x|^2 \\ & & + 2 |f^\star| |x| \left[\int_{t-\tilde c}^t\int_s^t |m(l)|^2 { \rm d}l { \rm d}s\right ] |m(t)| |m^t(t ) x| \\ & \leq & \ ! ( 2 f^\star \kappa + \tilde c ) |m^t(t ) x|^2 + \tilde c^2 \end{array}\ ] ] everywhere , by ii.(1 ) . by the triangle inequality ,@xmath307 so @xmath308 holds everywhere .recalling that @xmath285 everywhere and that @xmath309 is everywhere positive definite , the fact that @xmath303 satisfies the requirements of lemma [ k02 ] follows immediately from our choice of @xmath310 .remark [ gissue ] now gives : let ( [ pertid ] ) satisfy conditions i .- iii . above and let @xmath57 be as in lemma [ plem ] .then there exists a constant @xmath311 such that for each constant @xmath312 , @xmath313}(x , t):=v\left(\left[i-\frac{\sqrt{\alpha}}{2}\int_{t-2/\sqrt{\alpha}}^t\int_s^t d(l)\ , { \rm d}l\ , { \rm d}s\right]x , t\right)\ ] ] where @xmath314 is an iss lyapunov function for ( [ pertid ] ) .the following one degree - of - freedom mass - spring system from @xcite arises in the control of mechanical systems in the presence of friction : @xmath315 where @xmath316 and @xmath317 are the mass position and velocity , respectively ; @xmath318 , @xmath319 denote positive time - varying viscous , coulomb , and static friction - related coefficients , respectively ; @xmath320 is a positive constant corresponding to the stribeck effect ; @xmath321 is a positive definite function also related to the stribeck effect ; @xmath322 denotes a positive time - varying spring stiffness - related coefficient ; and sat@xmath323 denotes any continuous function having these properties : @xmath324 we model the saturation as the differentiable function @xmath325 where @xmath326 is a large positive constant .note for later use that @xmath327 for all @xmath328 .we assume the friction coefficients vary in time faster than the spring stiffness coefficient so we restrict to cases where @xmath329 .our precise mathematical assumptions on ( [ frictionexample ] ) are : @xmath322 and the @xmath273 s are bounded @xmath9 functions ; @xmath330 has a globally bounded derivative ; and there exist constants @xmath331 , with @xmath332 and @xmath333 for @xmath334 , and a @xmath38 function @xmath335 such that @xmath336 for all @xmath337 satisfying @xmath338 .although the @xmath273 s are positive for physical reasons , we will not require their positivity in the sequel .we show ( [ frictionexample ] ) satisfies the requirements of the version of theorem [ lef2 ] from remark [ gissue ] ( with @xmath154 for a constant @xmath130 ) when ( [ m2 ] ) is @xmath339 assuming this additional condition whose physical interpretation is that the spring stiffness is nonincreasing : @xmath340 to this end , set @xmath341 and @xmath342 where @xmath343/\tilde \sigma_1 $ ] . since @xmath344 , @xmath345 for all @xmath346 and @xmath33 .also , since @xmath347 everywhere , the derivative @xmath348 along trajectories of ( [ fbf ] ) satisfies @xmath349x_2-[2ax_2+x_1]\\\ ; \ ; \ ; \ ; \ ; \ ; \ ; \times\{\tilde{\sigma}_1x_{2}+ \left[\tilde\sigma_2+\tilde \sigma _ { 3}e^{-\beta _ { 1}\mu ( x_{2})}\right ] { \rm sat } ( x_{2})+k(t)x_{1}\ } \end{array}\ ] ] and therefore , by grouping terms , we also have @xmath350\\ = -b|x|^2-\frac{k_o}{2}\left(|x_1|-\frac{s}{k_o}|x_2|\right)^2\ ! + \left(\frac{s^2}{2k_o}+\frac{1}{2 } -a\tilde \sigma_1\right)x^2_2\\ \le -b|x|^2,\ ; { \rm where}\ ; b:=\min\{k_o/2,a\tilde \sigma_1 - 1/2\}. \end{array}\ ] ] the preceding inequalities imply that @xmath351 is a lyapunov function for ( [ fbf ] ) satisfying the requirements of lemma [ k02 ] .the integral bound requirement ( [ relate ] ) from our theorem follows from ( [ sigma1 ] ) and the sublinear growth of @xmath352 , since the integral bound can be verified term by term . the remaining bounds from ( [ m16 ] ) follow because @xmath353 and sat have globally bounded derivatives .we conclude that for sufficiently large constants @xmath1 , ( [ frictionexample ] ) admits the iss lyapunov function @xmath354}(\xi , t)=v\left(\xi_1 , \xi_2 + \frac{\sqrt{\alpha}}{2}\int_{t-\frac{2}{\sqrt{\alpha}}}^t\int_s^t \gamma_\alpha(l,\xi)\ , { \rm d } l\ , { \rm d}s , t\right ) \end{array}\ ] ] where @xmath57 is the lyapunov function ( [ qu1 ] ) for ( [ fbf ] ) and @xmath355 so ( [ frictionexample ] ) is iss for large enough @xmath1 , by remark [ gissue ] .[ const ] the preceding construction simplifies considerably if @xmath356 and @xmath357 in ( [ frictionexample ] ) are positive constants . in that case , the limiting dynamics ( [ m2 ] ) can be taken to be @xmath358 and the iss lyapunov function for ( [ frictionexample ] ) becomes @xmath359 with @xmath57 defined by ( [ qu1 ] ) , since the @xmath360 term in the difference @xmath361 is no longer present .simple calculations show that theorem [ altthm ] applies to ( [ nonuges ] ) and ( [ pertid ] ) without controls ( with the same choices of @xmath57 that we used in our earlier discussions of those systems ) , assuming in the latter case that requirements i .- iii .hold and for instance @xmath5 is a suitable periodic function and @xmath362 is bounded . we next show how theorem [ altthm ] also applies to cases that are not tractable by theorem [ lef2 ] .[ altillus ] simple calculations that we omit because of space constraints show that the one - dimensional dynamics @xmath363 satisfy h. using @xmath364 , @xmath365 , @xmath366 , @xmath367 , @xmath368 , and small enough @xmath369 , so ( [ ngs ] ) has the global lyapunov function @xmath370}(x , t)= \frac{x^4}{4 } - \left(\displaystyle\int_{t - 1}^{t } \left(\displaystyle\int_{s}^{t } 10 \cos(\alpha l ) dl\right ) ds\right)\frac{x^6}{1 + x^2}\ ] ] when @xmath1 is large enough .however , ( [ ngs ] ) is not covered by theorem [ lef2 ] since it is not globally lipschitz in @xmath75 .consider the nonautonomous scalar system @xmath371 where @xmath164 is an unknown , fast time - varying parameter satisfying @xmath372 for all @xmath373 and some constant @xmath374 and admitting a constant @xmath375 such that assumption h2. holds for all @xmath376 .we assume the control @xmath377 is amplitude limited in the sense that @xmath378 for some constant @xmath379 .we show that the saturated state feedback @xmath380 renders ( [ ex ] ) ugas .( a similar argument shows that @xmath381 stabilizes ( [ ex ] ) for any constant @xmath382 . ) for simplicity , let @xmath383 and @xmath384 .the derivative of @xmath385 along ( [ ex ] ) in closed loop with ( [ u ] ) is @xmath386 simple calculations allow us to verify the hypotheses of theorem [ altthm ] for the closed loop system using @xmath387 , @xmath388 , and small enough @xmath369 , so we know ( [ u ] ) indeed uniformly globally asymptotically stabilizes ( [ ex ] ) with control lyapunov function @xmath370}(x , t)=\dfrac{1}{2}x^{2}-\left ( \int_{t-1}^{t}\int_{s}^{t}p(\alpha l ) \ , { \rm d}l\ , { \rm d}s\right ) \frac{x^{3}}{1+x^{2 } } \label{clf}\ ] ] when @xmath1 is sufficiently large .hypotheses similar to those of ( * ? ? ?* theorem 3 ) are sufficient for uniform global exponential stability of rapidly time - varying nonlinear systems .we provided complementary results to @xcite by establishing uniform global asymptotic stability of fast time - varying dynamics without requiring local exponential stability of the limiting dynamics , and by constructing global lyapunov functions for fast time - varying systems .our lyapunov constructions are new even in the special case of exponentially stable dynamics , and are input - to - state stable lyapunov functions when the dynamics are control affine , under appropriate conditions .our results apply to dynamics that are not necessarily uniformly lipschitz in the state .we illustrated our methods using a friction example .edwards , h. , lin , y. , & wang , y. ( 2000 ) . on input - to - state stability for time - varying nonlinear systems ._ proceedings of the 39th ieee conference on decision and control _ , sydney , australia , ( pp .3501 - 3506 ) .grne , l. , sontag , e.d . , & wirth , f. ( 1999 ) .asymptotic stability equals exponential stability , and iss equals finite energy gain - if you twist your eyes . _systems and control letters _ , 38(2 ) , 127 - 134 .

<S> we explicitly construct global strict lyapunov functions for rapidly time - varying nonlinear control systems . </S><S> the lyapunov functions we construct are expressed in terms of oftentimes more readily available lyapunov functions for the limiting dynamics which we assume are uniformly globally asymptotically stable . </S><S> this leads to new sufficient conditions for uniform global exponential , uniform global asymptotic , and input - to - state stability of fast time - varying dynamics . </S><S> we also construct strict lyapunov functions for our systems using a strictification approach . </S><S> we illustrate our results using several examples . , , time - varying systems , input - to - state stability , lyapunov function constructions </S>

1207.5242

the periodically driven transverse ising model describes the dynamics of @xmath0 interacting two - level systems in a time - dependent transverse local field @xcite @xmath1 where @xmath2 are pauli matrices at the @xmath3th site and we assume periodic boundary conditions @xmath4 for @xmath5 . in the followingwe shall consider @xmath6 , and a monochromatic modulation of the transverse field with a static contribution @xmath7 .associated with the hamiltonian equation is a conserved parity @xmath8 , such that @xmath9=0 $ ] , which is given by @xmath10 in the thermodynamic limit @xmath11 , the undriven ising model @xmath12 exhibits a second - order qpt at @xmath13 from a symmetric paramagnetic phase ( @xmath14 ) to a symmetry - broken ferromagnetic phase ( @xmath15 ) @xcite .our aim in this paper is to study the new aspects of criticality under the effect of driving .+ in this section we provide the basics on the formalism used to describe the time - dependent ising model .in particular , we find a resonance condition related to @xmath16-photon processes under the effect of driving . in this paperwe consider the restriction of the hamiltonian equation to the subspace with even @xmath17 number of fermionic quasiparticles ( see appendix [ appendixa ] ) .after a jordan - wigner transformation and a discrete fourier transform of hamiltonian equation we obtain @xmath18(c_k^{\dagger}c_k+c_{-k}^{\dagger}c_{-k})-2g(t)\right\ } \nonumber \\ & & + \sum_{k>0}2 j\sin k(c_k^{\dagger}c_{-k}^{\dagger}+c_{-k } c_{k } ) = \sum_{k>0 } \hat{h}_{k}(t ) , \end{aligned}\ ] ] where @xmath19 and @xmath20 are fermionic operators @xcite . for finite size@xmath0 of the spin chain , the quasimomentum is restricted to @xmath21 . in the following ,we focus on the thermodynamic limit @xmath22 , where we have @xmath23 . even if we prepare the system initially in a ground state of the undriven model , under the effect of nonadiabatic external driving , the system will experience transitions to excited states . rather recently, a formalism has been developed to deal with this kind of dynamical situation @xcite .the idea is based on the fact that hamiltonian equation conserves momentum and parity . as a consequence, we can use the bcs ansatz for the evolution of the quantum state of the system @xmath24 , \ ] ] which implies that for a given quasimomentum @xmath25 , the quantum evolution is restricted to the nambu subspace @xmath26 , consisting of doubly occupied @xmath27 and unoccupied @xmath28 states of @xmath29 fermions @xcite .+ the matrix representation of the operator @xmath30 in the nambu subspace is given by @xmath31 where @xmath32 is the bogoliubov - de gennes ( bdg ) hamiltonian , @xmath33 , @xmath34 , and @xmath35 . by defining the spinor @xmath36 , and considering the bcs ansatz eq , it is possible to show that the coefficients @xmath37 and @xmath38 should satisfy the differential equation @xmath39 which constitutes the dynamical version of the bogoliubov - de gennes equation @xcite . at this point we have translated the many - body problem into the solution of the time - dependent schrdinger equation for an effective two - level system .+ under periodic driving , the floquet theorem states that the solution of equation can be written as @xmath40 where @xmath41 denote the floquet modes corresponding to the quasienergies @xmath42 .furthermore , in the extended hilbert space @xmath43 , where @xmath44 is the hilbert space of square integrable functions and @xmath45 is the space of time - periodic functions , the floquet states satisfy the eigenvalue problem @xmath46 where @xmath47 , @xmath48 is the floquet- bogoliubov - de gennes ( fbdg ) hamiltonian , @xmath49 are the quasienergies , and the floquet modes @xmath50 have the same period @xmath51 as the external driving @xcite . in the thermodynamic limit , the ising model is characterized by an infinite set of collective excitations . under the effect of an external driving, the possibility of multiphotonic resonances arises @xcite . to study such quantum resonances ,let us consider the system in the absence of driving @xmath52 . in this case, the floquet modes and the quasienergies become the stationary states @xmath53 and the excitation spectrum @xmath54 of the undriven system , respectively @xcite . therefore, the solution of equation can be written in the form of equation as follows : @xmath55 the energy gap is given by @xmath56 , where @xmath57 in the semiclassical theory of light - matter interaction , we can interpret a floquet state as a light - matter quantum state containing a definite , though very large , number of photons @xcite .multiple transitions between quantum states of the spin chain that are not directly coupled by the interaction can occur by means of intermediate states with a different number of photons present @xcite .in particular , @xmath16-photon transitions occur when the condition @xmath58 with integer @xmath16 is satisfied . for a parametric oscillator with fundamental frequency @xmath59 , equation is the usual resonance condition @xcite . in floquettheory , equation implies the existence of a crossing between the single - particle energy levels @xmath59 and @xmath60 when the energy spectrum is folded into the brillouin zone @xcite .such a crossing occurs at the wave vector @xmath61 where the resonance condition is fulfilled , as depicted in figure [ fig1 ] ( a ) .figure [ fig1 ] ( a ) depicts the energy dispersion relation of the undriven system , and the continuous lines in figure [ fig1 ] ( b ) , the corresponding folding of the energy spectrum into the first brillouin zone @xmath62 . in this paperwe focus on the weak spin - spin coupling limit @xmath63 . in this limitthe multiphoton resonance condition reads @xmath64 such resonance condition will be used in the next section to perform a description of the system based on an effective time - independent hamiltonian which is valid for parameters close to a multiphotonic resonance .( a ) typical gapped energy spectrum @xmath65 of the undriven system corresponding to the paramagnetic phase @xmath66 . in this case, the energy dispersion is slightly curved because the curvature is proportional to the spin - spin interaction strength @xmath67 .( b ) the continuous lines depict the spectrum @xmath65 when it is folded into the first brillouin zone , the crossing at @xmath68 is related to a two - photon resonance .the dashed lines represent the quasienergy spectrum @xmath69 for a driving amplitude @xmath70 .the parameters are @xmath71 , and @xmath72 . ]the dynamical bdg equations allow us to investigate the full quantum evolution of the driven system . as we are interested in the asymptotic quantum dynamics and the description of critical signatures , it is convenient to describe the dynamics of the system in a rotating frame . in the weak spin - spin interaction limit ,it is possible to neglect the fast oscillations in the rotating frame , and a description of the system based on the description for time - independent systems is possible via an approximate effective hamiltonian . motivated by the @xmath16-photon resonance condition in the weak spin - spin coupling limit , eq ., we perform a description of the system based on the rotating wave approximation @xcite .let us perform a unitary transformation of hamiltonian equation into a convenient rotating frame via the unitary operator @xmath73 , \end{aligned}\ ] ] where @xmath74 . in the rotating frame the dynamicsis governed by the hamiltonian @xmath75^\dagger \hat{\mathcal{h } } \hat{u}^{(m)}(t)$ ] , where @xmath76 $ ] is the floquet hamiltonian .the explicit form of this operator is given by @xmath77\right\}\sum_{i=1}^n \sigma_i^z \sigma_{i+1}^z-\frac{j}{2}\left\{1-\cos[4\alpha_{m}(t)]\right\}\sum_{i=1}^n \sigma_i^y \sigma_{i+1}^y \nonumber \\ & & + \frac{j}{2}\sin[4\alpha_{m}(t)]\sum_{i=1}^n \sigma_i^z \sigma_{i+1}^y + \frac{j}{2}\sin[4\alpha_{m}(t)]\sum_{i=1}^n \sigma_i^y \sigma_{i+1}^z , \end{aligned}\ ] ] where the detuning @xmath78 describes how far the system is from resonance , and @xmath16 is an integer that denotes the order of the multiphotonic resonance @xcite . by using the identity @xmath79 where @xmath80 is the @xmath81th - order bessel function @xcite, the hamiltonian equation can be written in the form @xmath82 in analogy with the standard rwa of quantum optics , we obtain an approximate hamiltonian to describe the @xmath16th resonance by neglecting all the terms in @xmath83 with oscillatory time dependence : @xmath84 . this approximation is valid as long as the condition @xmath85 holds @xcite .+ finally , we obtain the time - independent effective hamiltonian @xmath86 where the parameters @xmath87 $ ] and @xmath88 $ ] denote effective anisotropies in the rotating frame .interestingly , the effective hamiltonian equation corresponds to an exactly solvable model , i.e. , it is unitarily equivalent to the @xmath89 anisotropic spin chain in a transverse field @xcite .however , in our case , the anisotropies depend upon both the order @xmath16 of the resonance as well as the driving amplitude @xmath90. therefore , the driving amplitude of the local field now plays the role of a new parameter that influences the criticality of the system . under the rwa ,the hamiltonian @xmath91 and the floquet hamiltonian @xmath92 are isospectral operators , which implies that the eigenvalues of the effective hamiltonian correspond to the quasienergies . as we show in appendix[ appendixb] similarly to the ising model @xcite after a jordan - wigner transformation , and a discrete fourier transform , the hamiltonian equation can be written as follows : @xmath93 \nonumber \\ & + \sum_{k>0}(-1)^{m}\delta_k\mathcal{j}_{m}\left(\frac{4 g_{1}}{\omega}\right)(c_k^{\dagger}c_{-k}^{\dagger}+c_{-k } c_{k } ) \nonumber \\ & = \sum_{k>0 } \hat{h}^{(m)}_{0,k } .\end{aligned}\ ] ] the matrix representation of @xmath94 in the nambu subspace is given by @xmath95 the hamiltonian equation can be diagonalized via a bogoliubov transformation @xmath96 where @xmath97 ^ 2}.\ ] ] furthermore , the quasienergies are defined ( modulus @xmath98 ) by the equation @xmath99 as defined in equation . the quasienergy gap in the fermion picture is given by @xmath100 .therefore , when the gap closes , modulus @xmath98the effective hamiltonian exhibits a behavior which resembles the dynamics of a critical quantum system .the dashed lines in figure [ fig1 ] ( b ) depict the quasienergy dispersion relation for @xmath101 .we observe that the driving lifts the degeneracy giving rise to an anticrossing .based on the well - known results for the time - independent @xmath89 model that we summarize in appendix [ appendixb ] , we find that the system described by the effective hamiltonian equation exhibits an effective nonequilibrium ising - like qpt along the critical lines @xmath102 , and a nonequilibrium anisotropic qpt along the lines where @xmath103 , as long as the condition @xmath104 holds .the gapless quasienergy excitation spectrum for parameters along the critical lines is a direct consequence of coherent destruction of tunneling @xcite , i.e , of the existence of a generalized parity symmetry in the extended hilbert space @xmath43 , where @xmath44 is the hilbert space of square integrable functions and @xmath45 is the space of time - periodic functions @xcite .figure [ fig2 ] ( a ) depicts the character of the quasienergy excitation spectrum @xmath105 for parameters in the ferromagnetic phase fm@xmath106 , along the critical line , and in the ferromagnetic phase fm@xmath107 , respectively .figure [ fig2 ] ( b ) depicts the phase diagram for the nonequilibrium qpt in the neighborhood of the two - photon resonance .the white zones in the phase diagram correspond to the effective paramagnetic phase and are defined by the inequality @xmath108 , for @xmath109 , where @xmath110 denotes the maximum detuning for which the rwa is still valid .the anisotropic transition is characterized by two ferromagnetic phases , i.e. , for @xmath111 the system is in a ferromagnetically ordered phase along the @xmath112 direction fm@xmath107 , while it is the other way around in the fm@xmath106 phase . in the particular case@xmath113 , the effective hamiltonian equation is unitarily equivalent to the @xmath89 model in the absence of a transverse field .therefore , in this special case the system only exhibits the conventional anisotropic transitions between the ferromagnetically ordered fm@xmath107 and fm@xmath106 phases . in figure [ fig2 ] ( c )we plot the effective asymmetries @xmath114 and @xmath115 as a function of the driving amplitude @xmath90 in the case of a two - photon resonance .( color online ) nonequilibrium quantum phase transition in the driven ising chain in a time - dependent transverse field @xmath116 .@xmath117 depicts the quasienergy dispersion relations @xmath105 for parameters in the ferromagnetic phases fm@xmath107 and fm@xmath106 , and along a critical line .@xmath118 depicts the phase diagram of the nonequilibrium phase transition around the @xmath109 resonance as a function of the driving amplitude @xmath90 and the static local field @xmath119 .the white zones represent the paramagnetic phase .correspondingly , the blue ( dark gray ) zones represent the ferromagnetic phase fm@xmath107 and the green zones ( light gray ) the ferromagnetic phase fm@xmath106 .@xmath120 depicts the effective asymmetries in the @xmath112 direction @xmath121 [ blue ( dark gray ) curve ] and @xmath122 direction @xmath123 [ green ( light gray ) curve ] as a function of the driving amplitude @xmath90 . for this plot ,we have considered @xmath124 . ]as we previously mentioned , the eigenvalues of the effective hamiltonian correspond to the quasienergies of the system .however , the corresponding eigenstates do not necessarily correspond to floquet modes . in order to obtain the floquet modes, one should apply a unitary transformation back into the laboratory frame . in so doing ,the positive - quasienergy floquet mode around the @xmath16-photon resonance in the nambu subspace is given by @xmath125 and , correspondingly , the negative - quasienergy floquet mode is given by @xmath126 where @xmath127 now , let us investigate the quantum evolution in the laboratory frame around the @xmath16-photon resonance when the system is initialized in a paramagnetic state of the undriven model with all the spin polarized along the @xmath128 axis @xmath129 restricted to the nambu subspace for a given @xmath130 , such an initial state corresponds to the spinor @xmath131 , whose quantum evolution is given by @xmath132 where @xmath133 denotes the floquet states restricted to the nambu subspace . by using the exact quantum evolution of the initial paramagnetic state , we are able to calculate the dynamics of physical observables in the laboratory frame .the transverse magnetization density @xmath134 gives us information about the occurrence of a macroscopic polarization of the spins along the @xmath128 axis .let us consider the expectation value @xmath135 close to the @xmath16-th resonance , @xmath136 where @xmath137 and @xmath138 is the fbdg hamiltonian equation .( color online ) ( a ) time evolution of the dimensionless magnetization density @xmath134 in the thermodynamic limit ( black curve ) and exact numerical result for a finite system consisting of @xmath139 spins [ green ( light gray ) curve ] for @xmath140 , where @xmath51 is the period of the external driving .time evolution for parameters corresponding to ( a ) the nonequilibrium paramagnetic phase @xmath141 , for @xmath109 , where @xmath142 , ( b ) the ising - like critical line @xmath143 , for @xmath109 , where @xmath144 , and ( c ) time evolution for parameters corresponding to the nonequilibrium ferromagnetic phase ( fm@xmath107 ) @xmath145 , for @xmath109 , where @xmath146 . the insets show the detail of the magnetization curves .we have considered @xmath124 . ]figure [ fig3 ] depicts the dynamics of the magnetization density in the thermodynamic limit calculated using rwa ( black curve ) .in particular , figure [ fig3 ] ( a ) shows the dynamics for parameters in the nonequilibrium paramagnetic phase , figure [ fig3 ] ( b ) at the ising - like critical line @xmath147 , and figure [ fig3 ] ( c ) for the ferromagnetic phase fm@xmath107 .we observe that in the paramagnetic phase the system exhibits a stationary state which corresponds to a polarized state along the local field direction .in contrast , for parameters corresponding to the critical line and the ferromagnetic phase fm@xmath107 , the state is not totally polarized along this direction .furthermore , at the ising - like critical line , the magnetization density does not exhibit oscillations .the green curve ( light gray ) in figure [ fig3 ] depicts the result of exact numerical calculation ( see appendix [ appendixc ] ) of the magnetization density for a finite size system @xmath139 .+ in this dynamical scenario the connection with criticality is not obvious .rather , signatures of quantum criticality in the laboratory frame may appear in the asymptotic dynamics .let us consider now the time evolution of the expectation value of a general observable @xmath148 in general , following the argument established in @xcite , it is possible to show that @xmath149 , where @xmath150^{\dagger } \mathcal{o}_{k } \phi_{k , m}^{(\lambda)}(t)\ ] ] is the _ periodic _ contribution to the expectation value , which corresponds to synchronization with the external driving . herewe consider @xmath151 $ ] and @xmath152 $ ] .correspondingly , @xmath153^{\dagger } \mathcal{o}_{k } \phi_{k , m}^{(-)}(t)\right\}\ ] ] denotes the _ transient _ component , which decays to zero in the long - time limit as a consequence of destructive interference in time @xcite .therefore , the system tends to synchronize with the external control in the long - time limit .the particular case @xmath113 for @xmath154 has been discussed in ref .@xcite in the context of freezing of the response in a many - body system . in this case, the system only exhibits the conventional anisotropic transition , which is reflected in the behavior of the magnetization dynamics .furthermore , the anisotropic critical lines @xmath155 are related to the effect of maximal freezing discussed in ref .we conclude that in the asymptotic dynamics , the floquet modes determine the quantum critical behavior , as we discuss in the next section .we now define cycle - averaged expectation values of physical observables . in the case of a time - dependent hamiltonian @xmath156 , the energy is not conserved .therefore , to describe signatures of the quantum phase transition in the laboratory frame we define the averaged energy @xmath157 in the floquet state @xmath158 as @xmath159^{\dagger}h_{k}(t)\psi^{(\pm)}_{k , m}(t ) \nonumber \\ & = \int_{0}^{t}\frac{dt}{t}\int_{0}^{\pi}\frac{dk}{2\pi } \left(\varepsilon_{k , m}^{(\pm)}+\left[\phi_{k , m}^{(\pm)}(t)\right]^{\dagger } i\frac{\partial}{\partial t}\phi_{k , m}^{(\pm)}(t)\right ) .\end{aligned}\ ] ] ( color online ) cycle - averaged energy @xmath160 as a function of the driving amplitude @xmath90 .( a ) for parameters in nonequilibrium paramagnetic phase @xmath141 , for @xmath109 , where @xmath161 .( b ) for parameters inside the ladder of ferromagnetic phases @xmath145 , for @xmath109 , where @xmath162 .the insets depict the second derivative of the cycle - averaged energy as a function of the driving amplitude @xmath90 .we have considered @xmath124 . ] by using the analytic expressions for the floquet modes , eqs . and, we obtain the expression @xmath163 on exact resonance @xmath113 , we obtain an analytical expression for the cycle averaged energy @xmath164^{2}\right\ } , \ ] ] where @xmath165 $ ] is the complete elliptic integral of the second kind ( see appendix [ appendixb ] ) .this result confirms our prediction based on the description of the system in the rotating frame ( see figure [ fig2 ] ) .the cycle - averaged energy exhibits singularities at the zeros of the bessel function , i.e. , @xmath166 .this is a clear signature of a critical nonequilibrium behavior .( color online ) cycle - averaged dimensionless magnetization @xmath167 as a function of the driving amplitude @xmath90 .( a ) for parameters in the nonequilibrium paramagnetic phase @xmath141 , for @xmath109 , where @xmath161 .( b ) for parameters inside the ladder of ferromagnetic phases @xmath145 , for @xmath109 , where @xmath162 .the insets depict the second derivative of the cycle - averaged magnetization as a function of the driving amplitude @xmath90 .the parameters are @xmath109 and @xmath124 . ] finally , we calculate the cycle - averaged transverse magnetization in the floquet mode @xmath168 . by considering the extension of the hellmann - feynman theorem for floquet theory @xcite , we can compute the cycle - averaged magnetization @xmath169 in terms of derivatives of the quasienergies as follows : @xmath170^{\dagger}\left(\frac{\partial \mathcal{h}_k}{\partial g_0}\right)\psi^{(\pm)}_{k , m}(t ) \nonumber \\ & = -\int_{0}^{\pi}\frac{dk}{2\pi}\frac{\partial \varepsilon_{k , m}^{(\pm)}}{\partial g_0}=\mp\int_{0}^{\pi}\frac{dk}{\pi}\cos(2\phi_{k , m } ) . \end{aligned}\ ] ] figures [ fig4 ] and [ fig5 ] show the cycle - averaged expectation values of observables . similarly to the undriven case , the system exhibits nonanalyticities in the second derivative of observables as can be seen in the corresponding insets thus resembling a continuous phase transition .such nonanalyticities arise as a consequence of cdt @xcite , and therefore , from the gapless quasienergy spectrum .we have investigated the nonequilibrium critical behavior in a driven one - dimensional transverse ising model .we find that the fundamental signature of critical behavior is the existence of a gapless quasienergy spectrum , which is a direct consequence of cdt .the role of coherent destruction of tunneling in nonequilibrium qpt has been explored either theoretically or experimentally in the context of driven superfluidity @xcite . in this paper , we show that cdt induces a critical behavior which resembles a second - order qpt. in particular , the symmetry which is broken corresponds to a generalized parity in an extended hilbert space @xmath43 , where @xmath44 is the hilbert space of square integrable functions and @xmath45 is the space of time - periodic functions . in this nonadiabatic scenario ,the short time dynamics is governed by transient effects that tend to zero in the asymptotic limit as a consequence of destructive interference @xcite .the long - time dynamic is governed by the floquet modes , therefore the nonequilibrium quantum phases correspond to states which are synchronized with the external control . by means of a kramers - wannier self - duality transformation @xcite , it is possible to map hamiltonian equation into a dual periodically driven ising model with time - dependent exchange interaction .therefore , the study of the nonequilibrium qpt in the hamiltonian equation allows one to get a physical picture of the corresponding qpt in the dual model .a possible experimental implementation of our model could be achieved based on a configuration of superconductor quantum bits with programmable spin - spin interaction @xcite ; such a setup allows for a high degree of control of the system parameters .we anticipate that under an adiabatic change of the static local field @xmath171 and the driving amplitude @xmath172 our model could be interesting in the context of quantum annealing , as the effective hamiltonian equation corresponds to the @xmath89 model .another experimental setup can be realized by means of cold atoms @xcite , and in fully c - labeled sodium butyrate using liquid state nuclear magnetic resonance @xcite . the authors gratefully acknowledge discussions with a. b. bhattacherjee , g. engelhardt , m. hayn , v. a. leyton , c. nietner , m. plenio , b. regler , and m. vogl , and financial support from the daad and dfg grants br @xmath173 , @xmath174 , sfb @xmath175 , grk @xmath176 , and scha @xmath177 .in this section we introduce the fundamental tools used in the solution of the ising model following the methods and the notation of ref .@xcite .let us consider the hamiltonian equation in the case of even number of lattice sites @xmath0 . for convenience, we use the jordan - wigner representation of the pauli matrices @xmath178 under this representation of the angular momentum algebra , the parity operator equation acquires the form @xmath179 using this parity operator we are able to define projectors on the subspaces with an even @xmath17 and odd @xmath180 number of fermionic quasiparticles as follows @xmath181 the projectors @xmath182 satisfy the usual properties of orthogonal projection operators such as @xmath183 , @xmath184 , and @xmath185 . using these properties and the algebra of fermionic operators it is possible to show that the hamiltonian equation can be decomposed as follows @xmath186 where @xmath187 to perform the splitting we have defined antiperiodic boundary conditions in the even @xmath17 subspace @xmath188 and periodic boundary conditions in the odd @xmath180 subspace @xmath189 . in this paper we focus on the projection @xmath190 in the even subspace .translational invariance suggests the use of discrete fourier transform @xmath191 which is compatible with the antiperiodic boundary conditions when @xmath21 . the discrete fourier transform maps @xmath190 into hamiltonian equationhere we consider the critical behavior in a model described by the hamiltonian @xmath192 which is unitarily equivalent to the hamiltonian of an anisotropic @xmath89 spin chain in a transverse field @xcite .similarly to the ising model , after jordan - wigner transformation , and a discrete fourier transform , the hamiltonian equation in the even subspace ( the subspace with an even number of fermionic quasiparticles ) acquires the form @xmath193(c_k^{\dagger}c_k+c_{-k}^{\dagger}c_{-k})-2h\right\ } \nonumber \\ & & + \sum_{k>0}2 ( j_z - j_y)\sin k(c_k^{\dagger}c_{-k}^{\dagger}+c_{-k } c_{k } ) .\end{aligned}\ ] ] the diagonalization of this hamiltonian is completed after a bogoliubov transformation , @xmath194 where @xmath195 ^ 2+\left[\left(j_z - j_y\right)\sin k\right]^2}.\ ] ] the system exhibits an ising - like qpt along the lines @xmath196 and an anisotropic qpt along the line @xmath197 , providing that @xmath198 .the anisotropic transition is characterized by two ferromagnetic phases , i.e. , for @xmath199 the system is in a ferromagnetically ordered phase along the @xmath112 direction fm@xmath107 , while it is the other way around in the fm@xmath106 phase . similarly to ref .@xcite , we consider a reparametrization of the asymmetries @xmath200 where @xmath201 is a dimensionless parameter characterizing the degree of anisotropy in the @xmath202 plane . under this reparametrization, the ising - like critical lines correspond to @xmath203 , and the anisotropic transition occurs at @xmath204 , as long as @xmath205 .interestingly , in the absence of a transverse field , i.e. , for @xmath206 , the scaled ground - state energy can be written in the thermodynamic limit as @xmath207 , \end{aligned}\ ] ] where @xmath165 $ ] is the complete elliptic integral of the second kind @xcite .the scaled ground state energy exhibits a nonanalyticity of the second derivative at the critical line of the anisotropic transition @xmath204 @xcite , which is a generic characteristic of a second - order qpt @xcite .by using the bcs ansatz , eq . , we can solve the schrdinger equation for the ising model in terms of the solution of the schrdinger equation for an effective two - level system described by the bdg hamiltonian equation , which is parametrized by the quasimomentum @xmath208 .in the numerical calculation we assume that the system is prepared initially in the unoccupied state @xmath28 , which implies that @xmath209 .after numerical integration of the dynamical bdg equation , we find the spinor @xmath210 . to calculate the scaled expectation value of the transverse magnetization @xmath211 for a given system size @xmath0, we use the formula @xmath212^{\dagger}\sigma^{z}(k)\psi_{k}(t ) .\ ] ] in the last expression we have used the definition of @xmath213 given in equation . for example , to calculate the dynamics of the system for @xmath214 , we perform the numerical integration of equation for @xmath215 . based on the solution of this equation we find the solution of the schrdinger equation for hamiltonian equation using the bcs ansatz @xmath216 where @xmath217 and @xmath218 03 s. sachdev , _ quantum phase transitions _ ( cambridge university press , cambridge , england , 1999 ). m. w. johnson , et .al , nature ( london ) * 473 * , 194 ( 2011 ) .j. barreiro , m. mller , p. schindler , d. nigg , t. monz , m. chwalla , m. hennrich , c. f. roos , p. zoller , and r. blatt , nature ( london ) * 470 * , 486 ( 2011 ) . j. simon , b. s. bakr , r. ma , m. e. tai , p. m. preiss , and m. greiner , nature ( london ) * 472 * , 307 ( 2011 ) .r. coldea , d. a. tennant , e. m. wheeler , e. wawrzynska , d. prabhakaran , m. telling , k. habitch , p. smeibidl , and k. kiefer , science * 327 * , 177 ( 2010 ) .s. mostame and r. schtzhold , phys .* 101 * , 220501 ( 2008 ) .s. trotzky , p. cheinet , s. flling , m. feld , u. schnorrberger , a. m. rey , a. polkovnikov , e. a. demler , m. d. lukin , and i. bloch , science * 319 * , 295 ( 2008 ) .a. altland , v. gurarie , t. kriecherbauer , and a. polkovnikov , phys .a * 79 * , 042703 ( 2009 ) .a. polkovnikov , k. sengupta , a. silva , and m. vengalattore , rev .phys . * 83 * , 863 ( 2011 ) .w. h. zurek , u. dorner , and p. zoller , phys .lett . * 95 * , 105701 ( 2005 ) .j. gong , l. morales - molina , and p. hnggi , phys .103 * , 133002 ( 2009 ) . c. d. graf , g. weick , and e. mariani , europhys. lett . * 89 * , 40005 ( 2010 ) .a. eckardt , c. weiss , and m. holthaus , phys .* 95 * , 260404 ( 2005 ) .l. goren , e. mariani , and a. stern , phys .a. * 75 * , 063612 ( 2007 ) . h. lignier , c. sias , d. ciampini , y. singh , a. zenesini , o. morsch , and e. arimondo , phys .. lett . * 99 * , 220403 ( 2007 ) .n. h. lindner , g. refael , and v. galitski , nat .* 7 * , 490 ( 2011 ) .j. i. inoue , and a. tanaka , phys .lett . * 105 * , 017401 ( 2010 ) . v. m. bastidas , c. emary , b. regler and t. brandes , phys .* 108 * , 043003 ( 2012 ) l. jiang , t. kitagawa , j. alicea , a. r. akhmerov , d. pekker , g. refael , j. i. cirac , e. demler , m. d. lukin , and p. zoller , phys .106 * , 220402 ( 2011 ) j. h. shirley , phys . rev . * 138 * , b979 ( 1965 ) .s. ashhab , j. r. johansson , a. m. zagoskin , and f. nori , phys .a * 75 * , 063414 ( 2007 ) .t. dittrich , p. hnggi , g. ingold , b. kramer , g. schn and w. zwerger , _ quantum transport and dissipation _ ( wiley - vch , weinheim , 1998 ). m. grifoni and p. hnggi , phys .rep . * 304 * , 229 ( 1998 ) .f. grossmann , t. dittrich , p. jung , and p. hnggi , phys .lett . * 67 * , 516 ( 1991 ) .f. grossmann and p. hnggi , europhys . lett .* 18 * , 571 ( 1992 ) .j. dziarmaga , phys .lett . * 95 * , 245701 ( 2005 ) .v. mukherjee , u. divakaran , a. dutta , and d. sen , phys .b * 76 * , 174303 ( 2007 ) v. mukherjee , a. dutta , and d. sen , phys . rev .b * 77 * , 214427 ( 2008 ) t.w. b. kibble , j. phys . a * 9 * , 1387 ( 1976 ) ; phys .rep . * 67 * , 183 ( 1980 ) ; w. h. zurek , nature ( london ) * 317 * , 505 ( 1985 ) ; acta phys .b * 24 * , 1301 ( 1993 ) ; phys .rep . * 276 * , 177 ( 1996 ) .s. miyashita , k. saito , and h. de raedt , phys .. lett . * 80 * , 1525 ( 1998 ) k. hijii and s. miyashita , phys .a * 81 * , 013403 ( 2010 ) a. das , phys . rev .b * 82 * , 172402 ( 2010 ) s. bhattacharyya , a. das , and s. dasgupta , phys .b * 86 * , 054410 ( 2012 ) .a. russomanno , a. silva , and g. e. santoro , phys .. lett . * 109 * , 257201 ( 2012 ) .v. i. arnold , _ mathematical methods of classical mechanics _( springer - verlag , new york , 1978 ) .m. abramowitz and i. a. stegun , _ handbook of mathematical functions with formulas , graphs and mathematical tables _ , edited by m. abramowitz and i. a. stegun ( dover , new york , 1972 ) .e. lieb , t. schultz , and d. mattis , ann .( ny ) * 16 * , 407 ( 1961 ) e. barouch and b. m. mccoy , phys .rev . a * 3 * , 786 ( 1971 ) .j. e. bunder and r. h. mckenzie , phys .b * 60 * , 344 ( 1999 ) .y. niu , s. b. chung , c. h. hsu , i. mandal , s. raghu , and s. chakravarty , phys .b * 85 * , 035110 ( 2012 ) i. peschel , j. stat . mech . ( 2004 ) p12005. m. m. wolf , g. ortiz , f. verstraete , and j. i. cirac , phys .lett . * 97 * , 110403 ( 2006 ) j - s .lee , and a. k. khitrin , phys .a * 71 * , 062338 ( 2005 ) j - s .lee , t. adams , and a. k. khitrin , new journal of physics * 9 * , 83 ( 2007 )

<S> we establish a set of nonequilibrium quantum phase transitions in the ising model driven under monochromatic nonadiabatic modulation of the transverse field . </S><S> we show that besides the ising - like critical behavior , the system exhibits an anisotropic transition which is absent in equilibrium . </S><S> the nonequilibrium quantum phases correspond to states which are synchronized with the external control in the long - time dynamics . </S><S> one of the more intriguing hallmarks of many - body systems is that at zero temperature quantum fluctuations can drive the system to a drastic change of state , commonly known as a quantum phase transition ( qpt ) . </S><S> a paradigmatic model for qpts is the one - dimensional ising model @xcite . </S><S> recently , experimental realizations of one - dimensional spin chains have been suggested , where a quantum simulation of the system close to the phase transition is possible , and a wide freedom on the control of the parameters is achieved @xcite . the quantum control of many - body systems by a driving field has attracted considerable interest , both theoretical and experimental , with workers from very different communities beginning to look at driven models @xcite . </S><S> the possibility of manipulating the quantum state of a system by means of a classical external control allows one to explore novel states of matter and effective interactions which are absent in equilibrium @xcite . in the presence of an external control , quantum resonances and symmetries play an important role @xcite . </S><S> in particular , as a consequence of a generalized parity in the extended hilbert space @xcite , under the effect of periodic driving the tunneling can be slowed down or totally suppressed in a perfect coherent way , a phenomenon commonly referred to as coherent destruction of tunneling ( cdt ) @xcite . </S><S> rather recently , the extension of this concept to many - body systems has been addressed in the context of the mott - insulator - superfluid transition in ultracold systems both theoretically @xcite as well as experimentally @xcite , and in a two - mode bose - hubbard model with time - dependent self - interaction strength @xcite . </S><S> the dynamics of one - dimensional spin chains has been addressed extensively when the system is driven slowly through the critical point @xcite , where there is a diverging relaxation time and correlation length , and the dynamics can not be adiabatic in the thermodynamic limit . as a consequence of this , the final state of the system consists of ordered domains whose finite size depend upon the velocity of the parameter variation @xcite . </S><S> a nontrivial oscillation of the magnetization @xcite and the connection between symmetry and cdt @xcite has been investigated in a finite size periodically - driven ising model . </S><S> furthermore , under the effect of a nonadiabatic external control of the transverse field , the ising chain exhibits dynamical freezing of the response @xcite , and synchronization with the external driving in the asymptotic dynamics as a consequence of destructive interference in time @xcite . </S><S> our aim in this paper is to describe the nonequilibrium behavior of a one - dimensional ising model under the effect of a nonadiabatic monochromatic transverse field from the perspective of quantum criticality . in particular , we describe the dynamics by means of an effective hamiltonian which simulates the dynamics of an undriven system . </S><S> we show that in the asymptotic dynamics the nonequilibrium quantum phases correspond to states of the system which are synchronized with the driving . </S><S> in contrast to previous works @xcite , however , we describe the role of many - body cdt in the critical behavior by investigating signatures of criticality both in the laboratory frame as well as in the rotating frame . </S><S> the paper is organized as follows : in sec . </S><S> [ sectioni ] we discuss the equations of motion and describe quantum resonances by considering the excitation spectrum of the undriven system . in sec . </S><S> [ sectionii ] we describe the physics of the system by means of the rotating wave approximation ( rwa ) and discuss signatures of criticality based on the description of the quasienergy dispersion and cdt . in sec . </S><S> [ sectioniii ] we describe the quantum dynamics in the laboratory frame by considering the quantum evolution of the system when it is initialized in a paramagnetic ground state . furthermore , we study signatures of criticality by considering cycle - averaged expectation values of physical observables . </S><S> finally , a discussion of the results is presented in sec . </S><S> [ sectioniv ] . </S>

gr-qc0101065

for the past few decades , gravity of cosmic strings and gravitational waves from them have been investigated by many researchers@xcite@xcite .there are two physical contexts to investigate gravity of cosmic strings .one is the cosmological context and another is in the technical problem in general relativity : the treatment of concentrated line sources . in the cosmological context ,cosmic strings are topological defects associated with the symmetry breaking in unified theories@xcite . in the simplest case , the dynamics of cosmic stringsis governed by the nambu - goto action .if the self - gravity of strings is ignored ( test string case ) , the nambu - goto action admits oscillatory solutions .it is considered that these oscillations lead the emission of gravitational waves and strings gradually lose their kinetic energy by this emission@xcite .further , relic gravitational waves emitted by cosmic strings might be detectable by the gravitational wave detectors to be constructed in the near future@xcite . though this is a quite fascinating prediction , it is not clear whether or not it can be taken seriously . recently , by taking into account the self - gravity of the nambu - goto string , it is shown that the oscillatory behavior of a self - gravitating string coupled to gravitational wave is quite different from that of a test string@xcite .the dynamical degree of freedom concerning the perturbative oscillations of an infinite string is completely determined by that of gravitational waves and a self - gravitating infinite string does not oscillate spontaneously unlike an infinite test string . to give a completely precise dynamics of gravitating cosmic strings, we must solve the einstein equations and the evolution equations of the higgs and gauge fields from generic initial data , which is impossible at present .so it will be instructive to replace a cosmic string by a concentrated line source and clarify its gravity at first@xcite .however , it is known that the mathematical description of a self - gravitating thin string is delicate because the support of a string is a surface of co - dimension two . in general relativity , there is no simple prescription of an arbitrary concentrated line source where the metric becomes singular@xcite . in spite of these difficulties , a conical singularity , which appears in some exact solutions to the einstein equations ,is often regarded as a self - gravitating cosmic string .( for example , see ref.@xcite . )this is one of the idealizations of the concentrated matter sources .since a spacetime containing a straight gauge string has the asymptotically conical structure@xcite , this idealization seems plausible when the string thickness is much smaller than the other physical scale of interests . as an extension of this idealization, one might expect that an oscillating cosmic string can be also idealized by an oscillating conical singularity .however , as pointed out by unruh et al.@xcite , the worldsheet of a conical singularity is totally geodesic .this behavior is also seen in the time symmetric initial data of gravity of a loop conical singularity@xcite and similar results are also obtained in refs.@xcite .the dynamics of conical singularities is completely determined without their equations of motion .this geodesic property of a conical singularity shows that the idealization of cosmic strings by conical singularities might gives the different dynamics from that of cosmic strings in the early universe which have their finite thickness .further , this implies that the dynamics of cosmic strings is not idealized by that of conical singularities , though it seems natural to expect that loops of cosmic string can be regarded as conical singularities when the loop radius is much larger than the string thickness . in this paper ,we reconsider the above geodesic property pointed out in ref.@xcite using a time symmetric initial data containing a thick loop string .we show that the zero thickness of strings is essential to this geodesic property . because of the singular metric , gravity near the loop conical singularity is so strong that the loop itself is a geodesic on the initial surface . on the other hand ,the core of the thick loop string is not a geodesic on the initial surface and the curvature of a test loop on a flat space is naturally reproduced in the limit @xmath2 where @xmath3 is newton s gravitational constant .then , the geodesic property in ref.@xcite should not be taken seriously when we consider the dynamics of cosmic strings , since cosmic strings in the early universe have their own finite thickness .this also implies that we must take into account the effect of the string thickness when we discuss the dynamics of cosmic strings .this behavior of the gravity of a thick string loop also suggests that there is a criterion in the string thickness at which the linear perturbation theory around a flat background spacetime fails to describe the gravity of cosmic strings .we call this thickness `` _ critical thickness _ '' and evaluate this criterion in this paper . the geodesic property in ref.@xcite does not depend on the loop radius and its deficit angle . when the spacetime contains a sufficiently small loop conical singularity ,this loop is a geodesic on this spacetime and the other geodesics on this spacetime are quite different from those on the flat spacetime , on which geodesics are straight lines . on the other hand , gravity of a thick loop stringis not so strong as mentioned above .this suggests that the linear gravity is plausible if the string is sufficiently thick .then , for a fixed loop radius and a fixed deficit angle , there exists a critical thickness and the linear gravity on a flat background spacetime is valid if the string thickness is much greater than this critical thickness . to discuss the initial data of the gravity of a thick string loop and evaluate the above critical thickness, we first construct an example of the initial data in which the matter concentrates on the torus shell . since the gravity of a cosmic string has asymptotically conical structure , we use the initial data derived by frolov et al.@xcite as a solution outside the torus . using this example, we explicitly show that the core of the loop string is not a geodesic if the string has its finite thickness , while the loop approaches to a geodesic in the zero - thickness limit .though the matter distribution in this example is too artificial as a realistic cosmic string loop , our conclusion is also true even if the matter distribution is smooth .our example also shows that the initial data derived by frolov et al.@xcite may be regarded as the gravitational field produced by a smooth matter distribution .so , using this initial data as the geometry outside the matter field , we derive the above critical thickness of a string by comparing with a linear solution on the flat space background .this paper is organized as follows : in sec.[sec : general - formalizm ] , we briefly review the adm initial constraints and the treatment of the @xmath4-function distribution of the matter field on the initial surface to construct an example of the gravitational field produced by a loop matter distribution . in sec.[sec : loop - thick - string - initial ] , we construct an initial data of a thick loop string and discuss whether the string is a geodesic on the initial surface or not . we discuss the above critical thickness in sec.[sec : liniear ] . the final section ( sec.[sec : disucssion ] )is devoted to summary and discussions . throughout this paperwe work in relativistic units , so that the speed of light is equal to unity .in 4-dimensional spacetime @xmath5 , we consider a zero extrinsic curvature 3-dimensional spacelike submanifold @xmath6 .we denote the unit normal vector of @xmath7 in @xmath8 by @xmath9 .the induced metric @xmath10 on @xmath7 from the metric @xmath11 on @xmath8 is given by @xmath12 .the zero extrinsic curvature means the spacetime is momentarily static at @xmath6 .the adm initial - value constraints for a momentarily static hypersurface @xmath6 are reduced to@xcite @xmath13 where @xmath14 is the ricci scalar curvature of @xmath7 and @xmath15 is the energy density of the matter field on the hypersurface @xmath7 . in terms of a conformally transformed metric @xmath16 , the constraint ( [ hamiltonian_constraint-1 ] ) is given by @xmath17 where @xmath18 and @xmath19 are the covariant derivative and the scalar curvature associated with the metric @xmath20 , respectively . in this paper , we concentrate only on the geometry of @xmath7 using the constraint ( [ hamiltonian - constraint-2 ] ) ( or ( [ hamiltonian_constraint-1 ] ) ) . for simplicity ,we consider @xmath4-function distribution of matter energy density at the 2-dimensional hypersurface @xmath21 on @xmath7 to construct an example of solutions to eq.([hamiltonian - constraint-2 ] ) .to treat @xmath4-function distribution of the matter energy , we apply israel s condition@xcite on the initial surface , which is derived from the constraint ( [ hamiltonian - constraint-2 ] ) . to derive this condition, we first divide @xmath6 into two manifolds @xmath22 with the boundaries @xmath23 and then identify @xmath24 .this identification implies @xmath25 , where @xmath26 are the unit normal vector of @xmath23 on @xmath27 , respectively .the induced metrics @xmath28 on @xmath23 from @xmath29 are given by @xmath30 , respectively .since @xmath23 are diffeomorphic to each other , @xmath28 satisfy the condition @xmath31 by virtue of the constraint ( [ hamiltonian_constraint-1 ] ) , the @xmath4-function distribution of the matter energy density also gives the gap of the extrinsic curvatures @xmath32 of @xmath23 facing to @xmath27 .@xmath32 are defined by @xmath33 , where @xmath34 is the covariant derivative associated with the metric @xmath10 . introducing the gaussian normal coordinate @xmath35 in the neighborhood of @xmath21 by @xmath36 , the scalar curvature @xmath14 on @xmath7 is decomposed as follows : @xmath37 when @xmath38 is given by @xmath39 the integration of ( [ hamiltonian_constraint-1 ] ) over the infinitesimal interval of @xmath35 yields @xmath40 hence all conditions for the identification @xmath41 on the momentarily static initial data are ( [ constraint - junction - intrinsic ] ) and ( [ constraint - junction - extrinsic ] ) .here , we consider a momentarily static initial data of gravity produced by a thick loop string . in this paper , we assume that the vacuum region outside the loop is given by the initial data for a loop conical singularity derived by frolov et al.@xcite as mentioned in the introduction ( sec.[sec : introduction ] ) . in this section ,we first review the conformally flat version of the initial data for a loop conical singularity in ref.@xcite .second , replacing the conical singularity by a matter distribution , we construct an example of a gravitating thick loop string. then we discuss whether the loop is a geodesic on the initial surface or not .though we concentrate only on a conformally flat initial data , the similar behavior is also seen in the conformally non - flat version .( see appendix [ sec : c - non - flat ] . ) to discuss the initial condition for a loop conical singularity , we introduce the toroidal coordinate system .first , we consider the flat space metric @xmath42 defining the functions @xmath43 and @xmath44 by @xmath45 where @xmath46 and @xmath47 , the line element ( [ eucl - cylindrical ] ) becomes @xmath48 the geometrical meaning of the coordinates @xmath43 and @xmath44 is depicted in fig.[fig : fig1 ] . a @xmath43=constant surface is a torus with circular cross sections of radius @xmath49 ; the central axis of the tube forms a circle of radius @xmath50 in the plane @xmath51 .the ring @xmath52 , @xmath51 is the limit torus @xmath53 . to construct the initial data of a loop conical singularity at the limit torus @xmath53 , we remove a pair of spherical caps spanning the loop as shown in fig.[fig : fig2]@xcite .first , we fix an arbitrary angular deficit @xmath54 .we restrict @xmath44 to the range @xmath55 , identify these end points of this interval . applying a conformal factor @xmath56 to eq.([flat - toroidal ] ) , the physical line element associated with the metric @xmath10 is given by @xmath57 second , we rescale the coordinate @xmath44 by @xmath58 then the conformally transformed metric @xmath20 is given by @xmath59 though the exposed spherical faces have the same intrinsic geometry , their extrinsic curvatures have opposite sign in @xmath60 .this discontinuity disappears in @xmath6 if the conformal factor @xmath61 satisfies the following boundary conditions : 1 .@xmath61 is even in the inversion @xmath62 and smooth at @xmath63 ; 2 .@xmath64 at @xmath65 ; 3 . @xmath66constant at @xmath67 . by virtue of the condition ( i ) ,the extrinsic curvature of the surface @xmath63 in @xmath6 vanishes and no discontinuity is there .the condition ( ii ) guarantees that @xmath6 has the conical structure with the deficit angle @xmath68 at the limit tours @xmath53 .@xmath6 is asymptotically flat by virtue of the final condition ( iii ) . assuming @xmath69 is a killing vector on @xmath6 , eq.([hamiltonian - constraint-2 ] ) for the vacuum region @xmath70 in the coordinate system @xmath71 is given by @xmath72 the general solution to this equation with the boundary conditions ( i ) and ( ii ) has the form @xmath73 where @xmath74 is the second class legendre function and @xmath75 is the neumann factor . by comparing this with the expansion @xmath76one can easily see that @xmath77 is one of the simplest choice which continuously reduces to eq.([n - expand ] ) in the limit @xmath78 and satisfies the boundary conditions ( i ) and ( ii ) . from the asymptotic behavior of eq.([vacuum - solution-1 ] ) in the limit @xmath79 we can easily confirm that @xmath6 is asymptotically flat , i.e. , eq.([vacuum - solution-1 ] ) satisfies the boundary condition ( iii ) . since @xmath80 where @xmath81 the physical line element ( [ physical - three - metric ] ) is given by @xmath82 this shows that @xmath6 has asymptotically flat region with the adm mass @xmath83 here , we replace the above conical singularity by a matter distribution . for simplicity ,we consider an example in which the matter concentrates on the torus @xmath21 ( @xmath43=constant ) as eq.([tpp - def ] ) , where @xmath84 may depend on @xmath85 .we denote the outside the matter torus by @xmath86 and the inside by @xmath87 .@xmath86 includes asymptotically flat region and @xmath88 is given by eq.([physical - three - metric ] ) with the conformal factor ( [ vacuum - solution-1 ] ) .since we assume that @xmath87 is vacuum without conical singularities , the metric @xmath89 is given by @xmath90 where @xmath91 is a solution to eq.([hamiltonian - constraint - vacuum ] ) with @xmath92 .imposing the boundary conditions ( i ) , ( ii ) and @xmath92 , the solution to eq.([hamiltonian - constraint - vacuum ] ) is given by @xmath93 the coefficients @xmath94 in eq.([vacuum - solution - inside ] ) are determined by the boundary condition ( [ constraint - junction - intrinsic ] ) at @xmath21 . here, we consider the identification @xmath41 at the torus shell .we assume that @xmath95 is a killing vector on the whole initial surface @xmath7 and identify the orbits of @xmath95 at @xmath96 .since @xmath85 is a periodic coordinate with the period @xmath97 on each @xmath27 , we also identify the orbits of @xmath98 at @xmath41 . by these identifications ,the functions @xmath99 and @xmath85 are extended to smooth functions on @xmath7 . finally , we choose the coordinate function @xmath43 so that @xmath100 represent @xmath23 in @xmath27 , respectively .the intrinsic metrics @xmath28 on @xmath23 are given by @xmath101 where @xmath102 , respectively . then , eq . ( [ constraint - junction - intrinsic ] ) yields @xmath103 the first equation in eqs.([intrinsic - junctions ] )determineds @xmath104 as a function of @xmath105 and the deficit angle @xmath106 . the second one in eqs.([intrinsic - junctions ] ) determines the coefficients @xmath94 in @xmath107 : @xmath108 israel s junction ( [ constraint - junction - extrinsic ] ) determines the energy density @xmath84 as a function of @xmath85 : @xmath109 when the matter torus is sufficiently thin , i.e. , @xmath110 , @xmath111 behave @xmath112 from eqs.([phi+phi - asympt ] ) and eq.([a_n - relation ] ) , we easily see that @xmath113 .\label{rho - asymptotic - form - of - flat-2}\end{aligned}\ ] ] this shows that @xmath114 does not depend @xmath85 up to the order of @xmath115 .the leading term is just the relation of the proper line energy density of a straight string and the deficit angle .if the solution ( [ vacuum - solution-1 ] ) is extended to the limit torus @xmath53 , @xmath6 contains a loop conical singularity at the limit torus .this conical singularity is a geodesic on @xmath6 as pointed out by unruh et al.@xcite . to see this from eq.([vacuum - solution-1 ] ) , we consider the orbit of the killing vector @xmath95 and pay attention to the magnitude of the `` acceleration '' of this killing orbit . in the neighborhood of the limit torus @xmath53 , the norm @xmath116 of @xmath95 behaves @xmath117 then the magnitude @xmath118 of the `` acceleration '' is given by @xmath119 where @xmath120 is the proper distance from the killing orbits of @xmath95 with a finite @xmath43 to the limit torus @xmath53 .eq.([frolov - israel - unrhu - kappa ] ) shows that @xmath118 vanishes when @xmath121 .then the loop conical singularity is a geodesic on @xmath6 .we note that this is also true in the initial data of the conformally non - flat version derived in ref.@xcite .( see appendix [ sec : c - non - flat ] . ) here , we show that the core of a thick loop string ( the limit torus @xmath53 ) is not a geodesic on the initial surface @xmath6 containing a thick loop string in sec.[sec : thick - initial ] . in thisinitial data , the `` acceleration '' @xmath118 of the killing orbit of @xmath95 behaves @xmath122 in the limit @xmath65 .@xmath123 and @xmath124 are given by eqs.([a_n - relation ] ) and are constants determined by the outside deficit angle @xmath68 and the locus @xmath125 ( or @xmath105 ) of the matter torus .thus , the inside matter torus shell , the orbit of @xmath95 is not a geodesic on the whole initial surface @xmath7 if a string has its finite thickness .when the torus shell of the matter field is sufficiently thin ( @xmath110 ) , @xmath118 behaves @xmath126 this shows that the limit torus approaches to a geodesic on the initial surface @xmath6 as the string thickness approaches to zero , which is the similar situation to eq.([frolov - israel - unrhu - kappa ] ) .eq.([kappa - inside - thin - string ] ) also shows that @xmath127 when the outside deficit angle vanishes ( @xmath92 ) .this is the `` acceleration '' of the loop with the radius @xmath1 on a flat space , i.e. , a test string .thus , only by introducing the string thickness , we have obtained the natural behavior of @xmath118 which includes the case of a conical singularity and that of a test string .hence , we may say that the zero thickness of a string is the essential reason of the geodesic properties obtained in ref.@xcite .though the matter distribution in sec.([sec : thick - initial ] ) is too artificial for a gauge string loop , we may say that a cosmic string loop is not a geodesic if the loop has its finite proper length and if its energy density does not diverge at the core of the string ( @xmath53 ) . from the metric ( [ physical - three - metric ] ) , the proper length of the killing orbit of @xmath95 is @xmath128 .the core @xmath53 has a finite proper length only when @xmath129 . in this case , the right hand side of eq.([hamiltonian - constraint-2 ] ) behaves @xmath130 together with the hamiltonian constraint ( [ hamiltonian - constraint-2 ] ) , eq.([order - of - tpp ] ) yields that the energy density @xmath38 of matter fields does not contribute to @xmath131 nor @xmath132 of the conformal factor @xmath91 unless @xmath38 diverges at @xmath53 .when we solve eq.([hamiltonian - constraint-2 ] ) , we should impose the boundary conditions ( i ) , ( ii ) and @xmath92 from the regularity at the limit torus .then , the asymptotic form of @xmath107 in eq.([phi+phi - asympt ] ) is true up to the order of @xmath133 even if we consider the smooth distribution of @xmath38 .further , @xmath123 and @xmath124 in eq.([phi+phi - asympt ] ) remain constants .these constants @xmath123 and @xmath124 are determined by the boundary conditions at the surface of the matter distribution or the asymptotically flatness at the infinity as in our torus shell model .then , we obtain eq.([kappa - vacuum ] ) , again .thus , the limit torus @xmath53 must not be a geodesic even if the distribution of @xmath38 is smooth and a loop string has its finite proper length .in the last section , we have seen a thick loop energy distribution is not a geodesic on the initial surface , while a loop conical singularity is .the example in the last section also shows that the initial data derived in ref.@xcite may be regarded as the gravitational field produced a smooth matter distribution .further , as mentioned in the introduction ( sec.[sec : introduction ] ) , this also suggests that , for the fixed loop radius , there exists a criterion in the string thickness of the validity of a linear gravity .we call `` _ critical thickness _ ''. this critical thickness can be evaluated by the discrepancy of a linear gravity from the initial data in the previous section .the solution to the linearized hamiltonian constraint is derived by choosing the line element on @xmath7 so that @xmath134 the linearized hamiltonian constraint ( [ hamiltonian - constraint-2 ] ) is given by @xmath135 in the leading order of @xmath136 , where @xmath137 is the laplacian on the flat space .this is just newton s equation of gravity .however , we should comment that cosmic string does not generate the usual newton potential in the linearized einstein gravity because of its huge tension .we must note that @xmath136 in ( [ newton - eq ] ) does not have the usual meaning of the newton potential .we consider the matter energy density distribution given by @xmath138 .@xmath139 is the line energy density of the string and @xmath140 is the @xmath4-function whose support is on the loop with radius @xmath1 in the equatorial plane @xmath51 .the solution to eq.([newton - eq ] ) is given by @xmath141 where @xmath142 is the first class complete elliptic integral .to derive this solution ( [ newton - pot ] ) , we identified the line energy density @xmath139 with the hiscock mass @xmath143@xcite . comparing with eq.([asymptotic - metric ] ) and eq.([newton - line - element ] ), we easily see that @xmath144 coincides with @xmath136 in the asymptotically flat region @xmath145 . as an example , @xmath146 and @xmath136 on the equatorial plane with @xmath147are shown in fig.[fig : fig3 ] .we evaluate the validity of the linear perturbation theory with the flat space background by @xmath148 at the same circumference radius of the killing orbit of @xmath95 .we evaluate @xmath4 on the equatorial plane ( @xmath51 or @xmath149 ) .the discrepancy @xmath4 diverges at the locus of the loop due to the divergence in @xmath136 : @xmath150 while @xmath151 in the asymptotically flat region .this means that @xmath4 becomes order of unity at some points .these points give the `` critical thickness '' .if the string thickness is smaller than this critical thickness , we may regard that the linear perturbation theory on a flat space fails to describe gravity of a loop string .now , we evaluate this criterion .there are two circumference radii on the equatorial plane at which @xmath152 .we denote these by @xmath153 .@xmath154 is in the region @xmath155 ( @xmath156 ) and@xmath154 is in @xmath157 ( @xmath158 ) .we evaluate the critical thickness in the string thickness by @xmath159 .if a loop string has the larger thickness than @xmath160 , we may regard that the linear perturbation with a flat space background is plausible .the critical thickness @xmath160 for each deficit angle are shown in fig.[fig : fig4 ] .fig.[fig : fig4 ] shows that the linear perturbation theory with a flat space background is plausible when its thickness is larger than @xmath161 for @xmath162 ( gut string case ) .this critical thickness is also roughly estimated as follows : the divergence in @xmath4 is due to the logarithmic one in @xmath163 . since @xmath164 and @xmath165 near the loop , we obtain @xmath166 for @xmath167 . though this estimation is about one quarter of the result shown in fig.[fig : fig4 ] , it will be due to higher order corrections in @xmath136 or @xmath146 .we have considered the initial data of gravity produced by a loop string and showed that the core of a thick string is not a geodesic on the initial surface , while a loop conical singularity is . to see this, we first consider an example in which the matter concentrates on the torus surface .this geodesic property is unchanged if a loop string has its finite proper length and the matter energy density does not diverge .then , we may say that the geodesic property obtained in ref.@xcite should not be taken seriously when we consider the dynamics of cosmic strings in the early universe , because realistic cosmic strings have their own finite thickness .further , using the above example , we have seen that the loop approaches to a geodesic on the initial surface in the zero thickness limit .this behavior in our example also suggests that the metric perturbation on a flat space background loses its validity when a string is sufficiently thin .we also considered the critical thickness which is the criterion of the validity of the linear gravity around a flat space background .we evaluate this criterion using the circumference radii at which the discrepancy @xmath4 defined by eq.([delta - def ] ) becomes unity .fig.[fig : fig4 ] shows that a linear perturbation is plausible when the string thickness is larger than @xmath168 where @xmath1 is the string bending scale .it is not clear from the arguments in this paper whether all behaviors of the `` acceleration '' @xmath118 of the killing orbit in our example are maintained in generic situations . to confirm this, we must solve the hamiltonian constraint ( [ hamiltonian - constraint-2 ] ) with smooth matter energy distribution , numerically .however , we expect that these are also true even when the matter distribution is smooth because the coefficients @xmath123 and @xmath124 in eq.([kappa - vacuum ] ) are completely determined by the regularity at the limit torus and the asymptotically conical structure .fig.[fig : fig3](a ) also shows that there is the discrepancy of @xmath136 and @xmath146 in the region near loop axis @xmath51. this will be due to the degree of freedom of gravitational waves on the initial surface .this speculation is also suggested the fact that the usual newton potential does not generated by a cosmic string because of its huge tension as commented in the previous section . to confirm this speculation, we must solve the time evolution from this initial data .however , @xmath4 in this region is smaller than unity and this discrepancy will be described by the linear perturbation with the minkowski spacetime background .so this discrepancy is not important for the above criterion . in ref.@xcite , we discussed the dynamics of a fluctuating infinite string and showed that the displacement of a string vanishes for a zero thickness string .these results in ref.@xcite is due to the same situation discussed here . from the viewpoint of the idealization problem ofconcentrated line sources in general relativity , we may say that the thickness should be remained finite when we investigate the dynamics of self - gravitating strings and their gravitational wave emission .we must note that our conclusion here does not contradict to our conclusion in ref.@xcite : there is no dynamical degree of freedom of the free oscillations of a nambu - goto string within the first order with respect to its oscillation amplitude . in ref.@xcite, we have carefully excluded the gauge freedom of perturbations .the result obtained here suggests that we may use the linear perturbation theory with minkowski background if the string thickness is much larger than the above critical thickness , though we must bear in our mind that the effect of higgs and gauge fields constructing cosmic string will dominate in a realistic cosmic string . in this case, we have to exclude gauge freedom of the perturbations carefully as in ref.@xcite .this will be our future work .finally , we must also comment that our arguments here and in ref.@xcite are those for a string in the situation where its bending scale is much larger than its thickness .so , from the conclusion here , we can not say anything about the geometry of near cusps or kinks , which may arise in the complex dynamics of cosmic strings@xcite .when cusps or kinks appear by the dynamics from large loop , we will have to take into account the dynamics of higgs and gauge fields .this will be not the problem in general relativity but the field theory itself .the author thanks prof .minoru omote , prof .takashi mishima and prof .akio hosoya for their continuous encouragement .in this appendix , we consider the conformally non - flat version of the initial data discussed by frolov et al@xcite and show that , even in this initial data , a thick loop energy distribution is not a geodesic on the initial surface @xmath7 nevertheless a conical singurality is .the conformally non - flat version of the initial data discussed by frolov et al@xcite is characterized by the conformally transformed metric @xmath20 @xmath169 in this initial data , there are two unknown functions @xmath170 and the conformal factor @xmath61 and the hamiltonian constraint ( [ hamiltonian - constraint-2 ] ) is given by @xmath171 as a vacuum solution to eq.([hamil - con - toroical - non - flat ] ) with a loop conical singularity , @xmath172 is derived in ref.@xcite , where @xmath173 and @xmath174 denote the lesser and greater of @xmath175 and @xmath176 are given by @xmath177 eq.([51 ] ) guarantees that the limit torus @xmath53 is a conical singularity and the axis @xmath178 is not@xcite . in the solution given by eqs.([51])-([67 ] ) , we easily see that the conical singularity at @xmath53 is a geodesic on the initial surface @xmath7 . to see this, we consider the region @xmath179 . in this region ,eq.([51 ] ) yields @xmath180 and eq.([phi - sol - nonflat - vacuum ] ) is given by @xmath181 the `` acceleration '' of the killing orbit of @xmath95 behaves @xmath182 where the proper distance @xmath183 from the killing orbit with a finite @xmath43 to the conical singularity ( @xmath53 ) .this shows that the loop conical singularity is a geodesic on @xmath7 . here, we show that the limit torus @xmath53 is not a geodesic when we replace the above conical singularity by the matter distribution with the finite energy . for simplicity ,we consider a @xmath4-function distribution of a matter energy density at the torus @xmath184 , denoted by @xmath21 , as in the main text .the metric @xmath185 on @xmath70 ( outside the torus ) is given by eqs.([toroidal - nonflat-2 ] ) , ( [ 51 ] ) , and ( [ phi - sol - nonflat - vacuum ] ) .on @xmath186 , we assume that the metric @xmath10 is given by @xmath187 and eq.([vacuum - solution - inside ] ) . the normal vectors @xmath26 of @xmath23 are given by @xmath188 where @xmath111 and @xmath189 are metric functions @xmath91 and @xmath190 on the boundaries @xmath100 , i.e. , @xmath23 , respectively . the induced metrics @xmath28 on @xmath23 are given by @xmath191 the trace of the extrinsic curvatures @xmath192 of @xmath23 in @xmath27 are given by @xmath193 we only consider the case @xmath194 .then we may use @xmath195 , @xmath196 .@xmath197 is given by eq.([phi - near - string ] ) with @xmath198 and @xmath107 is given by eq.([vacuum - solution - inside ] ) with @xmath125 .the junction condition ( [ constraint - junction - intrinsic ] ) for the intrinsic metric yields @xmath199 as in the conformally flat version , the first condition gives the relation of @xmath200. the second condition determines the coefficients @xmath94 in eq.([vacuum - solution - inside ] ) : @xmath201 the junction condition ( [ constraint - junction - extrinsic ] ) for the extrinsic curvature @xmath202 gives the energy density on the torus shell as in the conformally flat version in the main text . in this initial data, we easily see that the limit torus @xmath53 in this model is not a geodesic on @xmath6 ( on @xmath87 ) .since the metric on @xmath186 is also given by eq.([inside - physical - metric ] ) and eq.([vacuum - solution - inside ] ) , the `` acceleration '' @xmath118 of the killing orbit of @xmath95 at @xmath53 is given by eq.([kappa - vacuum ] ) , again .since the coefficients @xmath94 in eq.([an_in_cnf - version ] ) are constants determined by the locus of the torus matter shell , we conclude that the limit torus @xmath53 is not a geodesic on @xmath6 . as in the case in sec.[sec : is - a - loop - string - is - geodesic ] , when the matter torus shell is sufficiently thin ( @xmath110 ) , @xmath118 behaves @xmath203 where @xmath204 is the proper distance from @xmath125 to the limit torus @xmath53 .eq.([kappa - conformally - nonflat - asympt ] ) shows that when the thickness of the string is sufficiently smaller than its curvature radius ( @xmath205 ) , the loop approaches to a geodesic on @xmath6 .these behaviors are same as that obtained in sec.[sec : is - a - loop - string - is - geodesic ] .99 a. vilenkin and e.p.s .shellard , _ `` cosmic strings and other topological defects '' _ , ( cambridge university press , cambridge , england , 1994 ) .e.w . kolb and m.s .turner , _ the early universe _ , ( addison - wesly publishing company , 1993 ) .a. vilenkin , and a.e .everett phys .lett . * 48 * 1867 , ( 1982 ) .t. vachaspati , a.e .everett and a. vilenkin , phys .rev . d*30 * 2046 , ( 1984 ) .t. vachaspati , nucl .b*277 * , 593 , ( 1986 ) .m. sakellariadou , phys .d*42 * , 345 , ( 1990 ) .r.r .. caldwell and b. allen , _ ibid ._ , * 45 * , 3447 , ( 1992 ) .p. casper and b. allen , _ ibid ._ , * 52 * , 4337 , ( 1995 ) . b. allen and a.c .ottewill , phys .rev . d*63 * , 063507 , ( 2001 ) .battye and e.p.s .shellard , class .quantum grav .* 13 * a239 , ( 1996 ) .caldwell , r.a .battye and e.p.s .shellard , phys . rev . d*54 * , 7146 , ( 1996 ) .k. nakamura a. ishibashi and h. ishihara , phys . rev . d*62 * , 101502r , ( 2000 ) .k. nakamura and h. ishihara , _ ibid ._ , * 63 * , 127501 , ( 2001 ) .w. israel , phys .rev . d*15 * , 935 , ( 1977 ) .m. hindmarsh , phys .b*251 * , 28 , ( 1990 ) .r. gregory , d. haws and d. garfinkle , phys .rev . d*42 * , 343 , ( 1990 ) .b. carter , phys .* 74 * , 3098 , ( 1995 ) .battye and b. carter , phys .b*357 * , 29 , ( 1995 ) .b. boisseau , c. charmousis and b. linet , _ ibid ._ , * 55 * , 616 , ( 1997 ). c.j.s .clarke , j.a .vickers and j.p .wilson , class .quantum grav . , * 13 * , 2485 , ( 1996 ) . j.a .vickers and j.p .wilson , _ ibid ._ , * 16 * , 579 , ( 1999 ). d. garfinkle , _ ibid ._ , * 16 * , 4101 , ( 1999 )battye and b. carter , _ ibid ._ , * 17 * , 3325 , ( 2000 ) .r. geroch and j. traschen , phys .rev . d*36 * , 1017 , ( 1987 ) .w. kinnersley and m. walker , phys .rev . d*2 * , 1359 , ( 1970 ) .d. kramer , h. stephani , m.a.h . maccallum and e. herlt , _ exact solutions of einstein s field equations _( cambridge : cambridge university press , 1980 ) .d. garfinkle , phys .rev . d*32 * , 1323 , ( 1985 ) .r. gregory , phys .59 * , 740,(1987 ) .t. futamase and d. garfinkle , phys . rev . d*37 * , 2086 , ( 1988 ) . w.g .unruh , g. hayward , w. israel and d. mcmanus , phys .lett . * 62 * , 2897 , ( 1989 ) .frolov , w. israel and w.g .unruh , phys ., d*39 * , 1084 , ( 1989 ) . b. boisseau , c. charmousis , and b. linet , phys .rev . d*55 * , 616 , ( 1997 ) .wald _ general relativity _( univ . of chicago press , chicago , 1984 ) .w. israel , nuovo cimento b*44 * ( 1966 ) , 1 .t. damour and a. vilenkin , phys .rev . lett . * 85 * , 3761 , ( 2000 ) .

<S> the initial data of the gravitational field produced by a loop thick string is considered . </S><S> we show that a thick loop is not a geodesic on the initial hypersurface , while a loop conical singularity is . </S><S> this suggests that there is the `` _ critical thickness _ '' of a string , at which the linear perturbation theory with a flat space background fails to describe the gravity of a loop cosmic string . using the above initial data </S><S> , we also show that the linear perturbation around flat space is plausible if the string thickness is larger than @xmath0 , where @xmath1 is the curvature radius of the loop </S><S> . [ section ] [ section ] [ section ] [ section ] [ section ] </S>

1104.0942

use of personal social networks to gather information is fundamental to purchasing behavior @xcite .it is something so common in our daily routine that we usually do not even make a note of it .when we make a purchase from a retail store , we often speak beforehand to the shopkeeper about suitable products .when we need to purchase something we are unfamiliar with , we consult our friends and family for advice .when we purchase a popular new product , we have an urge to tell everyone we know about it .although personal social networks are implicit in the offline shopping experience , their introduction to the online world is a relatively new phenomenon .e - commerce websites , such as amazon , ebay and epinions , have successfully integrated product reviews , recommendations , search and product comparison , but they have been much slower at adopting social networking features as a part of customer experience .recommendation engines and product comparison sites help consumers discover new products and receive more accurate evaluations , however they can not completely substitute for the personalized recommendations and information that one receives from a friend or relative .basic behavioral psychology drives consumers to value and trust their friends purchasing decisions more than anonymous opinions .for example , a lucid marketing survey found that 68% of individuals consulted friends and relatives before purchasing home electronics @xcite .understanding how social networks are used and how they shape purchasing decisions is one of the fundamental interests of e - commerce .only recently have social networks been used in e - commerce applications to some success .for example , group purchasing companies such as groupon and livingsocial allow consumers to come together to buy products in bulk and save money , while social shopping sites such as kaboodle provide consumers the ability to share shopping lists with each other .the use of social networks in online shopping provides marketers and businesses with new revenue opportunities , while providing consumers with product information and both economic and social rewards for sharing @xcite .* present work .* when discussing the relationship between electronic commerce and social networks , various questions come to mind .how do friends influence consumer purchasing decisions and product adoption ?what factors influence the success of word - of - mouth product recommendations ?how does social influence and reputation affect commercial activity ?in this paper , we will address these questions through a detailed study of the world s largest e - commerce website taobao .the fundamental process we focus upon throughout this study is what we term _ information passing _ : an individual purchases a product , then messages a friend , what is the likelihood that the friend will then purchase the product ? where will he purchase it from ? understanding the flow of social influence in commerce networks is an important question .for example , information passing provides insight into how companies can structure online viral marketing campaigns to target consumers .it can also be used to optimize algorithms within product recommendation engines .however important information passing is to electronic commerce , it still has not been well studied on a large scale due to the inaccessibility of suitable data . to facilitate our research , we obtained a dataset describing the behavior of one million users in the world s largest e - commerce network taobao .taobao connects buyers and sellers , and provides an integrated instant messaging platform for communication among all its users . by modeling taobao as a network of three types of edges ( trades , messages , contacts ) , we are able to directly study how social communication and commercial transactions are interrelated in an online setting .our study provides insights into three main aspects of the social - commercial relationship : information passing , the price of trust , and consumer choice prediction .we begin our study of social commerce by quantifying the presence of information passing through analysis of triadic closure processes .we show that the influence of information passing is directly proportional to message strength , and is inversely proportional to product price , as well as the time between the purchase and the recommendation .additionally , we explain how information passing varies greatly between different product categories .we then investigate the general edge formation process in the context of directed triadic closure , and demonstrate that the formation of triad - closing message and trade edges is highly dependent upon user roles ( buyer or seller ) .our results indicate that information passing via buyer - buyer communication is one of the primary drivers of purchasing .a subtle point regarding information passing is that the spread of product recommendations , through word - of - mouth , inherently relies upon a notion of buyer - buyer _trust_. trust , from the perspective of social psychology , can be defined as perceived credibility or benevolence to the target @xcite . in the context of electronic marketplaces , buyer - seller trust ,most directly encapsulated by seller reputations and ratings , are the natural concept to study .a fundamental idea behind the nature of trust is its price .how much extra will a buyer pay for transaction security with a highly - rated seller ?although an intuitive idea , initial studies did not find evidence for a price of trust @xcite , and only recently has a price of trust been established in small , controlled , and single product settings @xcite . in our study , we analyze transaction information across over 10,000 products . using the overall customer satisfaction ( i.e. , average rating ) of the seller, we observe a small but super - linear effect of the seller rating upon the price premium they can charge and still engage in transactions .to further study the relationship between social networks and consumer behavior , we then consider the question , `` how does an online consumer decide upon a seller to purchase from when there are many sellers offering the same product ? ''we model this question of _ consumer choice _ through a machine learning task and predict which particular seller a buyer will purchase from , given that the sellers all offer the same relevant product . utilizing primarily social networking features, we construct a model that can predict , for the case of a buyer choosing from among 10 possible sellers , the correct seller 42% of the time , approximately 4 times better than baseline .we also contrast a variety of feature sets ( both graph - based and product / seller metadata ) , and demonstrate that the social graph is the most important feature in predicting consumer choice . in particular , the social graph is a far better determinant of consumer behavior than metadata features such as seller reputation or product price .our results nicely connect to granovetter s work , which argues that economic transactions are embedded in dynamic social networks , and that an individual s social graph dictates how they choose sellers to transact with @xcite. * further related work . * [ sec : related ] for all of the importance of social networks in consumer shopping , though , their use in electronic commerce still is not well understood .previous research examined the use of social networks in e - commerce , but has mostly focused upon one aspect of the use of social networks , such as product recommendations @xcite , product recommendation engines @xcite , or have been based upon a limited set of data @xcite .the electronic marketplace ebay is perhaps the most well studied e - commerce website .various aspects of ebay including auction efficiency @xcite , product recommendations @xcite , seller strategies @xcite , and summarization @xcite , have been studied .closely related to our work on consumer choice prediction , wu and bolivar created a model to predict item purchase probability @xcite .the primary difference here is that we utilize the social networks of the buyers and sellers , along with product and user metadata , to perform consumer choice prediction , whereas they consider only seller and product information .our study of information passing and triadic closure builds upon classical works by rapoport @xcite and granovetter @xcite .triadic closure has been explored in various settings : community growth @xcite , link prediction @xcite , signed networks @xcite , and social @xcite and information @xcite network evolution .in contrast , we demonstrate the existence of implicit recommendation behavior in a network not specifically designed for information passing .the paper proceeds as follows . in section 2, we describe the taobao dataset . in section 3, we analyze dyadic relationships in the network . in section 4, we provide a detailed analysis of information passing and directed triadic closure processes . in section 5, we quantify a price for trust . andlast , in section 6 , we model the consumer choice prediction problem .the data we use in this study comes from the chinese website taobao , one of the world s largest electronic marketplaces , with over 370 million registered users at the end of 2010 . although transactions on taobao can be either business - to - consumer , business - to - business , or consumer - to - consumer , the bulk of the products are sold by online storefronts operated by small businesses or individuals . perhaps the most unique aspect of taobao is its integrated in - browser instant messaging platform , which allows us to correlate users communication patterns and purchasing behavior .any user can purchase goods from other users , add other users onto their contact list , and message other users .note that non - contacts can message each other as well ..dataset statistics .[ cols= " < , > , > , > , > " , ] * experimental results . *table [ table : predresults ] gives an overview of our experimental results where we compare the models using various feature sets and the baselines .first , we note that the model trained on all 23 features gives a 79% improvement over the p@1 of the random baseline and a 38% improvement over random s mrr .the model also gives a 13% improvement over the mostmsg baseline and a 93% improvement over the minprice baseline .for all our evaluation metrics , the model displays significantly better performance than the 3 baselines .it is interesting to note the poor performance of minprice compared to mostmsg .this suggests that communication links and the social graph are essential to understanding how consumers make purchasing decisions in social commerce networks .note that a natural explanation for the importance of the social graph can be that a buyer first messages a seller , then immediately trades with him .however , we control for this by discarding all communication on and after the trade date . to evaluate prediction performance in more detail , we graph the p@1 and mrr of the full model ( labeled as _ svm _ ) and the 3 baselines versus the number of sellers in the buyer - seller cluster ( i.e. , how many different sellers a buyer can choose from ) in figures [ fig : prediction](a),(e ) .as expected , the performance of all models and baselines decreases as the number of sellers in the cluster increases .observe that the performance gap between the full model and mostmsg , the strongest baseline , widens as the number of sellers to choose from increases and prediction becomes more difficult .if we look at the proportional p@1 improvement of the full model over mostmsg , we see only a 4.5% improvement when there are 2 sellers , but a 39.5% improvement when there are 10 sellers . in particular, we would like to highlight the full model s strong p@1 of 42.1% for the challenging prediction task with 10 sellers .in general , the full power of the model is not realized until the prediction problem becomes difficult for simple rule - based heuristics . *different feature sets . *having constructed a successful predictive model , we now ask , `` what features are most valuable when modeling consumer choice ? '' in the following set of experiments , we contrast the performance of svm models trained on different sets of graph and metadata features in order to better understand how consumers make purchasing decisions . when doing prediction , are network features as valuable as metadata features such as product price and seller rating ?we first compare the performance of a svm trained on network ( graph ) features versus a svm trained on metadata ( seller profile and product description ) features .we find that the p@1 of the graph features svm is only slightly worse than the full model , whereas the metadata features svm is not much better than random , as displayed in table [ table : predresults ] .most notably , there is a large performance gap between the graph features svm compared to the metadata features svm , as illustrated in figures [ fig : prediction](d),(h ) .this implies that prediction using only seller and product information is inadequate , the social and trade networks in the neighborhood of the buyer and seller must be taken into account .buyers likely do not just use seller profile and product information when making purchasing decisions . given that network features are essential when predicting consumer choice ,what type of network features are more valuable for prediction : direct features or indirect features ( i.e. , clustering coefficients , pageranks ) ? for this experiment , we train svm models on direct and indirect network features , the different graph feature classes we use are listed in table [ table : featureset ] . note that when comparing direct and indirect network features , we include all metadata features in both sets . figures [ fig : prediction](c),(g ) illustrate the large performance gap between the _ meta + direct _ svm vs the _meta + indirect _svm , labeled as _ direct _ and _ indirect _ respectively in the figures .we observe that direct graph information provided by buyer - seller edges is significantly more valuable than the collective information provided by other edges in the local neighborhoods .this is not surprising because powerful information content is present in direct edges .historical buyer - seller message volume can be indicative of an existing social relationship or historical product queries , while historical buyer - seller trade volume can be indicative of customer loyalty and trust with the seller .which network ( contact , message , trade ) is most useful for predicting consumer choice ? in this section , we contrast svm models trained on each of the separate taobao networks .we include all metadata features with each set of network features in this experiment as well , prediction results are displayed in table [ table : predresults ] .the performance of the 3 network feature sets versus the number of sellers in the cluster are displayed in figures [ fig : prediction](b),(f ) .our experiment demonstrates that the message network is the most valuable network to utilize when predicting consumer choice .one possible explanation for this finding is that historical message volume is an indicator of familiarity between buyer and seller , i.e. , an existing trust relationship between buyer and seller .historical message volume can also indicate previous potential purchase interest .we also observe that prediction with the trade network is slightly better than that with the contact network .this suggests that customer loyalty , the primary trade network feature we use , is a more important indicator of consumer choice than the network of contacts .our explanation for this is that , inherently , contact links are less valuable than trade or message links in such social networks .it takes little effort to add someone to a friends list , as it is a one - time operation .in contrast , maintaining a conversation requires an investment of time and mutual interest on the part of both parties .forming a trade link is arguably the most costly as it requires currency and an actual transaction to make the connection .* per - category performance .* after performing prediction with a single svm ranking model , the next question to ask is , `` can we perform better prediction through the use of multiple models ? '' to answer this question , we segment all historical transactions in our dataset into their respective product categories , and train separate svm ranking models for each category .the results of our model testing are displayed as _ aggregated _ in figure [ fig : prediction](a),(e ) . as expected , the aggregated performance of the category svms is slightly better than the single full svm , with a p@1 of 0.58 compared to 0.56 .our study of consumer choice prediction demonstrates that in social commerce sites such as taobao , user communication and social activity is the primary influence upon consumer choice . utilizing primarily social networking features , we are able to construct an svm model that can predict , for the case of a buyer choosing from among 10 possible sellers , the correct seller 42% of the time , approximately 4 times better than random . when faced with a selection of substitute goods offered by different sellers , buyers will not just choose their preferred seller through simple heuristics regarding price or rating .we can imagine that buyers utilize many sources of information ( seller history , advice of friends , seller s messages ) , and each buyer processes the information in their own way in order to make a personal purchasing decision .although we can not say with certainty what buyers are thinking , we can definitively state that the social graph in which the buyer and seller are embedded is the best feature to look at when predicting consumer choice .our work analyzes the activities of one million users of the chinese social commerce site taobao . through the study of directed closure rules, we empirically verify that implicit information passing is present in the taobao network , and show that communication between buyers is a fundamental driver of purchasing activity .we then investigate the directed triadic closure process and explain how link formation is highly dependent upon the distribution of buyer / seller roles for the nodes of a social commerce network .third , we use taobao review data to demonstrate how high seller ratings are associated with product price premiums , and thus quantify a price for trust .finally , we develop a machine learning model to accurately predict consumer choice , and demonstrate that the social network is the most important feature in predicting how consumers choose their transaction partners .we hope that our study will motivate future research into social shopping , as well as give impetus to established e - commerce companies to add more social networking features .future areas of related study include : analysis of user browsing data to develop refined consumer choice models for social commerce , study of information passing while factoring in both buyer - buyer and buyer - seller trust relationships , and viral marketing to influence consumer choice in social commerce .

<S> while social interactions are critical to understanding consumer behavior , the relationship between social and commerce networks has not been explored on a large scale . </S><S> we analyze taobao , a chinese consumer marketplace that is the world s largest e - commerce website . </S><S> what sets taobao apart from its competitors is its integrated instant messaging tool , which buyers can use to ask sellers about products or ask other buyers for advice . in our study , we focus on how an individual s commercial transactions are embedded in their social graphs . by studying triads and the directed closure process </S><S> , we quantify the presence of information passing and gain insights into when different types of links form in the network . using seller ratings and review information </S><S> , we then quantify a price of trust . </S><S> how much will a consumer pay for transaction with a trusted seller ? </S><S> we conclude by modeling this consumer choice problem : if a buyer wishes to purchase a particular product , how does ( s)he decide which store to purchase it from ? by analyzing the performance of various feature sets in an information retrieval setting , we demonstrate how the social graph factors into understanding consumer behavior . </S><S> = 10000 = 10000 * categories and subject descriptors : * h.2.8 database management : database applications data mining * general terms : * measurement ; experimentation . * keywords : * e - commerce , viral marketing , recommender systems , triadic closure , price of trust . </S>

nucl-th0601014

we would like to thank pavel grigorov for technical assistance .remind that a commutation relation between two noncommuting operators imposes a definite constraint upon their dispersions , i.e. , the mean square deviations of these quantities from the corresponding expectation values .in fact , let us consider the two hermitean operators @xmath159 and @xmath160 that meet @xmath161 = \imath \hat c\ , \eqno({\rm a.1})\ ] ] where @xmath162 is also an hermitean operator . in particular , if @xmath163 and @xmath164 , then the operator @xmath165 . ] by definition , the expectation values with respect to an arbitrary state @xmath166 are @xmath167 let us introduce the deviations @xmath168 obviously they satisfy the relation @xmath169 = \imath \hat c\ .\eqno({\rm a.4})\ ] ] from eq.(a.4 ) it follows the uncertainty relation @xmath170 note also that for operator @xmath171 , @xmath172 if @xmath173 . in the case of interest for the relativecoordinates @xmath174 and the canonically conjugate momenta @xmath175 the corresponding relation looks as @xmath176 or @xmath177 for any state @xmath178 normalized to . here , in accordance with eq.(a.6 ) , @xmath179 and @xmath180 since @xmath181 .thus , the general result ( a.5 ) leads to the condition ( a.7b ) for the pair @xmath182 and @xmath183 that obeys the commutation rules , @xmath184 = \imath { \delta}_{jk } \frac{a-1}{a } \quad ( j , k = 1,2,3 ) . \ \eqno({\rm a.10})\ ] ] eqs.(a.10 ) are different from @xmath185 = \imath { \delta}_{jk } \quad ( j , k = 1,2,3 ) .\ \eqno({\rm a.11})\ ] ] by the way , this fact means that the transformation @xmath186 and @xmath187 is nonunitary , i.e. , it can not be performed via a unitary operator @xmath188 , viz . , putting @xmath189 and @xmath190 . under the latterthe commutation relations would not change .it follows from ( a.5 ) and ( a.11 ) that @xmath191 for the expectation values @xmath192 and @xmath193 in the shell model ground state @xmath166 ( in general , any state @xmath166 ) such that @xmath194 .so , we have relation ( a.7b ) for the mean square radius and mean square momentum in an exact translationally invariant state versus relation ( a.12 ) obtained for similar quantities within an approximate treatment of the system ( say , with the wavefunction which has a violated symmetry under space translations ) . in comparison with eq.(a.12 ) , the r.h.s . of eq.(a.7b ) is modified by the factor @xmath195 .this modification is closely connected with nonunitarity of the transformation from the usual coordinates and canonically conjugate momenta to their relative counterparts .thus the simultaneous shrinking of the density and momentum distributions , shown in the paper within the hom , is consistent with the model independent uncertainty relations ( a.7 ) .one has to deal with the expectation values of type @xmath196 @xmath197 where , for instance , @xmath198 .one can write @xmath199 we have used the properties @xmath200 and @xmath201=0 $ ] for any vectors @xmath127 and @xmath202 . if @xmath20 is a slater determinant , i.e. , @xmath203 with the antisymmetrization operator @xmath204 which has the property @xmath205 then @xmath206 with @xmath207 furthermore , using the permutation symmetry of the determinants involved , viz ., @xmath208 @xmath209 it is easily seen that @xmath210 @xmath211 @xmath212 now , taking into account eq.@xmath213 and the relation @xmath214=0\ ] ] for any a operators @xmath215 , we find @xmath216 where we have introduced the two sets @xmath217 and @xmath218 of new orbitals @xmath219 and @xmath220 expression ( b.10 ) explicitly reads @xmath221 @xmath222 @xmath223.\ ] ] the latter is equivalent to eq.(18 ) .in order to simplify evaluation of the @xmath224 configuration determinants involved in the r.h.s . of eq.@xmath110 ,let us consider a sparse * nm * @xmath225 * nm * matrix @xmath226 that consists of @xmath227 diagonal * * m**@xmath225*m * block matrices @xmath228 \eqno{(\rm c.2a)}\ ] ] or @xmath229 in the representation ( c.2b ) the diagonal elements @xmath231 ( l=1,2, ,m ) which are , in general , arbitrary and different from one another , are marked by the symbol @xmath232 .one can show that @xmath233 is equal to the product @xmath234 of the * m * determinants of the * * n**@xmath235*n * matrices @xmath236 .each of @xmath237 is composed of the @xmath238 diagonal elements of the matrices @xmath239 ( namely , elements @xmath232 ) , which one encounters when passing in the clockwise ( or counter - clockwise ) direction the corresponding spiral - like contour as displayed in fig . 2 .the relationship ( c.3 ) can be proved using the simple properties of determinants . as an illustration ,let us demonstrate relation ( c.3 ) for @xmath243 and @xmath244 : @xmath245 @xmath246 @xmath247 therefore @xmath248 the aforementioned prescription yields the same result but much quicker. in the case of interest one has to deal with the matrices of the type @xmath249 where each of the sixteen @xmath250 blocks is diagonal , e.g. , @xmath251\ ] ] with the four equivalent dispositions : @xmath252 , \ ] ] @xmath253 , \ ] ] @xmath254 , \ ] ] @xmath255 .\ ] ] taking into account this equivalence we get expression ( 19 ) which is much simpler compared to eq.(18 ) .ernst d.j ., shakin c.m ., thaler r.m .center - of - mass motion in many - particle systems .- phys.rev ., 1973 , * c7 * , no 3 , pp .925 - 930 ; center - of - mass motion in many - particle systems .critique of the gartenhaus - schwartz transformation .- phys.rev ., 1973 , * c7 * , no 4 , pp .1340 - 1343 .dementiji s.v . ,ogurtzov v.i . ,shebeko a.v ., afanasev g.n .effect of center - of - mass motion in quasielastic electron scattering by the @xmath256 nucleus .- yad .fiz . , 1975 ,* 22 * , pp .( english translation in : sov . j. nucl .phys . , 1976 ,* 22 * , pp .

<S> analytical expressions for the elastic and dynamic form factors ( ffs ) are derived in the shell model ( sm ) with a potential well of finite depth . </S><S> the consideration takes into account the motion of the target - nucleus center of mass ( cm ) . </S><S> explanation is suggested for a simultaneous shrinking of the density and momentum distributions of nucleons in nuclei . </S><S> the convenient working formulae are given to handle the expectation values of relevant multiplicative operators in case of the @xmath0 shell nuclei . </S><S> + * 1*. it is known that nuclear sm wave functions ( wfs ) do not possess the property of translational invariance ( ti ) . </S><S> several methods in earlier and recent studies of nuclear structure ( see , e.g. , @xmath1 $ ] ) have been proposed to transform any wf into one which is translationally invariant . of great interest among these methods </S><S> is the projection procedure considered in @xmath2 $ ] . </S><S> along with other attractive features shown in @xmath2 $ ] , this procedure enables a comparatively simple evaluation of the corresponding cm correction to the purely shell quantities ( see ref.@xmath3 $ ] ) . </S><S> the approach developed in @xmath3 $ ] is extended here to calculate the cross sections of the elastic and quasifree electron scattering on atomic nuclei with single - particle ( s.p . ) configurations more complex than the @xmath4 one . </S><S> in particular , we pay special attention to the physical interpretation of a simultaneous shrinking of the density and momentum distributions of nucleons in a nucleus due to the employed separation of its cm motion . </S><S> * 2*. by definition , the elastic ff in question is @xmath5 \mid \phi_{intr } \rangle , \eqno{(1)}\ ] ] while the dynamic ff can be written as in @xmath3 $ ] , @xmath6 @xmath7\mid \phi_{intr}\rangle,\ ] ] @xmath8 where @xmath9 is the energy ( momentum ) transfer , @xmath10 the nucleon mass , @xmath11 the cm coordinate ( total momentum ) operator of the nucleus composed of @xmath12 nucleons so that @xmath13 the coordinate ( momentum ) operator for nucleon number @xmath14 , and @xmath15 the intrinsic wf of the nuclear ground state ( g.s . ) . </S><S> following @xmath2 $ ] , we take as @xmath16 the vector @xmath17^{-1/2 } , \eqno{(3)}\ ] ] for a given trial ( approximate ) wf @xmath18 that may be nontranslationally invariant ( nti ) . here , a round bracket,@xmath19 , is used to represent a vector in the space of the cm coordinate only . in the harmonic oscillator ( ho ) model , where the slater determinant @xmath20 is `` pure '' in the space of the cm coordinate ( the bethe - rose - elliot - skyrme theorem @xmath21 $ ] ) </S><S> , one has @xmath22 with @xmath23 and @xmath24 with @xmath25 here @xmath26 is the oscillator parameter , @xmath27 , @xmath28 . </S><S> result @xmath29 is widely used in applications starting from the work @xmath30 $ ] . </S><S> note also that eqs.@xmath29 and @xmath31 are valid within the ho model , being independent of any specific way to separate the cm motion ( e.g. , the ernst - shakin - thaler prescription ( 3 ) that is equivalent to the so - called `` fixed cm approximation '' in case of finite nuclei(bound systems ) ) . </S><S> now , before finding some analogs of relations @xmath32 with an arbitrary wf @xmath18 ( in particular , the slater determinant constructed of the s.p . </S><S> orbitals in a potential well of finite depth , say , the woods - saxon or hartree - fock ones ) , we would like to trace the cm corrections of the density and momentum distributions @xmath33 and @xmath34 within the ho model . in this connection , </S><S> let us recall the general definitions for these quantities of primary concern : @xmath35 and @xmath36 @xmath37 for the @xmath38-@xmath39 shell nuclei we find in the ho model , @xmath40 @xmath41 substituting these expressions , respectively , into eq.@xmath29 and eq </S><S> . @xmath31 we get with the help of formulae @xmath42 : @xmath43 @xmath44 @xmath45 the intrinsic distributions without any cm correction are @xmath46 @xmath47 by comparing eqs.@xmath48 and @xmath49 one can see that the density and momentum distributions are subject to the equal changes . it does not seem to be surprising if we invoke the well known symmetry between the coordinate and momentum representations of the ho model hamiltonian . </S><S> one should point out that , apart from the `` most symmetrical '' nucleus with @xmath50 , these cm corrections are not reduced to a simple renormalization of the oscillator parameter ( e.g. , the change @xmath51 ) . </S><S> the dependences @xmath33 and @xmath52 calculated by formulae @xmath48 and @xmath49 with @xmath53 are depicted in fig.@xmath54 in this context , we show the ratio @xmath55\exp(-\frac{r^2}{r_0 ^ 2})\ ] ] versus the ratio @xmath56^{5/2}[1+\frac29(a-4)\frac{r^2}{\bar{r}_0 ^ 2 } ] \exp(-\frac{r^2}{\bar{r}_0 ^ 2})\ ] ] we see that an additional correlation between the nucleons , incorporated into ( inherent in ) the intrinsic density @xmath33 , gives rise to their redistribution between the nuclear shells ( from the @xmath38-shell to the @xmath39-one , while the center of the @xmath39-distribution is shifted toward larger @xmath57-values ) . </S><S> this rearrangement of the nuclear interior is accompanied by a decrease of the nuclear density in its peripheral region . </S><S> it implies the corresponding increase of the probability to find the nucleons in the central part of the nucleus . </S><S> remind that both the density distributions ( dds ) @xmath58 and @xmath59 are normalized to unity . </S><S> we interpret these features of the intrinsic dd @xmath33 as its shrinking in comparison with the dd @xmath60 , which embodies the spurious cm motion modes . </S><S> the same interpretation can be applied to the momentum distribution ( md ) @xmath34 vs @xmath61 . </S><S> such a simultaneous shrinking of the dd and md becomes more tractable if one evaluate the respective r.m.s . </S><S> radii and momenta . by definition , one has @xmath62 it is readily seen that @xmath63 @xmath64 where the vector @xmath65 is used to describe the `` zero '' ( lowest - energy ) cm oscillations with respect to the origin of coordinates . </S><S> a complementary smearing of @xmath60 and @xmath61 compared respectively with @xmath33 and @xmath34 is due to the nonphysical motion mode . </S><S> further , let us consider the commutation relations for the relative coordinates @xmath66 and the canonically conjugate momenta @xmath67 @xmath68 : @xmath69=\imath \delta_{kl}(1-a^{-1 } ) ~(k , l=1,2,3 ) \eqno{(11)}\ ] ] along with the original ones : @xmath70=\imath \delta_{kl}~(k , l=1,2,3 ) \eqno{(11')}\ ] ] the corresponding uncertainties meet the equations ( see , for instance , [ 9 ] , p.54 , and also suppl.@xmath12 to this translation ) : @xmath71 @xmath72 for any state @xmath20 normalized to unity , where we have introduced the expectation values @xmath73 and @xmath74 . </S><S> the latter are converted into the values @xmath75 and @xmath76 if @xmath77 . alternatively , </S><S> according eqs.@xmath78 we find in the ho model that @xmath79\left[1-\frac{(000\mid { \hat{\vec{p } } } ^2 \mid 000 ) } { a^2\langle p^2\rangle_0}\right ] \eqno{(13)}\ ] ] then , taking into account that @xmath80 and that in the ho model for a nucleus with fully occupied ( closed ) shells @xmath81 @xmath82 i.e. , @xmath83 where n is the principal quantum number , we get @xmath84 @xmath85 ^ 2\geq \frac94 \left(\frac{a-1}{a}\right)^2 \eqno{(15')}\ ] ] thus we arrive again to relations @xmath86 and @xmath87 . from this derivation </S><S> it follows that eqs . @xmath88 and @xmath89 without the signs of inequality </S><S> are permitted only for the @xmath4 configuration in the ho model . </S><S> such a minimization of the uncertainty relations is retained after making the cm correction . * 3*. further , using the algebraic procedure applied in [ @xmath90 for the calculation of the expectation values @xmath91 and @xmath92 with the slater determinant @xmath93 of the simple ho orbitals , we find @xmath94 @xmath95 @xmath96 with the operators @xmath97 where @xmath98 , @xmath99 is the vector whose components are the creation ( annihilation ) operators for oscillator quanta in the three different space directions . </S><S> analogously , one can show that @xmath100 now , let us assume that the many - body state @xmath20 is a slater determinant of s.p . </S><S> states @xmath101 which describe completely occupied bound states of the nucleons in a spherically symmetric potential well ( e.g. , woods - saxon potential or hartree - fock field ) </S><S> . then ( see suppl.b ) @xmath102 where @xmath103 is the determinant that is deduced from the determinant @xmath104 replacing the vector @xmath105 by @xmath106 in its column with the label @xmath107 . as an illustration , </S><S> let us consider the @xmath108 configuration in the @xmath109-coupling scheme . </S><S> from eq.@xmath110 it follows that ( see suppl.@xmath111 ) @xmath112 where @xmath113 is the @xmath114 determinant , @xmath115 the spatial part of the vector @xmath116 in the shell with the radial quantum number @xmath117 , the orbital angular momentum @xmath118 and its projection @xmath10 , and @xmath119 are deduced from @xmath120 just as @xmath103 from @xmath121 . </S><S> subsequent simplifications can be achieved owing to the transformation properties of the matrix elements @xmath122 with respect to the rotation group . </S><S> in fact , we have @xmath123 @xmath124 with @xmath125 . here </S><S> @xmath126 are the spherical components of the vector @xmath127 . in their turn , the determinants @xmath128 can be expressed in terms of the scalar functions @xmath129 and @xmath130 if one takes into account that the quantities @xmath131 and @xmath132 , each separately , are invariant under rotations , i.e. , they depend on @xmath133 and @xmath134 . </S><S> this property enables us to write down , @xmath135 @xmath136 @xmath137 @xmath138 with @xmath139 , where @xmath140 is the unit vector along the z - axis </S><S> . the scalars @xmath141 and @xmath142 satisfy the relations @xmath143 @xmath144 from which it follows that @xmath145 @xmath146 with @xmath147 where @xmath148 is the radial part of @xmath149 and @xmath150 is the spherical bessel function of @xmath151 . in eqs . </S><S> @xmath152 @xmath153 . </S><S> thus the initial cumbersome task of handling the expectation value @xmath154 is reduced to calculation of the simple overlap integrals . </S><S> all these formulae can be useful when studying cm corrections of the cross sections of the elastic and quasifree electron scattering from nuclei more complicated than the @xmath155 nucleus . </S><S> an attractive feature of similar studies is to proceed with one and the same corrected many - body wf of the nuclear g.s . in evaluating different structure functions like the dd and md . </S><S> of certain interest might be deviations of the @xmath156-dependence of the ratio @xmath157 beyond the ho model from that which is given by the canonical tassie - barker factor @xmath158 . </S>

1602.04293

in the last decades , high - z quasar surveys have discovered supermassive black holes ( smbhs ) with @xmath11 at @xmath12 ( e.g. * ? ? ? * ; * ? ? ? * ) , which pose questions about the earliest coevolution of smbhs and their host galaxies .the observations indicate that at least a minor fraction of high - z bhs experience an extremely efficient mass gain .several scenarios have been proposed ( e.g. , * ? ? ?* ; * ? ? ?* and references therein ) .in order to discriminate them , it is important to clarify observational signatures of rapidly growing high - z bhs in the light of the capability of ongoing and upcoming facilities . here , we focus on high - z smbh formation through collapse of supermassive stars ( smss ) with @xmath13 .smss can be formed from pristine gas in the so called atomic - cooling halo where radiative cooing by atomic hydrogen ( h ) is relevant instead by molecular - hydrogen ( h@xmath14 ) @xcite .formation of h@xmath14 can be suppressed by lyman - werner and h@xmath15 photodissociation photons from nearby star - forming galaxies @xcite , and/or collisional dissociation triggered by galactic shocks @xcite .once @xmath13 of such a gas is assembled , it becomes jeans unstable and collapses to form a supermassive protostar at the center @xcite .after @xmath16 of gas accretion , the sms becomes as massive as @xmath17 @xcite and can directly collapse into a smbh due to the general - relativistic instability @xcite .such direct - collapse black holes ( dcbhs ) are an attractive candidate of seeds of observed high-@xmath18 smbhs . in the very early growth stage of a dcbh ,gas accretion onto the nascent dcbh will result in feedback on the surrounding medium , which might be probed by james webb space telescope ( jwst ; * ? ? ? * ) or future 30 - 40 meter class telescopes , as argued in e.g. , @xcite .however , given the field of view ( @xmath19 for jwst nircam ) , a blind search is not necessarily an efficient way to identify newborn dcbhs and their host halos because dcbh formation may be rare in the early universe .also we note that x - ray emission from steady accretion onto dcbhs may be challenging to detect even with chandra ( e.g. , * ? ? ?an alternative strategy is to first find candidates by some transient signatures with survey facilities e.g. , _ swift _ bat @xcite and then to do followup observations with e.g. , jwst .for example , @xcite argued that ultra - long gamma - ray bursts may accompany dcbh formation . here, we propose that tidal disruption events ( tdes ) of stars formed in the nuclear accretion disk by a newborn dcbhs could occur . as we discuss later, associated x - ray bursts and radio afterglows may be much brighter than tdes in the universe today and could be detected by all - sky instruments such as bat and erosita ( merloni et al .2012 ) and on - going radio facilities ike the expanded very large array ( evla ) even from @xmath10 .this paper is organized as follows . in sec .[ sec : disk ] , we analyze a nuclear accretion disk around a newborn dcbh .first we study disk fragmentation and massive population iii ( pop iii ) star formation in the outer disk , and then discuss properties of the inner slim disk . in sec .[ sec : tde ] , we show that tdes of the massive pop iii stars by the central dcbh are feasible in the early growth stage and consider the observational signatures .we summarize our paper in sec .[ sec : summary ] .we consider a h@xmath14-free gas cloud with @xmath20 collapsing isothermally by h cooling ( ly@xmath21 , two - photon , h@xmath15 free - bound , free - free emission ) in an atomic - cooling halo with typically @xmath22 .first , a proto - sms with @xmath23 and a radius of @xmath24 is formed at the core @xcite .the gas accretion rate onto the core is very high ; @xmath25 where @xmath26 is the sound velocity . as for the mean molecular weight , we set @xmath27 assuming the pristine abundance ( see sec .[ sec : metal ] for the effect of metal and dust ) .note that the above accretion rate is conservative and can be larger than @xmath28 for massive atomic - cooling halos @xcite .the proto - sms grows via accretion through a massive self - gravitating disk with a size of @xmath29 pc ( e.g. , * ? ? ?* ) , to become a sms with @xmath30 ( e.g. , * ? ? ?* ; * ? ? ?* ; * ? ? ?* ; * ? ? ?* ; * ? ? ?* ) , and then , to collapse into a dcbh ( e.g. , * ? ? ?* ; * ? ? ?the total gas mass in the atomic cooling halo is sufficient to supply gas onto the nuclear disk at a rate as high as eq .( [ eq : m_dot ] ) even after the dcbh is formed unless radiative feedback has been turned on ( see sec .[ sec : irr ] ) .the outer part of the nuclear accretion disk will be self - gravitating and the gas angular momentum can be transferred mainly by turbulent viscosity due to gravitational instability .the turbulence also heats the disk and can keep the disk marginally stable , i.e. , the toomre parameter @xmath31 as long as the cooling time of the disk is sufficiently longer than the dynamical time . here , @xmath32 is the orbital frequency of the disk and @xmath33 is the disk surface mass density , which is given by @xmath34 we parameterize the disk viscosity as @xmath35 @xcite , where @xmath36 is the disk scale height .the equation of the heat balance is given by @xmath37 where the left and right hand side represents the viscous heating and radiative cooling , respectively .as for the radiative cooling rate ( in units of @xmath38 ) , we use an approximate formula @xcite .this makes the disk fragmentation radius a factor smaller than that obtained by @xcite with @xmath39 . ] ; @xmath40 where @xmath41 is the disk temperature in unit of kelvin and @xmath42 is the number density of hydrogen nuclei in unit of @xmath43 .( [ eq : lambda ] ) is valid for dense ( @xmath44 ) and warm ( @xmath45 ) pristine gas , which is predominantly neutral and optically thin , and the main cooling process is free - bound emission of h@xmath46 ions ( @xmath47 , * ? ? ?* ) . from eqs .( [ eq : q]-[eq : n ] ) , we can obtain the radial profile of the outer disk , @xmath33 , @xmath48 , @xmath41 and @xmath21 for a fixed ( @xmath49 , @xmath50 ) as a function of @xmath32 or distance from the center , @xmath51 , where @xmath52 is the mass of the dcbh .once the cooling time and dynamical time becomes comparable , the disk will fragment into clumps at around a characteristic radius @xmath53 ( e.g. , * ? ? ?* ; * ? ? ?? * ; * ? ?? * ; * ? ? ?* ; * ? ? ?the cooling time of the disk is estimated as @xmath54 on the other hand , from eqs .( [ eq : q]-[eq : h ] ) , the effective @xmath21 parameter is given by @xmath55 in the outer disk , the gas temperature gradually decreases toward the center , the value of @xmath21 increases inward and finally exceeds a critical value ( i.e. @xmath56 ) , which has been numerically calculated to be @xmath57 ( e.g. , * ? ? ? * ; * ? ? ?* ; * ? ? ?* ; * ? ? ?* ; * ? ? ?for our fiducial parameter set ( @xmath58 , @xmath59 , @xmath60 , @xmath61 ) , the disk fragmentation occurs at @xmath62 where the orbital period is @xmath63 yr .we find that the disk properties around the fragmentation radius are @xmath64 @xmath65 @xmath66@xmath67 typical mass of the clumps can be estimated as @xmath68 where @xmath69 corresponds to the wavelength of the most unstable mode .after fragmentation , the clumps grow in mass and migrate inward by interacting with the disk .the mass accretion rate onto the clump can be estimated as @xmath70 where @xmath71 is the hill radius of the clump and @xmath72 .the initial migration time of the clump , during which the rotation radius typically decreases to @xmath73 , can be approximately given by type i migration time @xcite , @xmath74 where @xmath75 with @xmath76 being the power index of the surface density , @xmath77 , and @xmath78 @xcite . from eqs .( [ eq : dotmc]-[eq : tmig ] ) , the total clump mass could increase to @xmath79 during the migration time ( @xmath80 ) .this total mass is comparable to the isolation mass ( e.g. , * ? ? ?* ) ; @xmath81 and a gap can be formed around the clump .this also means that a reasonable fraction of the gas at the fragmentation radius can accrete onto at most a few clumps .typical clump mass at @xmath73 will range from @xmath82 , which is broadly consistent with numerical simulations of a self - gravitating disk around a sms @xcite .just after the fragmentation , the central core of the clump collapses in a runaway fashion via optically - thin h@xmath46 free - bound emission .the core finally becomes optically thick and a quasi - hydrostatic equilibrium state , i.e. , a protostar is formed , within a free - fall time of @xmath83 . during the migration time , the clumps grow via gas accretion at a rate of @xmath84 ( eq . [ eq : dotmc ] ) , which is higher than a critical value , @xmath85 @xcite . in this case , the protostellar structure consists of two parts ; a bloated convective envelope with a radius of @xmath86 for @xmath87 and a radiation dominated core which dominates the mass @xcite . the core contract within a kelvin - helmholtz ( kh ) time ; @xmath88 and evolves to a pop iii main - sequence star with a mass of @xmath89 and a radius of @xmath90 where @xmath91 is the dimensionless parameter roughly corresponding to the ratio of the gas pressure to the total pressure @xcite .for example , we can estimate @xmath92 yr , @xmath93 , and @xmath94 .after a migration time ( eq . [ eq : tmig ] ) , the clump mass is close to the isolation mass of the disk and a gap will be formed .then , gas accretion onto each clump will significantly decrease , which sets the maximum mass of the star to be a few @xmath95 .we note that , although multiple stars can be formed from a single clump , the mean stellar mass is not much smaller than the total clump mass if the mass function is top - heavy , which is likely in the case of star formation from metal - poor gas ( e.g. * ? ? ?the main - sequence time of stars with @xmath96 is @xmath97 where @xmath98 is the energy conversion rate in nuclear burning and @xmath99 is close to the eddington luminosity @xcite .massive pop iii stars with a sufficient rotation would evolve into red supergiants after the main - sequence phase .this is primarily due to a rotational and convective mixing of cno elements into the outer envelope ( e.g. , * ? ? ?* ; * ? ? ?finally , such massive pop iii stars collapse and supernova explosions occur , leading to metal enrichment in the nuclear region of the host halo , although the outcome will be sensitive to the stellar mass , rotation , and magnetic field ( e.g. , * ? ? ?* ; * ? ? ?* ; * ? ? ?* ) . since we are interested in a few myr after the dcbh formation , which is comparable to the main - sequence lifetime ( eq .23 ) , hereafter we only consider the main - sequence phase. in the inner region , the disk finally becomes optically thick to h@xmath46 free - bound absorption when @xmath100 , which is @xmath101 for our fiducial case .here , @xmath102 is the absorption coefficient and @xmath103 .note that we take @xmath104 instead of @xmath48 as the path length of the photon .the photons can escape mainly in the radial direction since optically thick ionized layers are formed above the thin disk due to the irradiation as we show in the next subsection .once the disk becomes optically thick to h@xmath46 free - bound absorption , both of the temperature and ionization fraction of the disk start to increase due to viscous heating .the accretion rate is super - eddington with respect to thomson scattering in fully ionized gas , @xmath105 .thus , once the disk is significantly ionized , the inner disk becomes the so - called slim disk @xcite .the scale height of the slim disk takes its maximum at a radius where the radiation pressure dominates the gas pressure . in our case, this occurs at @xmath106^{1/3 } \sim 10^{-4 } \\rm pc , \ ] ] with @xmath107 being the effective temperature at the outer edge of the inner disk . in @xmath108 ,the temperature scales as @xmath109 and the aspect ratio of the disk becomes @xmath110 .we note that ionization instability can occur in this transition region between the outer thin and inner slim disk , which may result in an episodic accretion ( e.g. , * ? ? ?given that the accretion rate is well above the eddington rate , a radiation - driven wind will be launched from the inner disk .also , if large - scale magnetic fields exist in the innermost region , a relativistic jet can be launched by the blandford - znajek mechanism @xcite .the total bolometric luminosity can be as large as @xmath111 , or @xmath112 the efficiency @xmath113 is estimated to be @xmath114 for the disk wind ( e.g. , * ? ? ?* ; * ? ? ?* ; * ? ? ?* ) and can be even larger for the jet @xcite. such disk wind and jet can give significant radiative and mechanical feedback on the parent halo , especially in the polar region .although direct emission from the wind and jet are not detectable from @xmath1 , the feedback effects can be indirectly probed by observing reprocessed emission , in particular h@xmath21 and heii @xmath1151640 with jwst ( e.g. , * ? ? ?* ) . here , we consider the effect of irradiation of euv photons from the inner region on the outer neutral disk . inside a critical radius , @xmath116 ,optically - thick ionized layers with a scale height of @xmath117 is formed above the outer disk . here , @xmath118 is the sound speed in the ionized region . outside the critical radius , @xmath119 ,the disk is photo - evaporated by the irradiation .the evaporation rate is estimated as @xmath120 , where @xmath121 is the ionizing photon number , and gas supply from large scales onto the outer disk can be suppressed when @xmath122 @xcite .the critical ionization photon number flux at the disk radius is @xmath123 .before the direct collapse occurs , the main irradiation source is the central sms .the accretion disk merges to the sms at the surface , @xmath124 .the surface temperature of the sms can be as high as @xmath125 just before the direct collapse .the ionization photon flux can be @xmath126 @xcite , which is still below the critical value .once the dcbh is formed , the accretion disk extends inward to the bh horizon scale .high - energy photons emitted from the inner slim disk are shielded by the slim disk itself . instead , the outer disk is mainly irradiated by the outer edge of the slim disk , where @xmath127 and @xmath128 ( see fig .[ mdot ] and [ sec : inner ] ) .the corresponding ionization photon number flux is @xmath129 , which is also below the critical value .finally , massive pop iii stars formed by disk fragmentation can be an important irradiation source .the ionization flux from each star is @xmath130 .about @xmath131myr after the dcbh formation , the total stellar mass becomes @xmath132 , where @xmath133 is the star formation efficiency , and the number of stars can be @xmath134 .then , the total ionization flux can be @xmath135 , which is larger than the critical value .thus , we can concluded that the radiative feedback from the nuclear star cluster can significantly suppress the disk accretion rate after a mass doubling time of the dcbh .we note that massive stars also ionize the gas within the disk thickness , but the size of the ionized regions will be much smaller than the disk radius @xcite .the outer thin disk is still predominantly neutral except near the disk surface even after the star formation .thus our assumption in sec .[ sec : outer ] is justified . in sec .[ sec : outer ] , we assume that the disk consists of pristine gas , and consider only h@xmath46 emission as the radiative cooling process . here , we discuss whether this assumption would be valid for a low - metallicity environment as well .even at @xmath1 , the metallicity in an atomic - cooling halo could be as high as @xmath136 due to past supernova explosions ( e.g. * ? ? ?* ; * ? ? ?* ; * ? ? ?* ; * ? ? ?* ; * ? ? ?* ) , although the uncertainties are fairly large depending on the evolution history of each halo and on stellar progenitors which occur the explosions .we can neglect metal line cooling in the atomic - cooling halo as long as @xmath137 . on the other hand , thermalemission from dust grains could be an important coolant even if the metallicity is as low as @xmath138 @xcite . in a low density region in the atomic cooling halo , the dust and gasare essentially decoupled . due to the efficient emission, the dust temperature is kept much lower than surrounding gases , @xmath139 .when the gas density increases toward the center of the nuclear disk , the dust and gas start to exchange heat by collisions within the dynamical time , which occurs when @xmath140 @xcite , where @xmath141 is the dust depletion factor . from eqs .( [ eq : n ] ) and ( [ eq : t ] ) , the above condition is realized at @xmath142 for @xmath143 .on the other hand , dust grains are sublimated once the dust temperature exceed a critical value of @xmath144 . considering the inner slim disk as the main irradiation source ,the dust temperature is estimated as @xmath145 .therefore , the effect of dust cooling could be neglected within @xmath146 , which is close to the typical star formation radius , @xmath147 . from the above arguments ,we conclude that the dust cooling can be neglected if the gas metallicity in the atomic cooling halo is as low as @xmath148 .otherwise , the dust cooling becomes important at @xmath149 and the nuclear accretion disk is more likely to fragment into clumps at larger radii than we showed in sec .[ sec : outer ] .the initial mass function of the stars formed by disk fragmentation would be less top - heavy compared with the metal - zero case @xcite . also , the size of the star cluster would be larger , resulting in a longer relaxation time .thus , these effects could reduce the total event rate of energetic tdes by dcbhs ( see sec .[ sec : tde ] ) .the orbits of the clumps and stars after the initial migration time is highly nonlinear and stochastic ( e.g. * ? ? ?three - dimension hydrodynamical simulations are required to predict the evolution. if the gas - to - star conversion efficiency is not extremely high , interactions between the stars and residual gas disk will be still effective .the stars may migrate further inward with a timescale of eq .( [ eq : tmig ] ) , or a viscous timescale once a gap is formed around the star ( e.g. , * ? ? ?* ; * ? ? ?* ) ; @xmath150 such migration may proceed down at least to the radius where the disk mass becomes comparable to the clump mass , i.e. , @xmath151 for our fiducial case . before the direct collapse , most of the clumps may merge with the central sms with a stellar radius of @xmath152 , as observed in numerical simulations @xcite . on the other hand , after the direct collapse , a dominant fraction of the stars will not directly migrate to the tidal radius of the dcbh , @xmath153 . instead , interactions between the clumps and stars finally become more relevant , resulting in forming a nuclear star cluster . motivated by the argument in sec .[ sec : star_formation ] , we assume the effective stellar mass @xmath154 and size of the cluster @xmath155 , respectively . initially , most of the stars are alined with the disk .the relaxation time of the eccentricity can be estimated as @xcite @xmath156 here , @xmath157 is the mean eccentricity , @xmath158 is the surface density of the stars , and @xmath159 .the inclination also relaxes with a timescale of @xmath160 . a dozen of disk stars are formed within @xmath161 .hence the disk stars will evolve into a quasi - spherical stellar cluster within @xmath162 .the relaxation time of the spherical cluster can be estimated as @xmath163 as for eq .( [ eq : t_relax_cluster ] ) , we consider non - resonant two - body interaction @xcite , substituting @xmath164 , @xmath165 , and @xmath166 .some stars in the cluster are scattered onto the loss cone orbits within @xmath167 where @xmath168 is the angular size of the loss cone ( e.g. , * ? ? ?* ) . from eqs .( [ eq : t_life ] ) and ( [ eq : t_tde ] ) , tdes can occur within the lifetime of a massive pop iii star .namely , we can expect @xmath169 of tdes by each dcbh .we note that eq .( [ eq : t_relax_cluster ] ) is derived assuming a continuous distribution function of stars . in our case ,the number of stars @xmath170 may not be large enough , resulting in a relatively large scatter in the number of tde per dcbh .pop iii stars disrupted by newborn dcbhs are massive and compact ( see eqs . [eq : mast]-[eq : rast ] ) .consequently , the accretion luminosity of such tdes can be significantly higher than the low - z events , and some associated emissions could be detectable even from high redshifts .the fallback accretion rate is estimated as @xmath171 where @xmath172 is the fallback time , where @xmath173 is a factor of @xmath174 related to structure of disrupted star , @xmath175 is the ratio between the tidal radius and pericenter distance @xcite .the peak accretion luminosity is @xmath176 times larger than the mean accretion rate .if large - scale magnetic fields are amplified during the disruption , a relativistic jet can be launched by the blandford - znajek mechanism .the peak luminosity of the jet is estimated as @xmath177 , or @xmath178 with the super - eddington accretion rate associated with tdes , the innermost disk can be a magnetically arrested disk with an advection - dominated accretion flow @xcite . in this case , the jet efficiency can be described as @xmath179 @xcite .the kerr parameter of dcbh is expected to be as large as @xmath180 @xcite .thus , @xmath181 is a viable assumption .although the prompt emission mechanism of tde jets is still highly uncertain , the peak energy in the engine rest frame will be in hard - x - ray bands @xcite , and observed in soft x - ray bands @xmath182 due to redshift . for an emission efficiency of @xmath183 and a jet beaming factor of @xmath184 ,the peak isotropic luminosity is @xmath185 , which corresponds to an observed flux of @xmath186 from @xmath187 and @xmath188 from @xmath189 .the observed peak duration is @xmath190 for a disrupted star with @xmath191 .such emissions are good targets of soft - x - ray survey telescopes like erosita @xcite and hiz - gundam @xcite .for example , erosita has a limiting flux of @xmath192 , though the value depends on the emission spectrum ( e.g. , * ? ? ? * ) , and can detect the above emission even from @xmath10 .the early tail emission @xmath9 is also detectable and can be used to distinguish this type of tdes from other high - z transients ( e.g. , * ? ? ?* ; * ? ? ?if the intrinsic spectrum is hard enough , the signal can be also detectable up to @xmath10 by _ swift _ bat , but only with an integration time of @xmath193 @xcite. a bat archival data search of dim ( in terms of the observed flux ) and ultra - long transients will be interesting. a promising counterpart may be the afterglow emission ( e.g. , * ? ? ?* ; * ? ? ?the kinetic energy of the jet is as large as @xmath194 .the jet kinetic energy is comparable or larger than the binding energy of the parent halo .thus , tdes in collapsing atomic - hydrogen - cooling halos will also give significant mechanical feedback . the synchrotron emission from the decelerating jet , especially in radio bands , could be detectable even from @xmath10 by e.g. , evla as we show below . in this subsection, we use the notation @xmath195 in cgs unit unless we note .the typical time scale of the afterglow emission corresponds to the deceleration time of the tde jet , @xmath196 , or @xmath197 in the observer frame .here , @xmath198 is the lorentz factor of the jet and @xmath199 , or @xmath200 is the deceleration radius .we assume a wind - like density profile ; @xmath201 , and assume that @xmath202 and @xmath203 as fiducial , which corresponds to that a 10 % of the accreted matter has been ejected from the slim disk with an escape velocity at the inner most region . notethat the first tde typically occurs @xmath204 after the dcbh formation when the wind region extends far beyond the deceleration radius , @xmath205 . at the decelerating shock , magnetic field amplification and electron acceleration can occur .the comoving magnetic field strength at the forward shock is estimated as @xmath206^{1/2}$ ] , or @xmath207 where @xmath208 is the magnetic field amplification efficiency .we note that the energy density of the cosmic microwave background ( cmb ) in the comoving frame is much smaller than that of the amplified magnetic field , thus the effect of inverse compton cooling can be neglected , unlike extended lobes produced by steady jets from high - z smbhs whose radio emission may be muted by the cmb @xcite .the minimum lorentz factor of the non - thermal electrons is given by @xmath209 where @xmath210 is the acceleration efficiency .hereafter we set the power low index of the non - thermal electrons as @xmath211 .the characteristic synchrotron frequency in the observer frame is @xmath212 $ ] , or @xmath213 and the corresponding flux can be estimated as @xmath214 where @xmath215 is the luminosity distance to the source . for @xmath216 , the flux proportional to @xmath217 ( e.g. , * ? ? ?note that in our fiducial case , the shocked plasma is in the slow cooling regime and synchrotron self absorption is not relevant for @xmath218 at @xmath219 . for example , at @xmath220 , the anticipated peak flux @xmath221 after the tde can be @xmath222 mjy even from @xmath10 , which can be detectable by evla and ska .finally , the quasi - thermal disk emission can be also enhanced by the tdes . in our case ,the emission temperature is in euv or soft - x - ray bands and the luminosity may be still close to the eddington value , @xmath223 @xcite . given the intergalactic absorption , such emissions from high - z events are difficult to be detected by on - going and up - coming transient surveys .we briefly estimate the rate of such high - z tdes . the comoving number density of smss or dcbhs in the early growth is estimated to be @xmath224 for @xmath1 @xcite , which corresponds to @xmath225 of such systems in the sky . given @xmath226 tdes by each dcbh and a jet beaming factor of @xmath227 , the total rate at @xmath228is estimated to be a few @xmath229 .note that uncertainties of the dcbh number density are fairly large at this stage and can be significantly larger than the above value ( see e.g. , * ? ? ?* ; * ? ? ?also , if the inward migration of the stars before joining the cluster is more efficient , the size of the star cluster becomes significantly smaller than the disk fragmentation radius , in which case the tde rate is enhanced . on the other hand, relativistic jets may not accompany in some cases e.g. , due to absence of large - scale magnetic field amplification in the tde disk .in fact , off - axis radio afterglow observations of local tdes suggest that only @xmath230 of them are accompanied by powerful relativistic jets @xcite .we analytically calculate properties of a nuclear accretion disk around a direct collapse black hole ( dcbh ) within a few myr after its formation .the outer disk is gravitationally unstable and fragments into clumps at @xmath3 .the clumps evolve into pop iii stars with a typical mass of @xmath231 , which will form a dens star cluster .the relaxation time of the cluster is estimated to be @xmath232 and shorter than the stellar lifetime of a few myr .we can expect that @xmath226 of massive meal - poor stars are tidally disrupted by each dcbh .if a relativistic jet is launched by such a tidal disruption event , bright x - ray transients with a duration of a few months to @xmath131 yr could be produced and detectable by _ swift _ bat and erosita even from @xmath10 . given a formation rate of dcbhs @xmath224, the all sky event rate could be a few times per year , although the uncertainties are fairly large . around the time when the dcbh mass is doubled ,gas accretion will be strongly suppressed by various agn feedback effects ; the radiative and mechanical feedback from the inner slim disk , star cluster , and tde jets , which can be probed by follow - up observations by evla and jwst .we thank zolt@xmath233n haiman , nicholas stone , kunihito ioka , and amy lien for valuable comments .kk is supported by nasa through einstein postdoctoral fellowship grant number pf4 - 150123 awarded by the chandra x - ray center , which is operated by the smithsonian astrophysical observatory for nasa under contract nas8 - 03060 .ki is supported by the simons foundation through the simons society of fellows . ,d. , mihara , t. , sawano , t. , et al .2014 , in society of photo - optical instrumentation engineers ( spie ) conference series , vol . 9144 , society of photo - optical instrumentation engineers ( spie ) conference series , 2

<S> we analyze the early growth stage of direct - collapse black holes ( dcbhs ) with @xmath0 , which are formed by collapse of supermassive stars in atomic - cooling halos at @xmath1 . </S><S> a nuclear accretion disk around a newborn dcbh is gravitationally unstable and fragments into clumps with a few @xmath2 at @xmath3 from the center . </S><S> such clumps evolve into massive population iii stars with a few @xmath4 via successive gas accretion and a nuclear star cluster is formed . </S><S> radiative and mechanical feedback from an inner slim disk and the star cluster will significantly reduce the gas accretion rate onto the dcbh within @xmath5 . </S><S> some of the nuclear stars can be scattered onto the loss cone orbits also within @xmath6 and tidally disrupted by the central dcbh . </S><S> the jet luminosity powered by such tidal disruption events can be @xmath7 . </S><S> the prompt emission will be observed in x - ray bands with a peak duration of @xmath8 followed by a tail @xmath9 , which can be detectable by _ swift _ bat and erosita even from @xmath10 . </S><S> follow - up observations of the radio afterglows with , e.g. , evla and the host halos with jwst could probe the earliest agn feedback from dcbhs . </S>

1602.03524

dwarf galaxies are the lowest - mass inhabitants of the extragalactic universe .their haloes are predominantly made up of dark matter ( dm ) , which makes them the ideal probing ground for dark matter or alternative gravity theories .moreover , their relatively shallow gravitational potential makes their evolution very sensitive to all sorts of internal and external physical processes , ranging from stellar feedback by supernovae to the heating by an external uv - background ( uvb ) . for this reasons ,numerical simulations of the formation and evolution of dwarf galaxies provide a promising way to solve many questions about the existence and character of dark matter @xcite , the strength and timing of the uvb @xcite and the specifics of star formation and stellar feedback @xcite .there are two major classes of numerical simulations of dwarf galaxies .cosmological zoom - simulations start with a large cosmological box filled with dark matter and optionally baryons , and locally increase the resolution by means of refinement levels @xcite .some of these simulations focus on more massive haloes and mainly study dwarf galaxies as satellites or close companions of these haloes .these satellite haloes are influenced by the presence of their host and it remains unclear if this makes them fundamentally different from the isolated field dwarf galaxies @xcite .since the best observational data about dwarf galaxies comes from milky way satellites , this type of constrained simulations is very useful .there are also zoom simulations of individual isolated dwarf galaxies @xcite or groups of dwarf galaxies @xcite . setting up a cosmological zoom simulation however requires a careful selection of an appropriate zoom region , which is an expensive process that requires a suite of preliminary simulations to be run and even then allows for little control over the final properties of the selected halo .simulations of isolated dwarf galaxy haloes either start with an idealized set - up containing a dark matter halo with a theoretically derived density profile and a gas halo embedded therein to study the self - consistent formation of the galaxy @xcite , or start with a fully self - consistent galaxy including a stellar disk and subject this galaxy to external processes @xcite . because these simulations only contain the mass that resides in the dwarf halo , it is possible to obtain much higher resolutions than for zoom - simulations , with less computational cost .this also allows for a large parameter survey where not only model parameters but also structural parameters can be easily changed .however , these simulations do not take into account cosmological effects like the large scale gravitational potential or the effects of mass accretion , apart from the initial density profile , which is derived from cosmological simulations .this drawback can be alleviated by running merger simulations which take into account the merger history of the isolated halo @xcite .the ultimate goal of numerical dwarf galaxy simulations is to produce model data that can be compared to observed dwarf galaxies to constrain theoretical models .we therefore need to find model quantities that can be easily and unambiguously observed .an often quoted relation is the so called @xmath0 relation @xcite , which links haloes from large scale cosmological ( zoom-)simulations to observed stellar mass aggregates by means of an abundance matching technique .this relation is not applicable in the dwarf regime however , since ( a ) halo masses can not easily nor unambiguously be observed , and ( b ) the starting premise of the technique , i.e. the most massive halo corresponds to the most luminous stellar aggregate , breaks down at low masses @xcite .a more accessible tracer of the halo potential is the circular velocity , which can be measured from resolved rotation curves .together with the observed stellar and neutral gas mass , this quantity yields the so - called baryonic tully - fisher relation ( btfr ) @xcite .this purely observational relation is a more promising candidate for model comparisons , since it relates a halo property to baryonic properties .@xcite discuss the position of a set of simulated dsph satellites on the tully - fisher relation , but to our knowledge , no simulated btfr in the dwarf regime has been published before . without an uvb , numerically producing dwarf galaxies that have realistic observational properties poses no real challenge @xcite , especially since there is a degeneracy between the stellar feedback parameters and the position of the resulting dwarf galaxy on a range of observational scaling relations @xcite . however , these systems tend to have unrealistically high stellar masses and bursty sfhs @xcite .when the uvb is taken into account , the evolution of these galaxies looks a lot different : star formation is limited to very early in the simulation , after which the galaxies become gas poor and star formation stops @xcite . or star formation is delayed until very late in the simulation , yielding very exotic dwarf galaxies @xcite .cosmological zoom - simulations predict that only half of the dark matter halos with circular velocities of @xmath1 host galaxies that are able to form stars , let alone keep neutral gas , and this fraction rapidly decreases towards lower circular velocities @xcite .this is in disagreement with observed dwarf irregular galaxies ( dirrs ) , which have continuous star formation histories @xcite , are able to keep neutral gas throughout their evolution , and show no clear imprint of reionization @xcite .the number density of these systems roughly equals that of the dark matter halos from cosmological simulations @xcite , indicating that much more dark matter halos host gas rich , star - forming , low mass dwarf galaxies than predicted .furthermore , ultra - faint dirrs like leo p @xcite , leo t @xcite and pisces a @xcite have a large neutral gas content and circular velocities of @xmath2 .this disagreement has two possible explanations : ( a ) there is something wrong with the simulations and especially the sub - grid model that is employed , or ( b ) there is something wrong with how models and observations are compared , e.g. the halo masses inferred from observations are lower than the corresponding halo masses in simulations . in this work, we will run a large suite of simulations of isolated dwarf galaxy haloes to try to reproduce the observed btfr and address this problem .we will discuss the differences between theoretical quantities and the quantities that are actually observed and try to produce mock observations that resemble the real observations as closely as possible .this should allow us to reliably compare our models with observed field dirrs and allow us to constrain the details of our galaxy evolution model .we will focus on three different aspects of our model : the stellar feedback strength , the strength and timing of the external uvb , and the specifics of a new population iii feedback model .we will explain how these parameters affect the galaxy star formation history ( sfh ) and how this influences the final position of the galaxy on the btfr .the paper is structured as follows : section [ section_model ] gives the details of our simulations and the model parameters .section [ section_analysis ] gives details about our analysis of the results .section [ section_results ] gives an overview of the results for the different parameters , including a discussion of the effect of stochasticity and resolution on the final result .finally , section [ section_conclusion ] will present our conclusions .we ran a large suite of numerical simulations of isolated dwarf galaxy halos using an adapted version of the n - body / sph code gadget2 @xcite .the adaptations consist of a model for star formation using sink particles @xcite , gas cooling and heating using a 5 parameter model that includes metal dependent cooling and heating by an ionizing uvb @xcite , stellar feedback from snii , snia , stellar winds and optionally population iii stars , and an advanced equation of state that takes into account the ionization state of the gas @xcite . in this section ,we describe the initial conditions of our models , and discuss the different subgrid models and parameters that are varied during our study .we conclude with a general overview of our models and simulations .the initial conditions for our isolated halos are generated using a monte carlo sampling technique for both the initial dm - halo and gas halo .the former is set up as a nfw - halo @xcite with a concentration parameter given by @xcite : @xmath3 with @xmath4 the total mass of the halo .the latter is set to be in pseudo thermal equilibrium , with a density profile of the form @xcite : @xmath5 where @xmath6 is set to equal the scale length of the nfw - halo , while the central density @xmath7 is related to the central density of the nfw - halo through @xmath8 with @xmath9 @xcite .initially , the particles that sample the dark matter receive random velocities drawn from the isotropic distribution function corresponding to the nfw density profile . to prevent these velocities from erasing the nfw cusp in the scarcely sampled central part of the distribution ,we construct so - called ` quiet ' initial conditions , in which the velocities are assigned in a symmetric way @xcite .the gas halo is initially in rest , but is optionally set up with a constant solid body rotation , to study the effect of angular momentum on the dwarf galaxy @xcite .the rotation velocity @xmath10 is a model parameter .all models are run with 50,000 dark matter particles and 50,000 smoothed particle hydrodynamics ( sph ) gas particles ( which we will call _ low resolution_ runs ) . to verify that this is indeed enough to obtain a qualitatively correct behaviour of the resulting dwarf galaxy ,we ran some of our models with 200,000 particles of each type ( _ high resolution _ runs ) .note that low and high resolution do not explicitly denote an actual physical resolution difference , but rather a fixed relative resolution difference of a factor of 4 .the actual mass resolution for the particles is set by the desired total mass of the dark matter halo .the smoothing length of the gas particles is dynamically set by requiring a fixed number of @xmath11 neighbours for all particles .the softening length of all components ( dm , gas and stars ) is set to a fixed value , which roughly corresponds to the smoothing length of a gas particle that has a density equal to the density treshold for star formation ( calculated as the radius of a sphere with this density , containing 50 gas particles ) . for clarity , the mass resolution and softening lengths for the different modelshave been specified in table [ table_ics ] .our models start at a redshift of 12 , which corresponds to a lookback - time of 13.37 gyr with the cosmological parameters employed @xcite , and are run until a redshift of 0 , or until the total computation time exceeded the arbitrary limit of 3 months , which happened for some of the high resolution simulations due to physical reasons .all simulations were run on our local computing infrastructure , which consists of 5 computing nodes with intel cpus with between 16 and 64 physical cores .the initial conditions for our simulations contain only 3 free parameters : the total halo mass @xmath12 that is sampled , the initial rotation velocity @xmath13 and the number of particles used to sample the distributions .due to the idealized set up of the isolated initial conditions , we might potentially miss two important effects that might shape the btfr : the accretion of cold gas that will fuel star formation and the angular momentum acquisition of the halo due to mergers . to effectively resolve these effects ,cosmological zoom simulations are needed .it is however very computationally expensive to run a parameter study of the size considered in this work with this type of simulations .to get an idea of the effect of including these effects , we will also consider a merger simulation . instead of simulating a single isolated halo , we set up a system of smaller halos , that are then allowed to merge according to a merger tree that is sampled using the extended press - schechter theory , with a conditional mass function that is fitted to the merger trees from the millenium simulation @xcite .this allows us to incorporate cosmological effects into our simulations without using full blown cosmological zoom simulations .the halos we consider are relatively light , so we expect them to only accrete gas through this type of mergers with other small halos , since they lack the mass to acquire unbound uvb heated gas .the standard uvb we use is the one provided by @xcite , which kicks in at a redshift of 10.5 and reaches its peak strength at a redshift of 2 . since our runs start at a redshift of 12 , the start of the uvb coincides with the onset of star formation in our models .we therefore also experimented with a uvb that only starts at a redshift of 7 ( we will call this _ late uvb _ runs , contrary to the standard _early _ uvb ) . to qualify the effect of the uvb on the star formation , we also ran models with a uvb with a strength that is only 10 per cent of the normal strength ( _ low uvb _ runs ) . the uvb acts as a heating term @xcite and also influences the ionization equilibrium , which introduces a redshift dependence for the cooling curves and the multiphase equation of state ( this is an extension of the model of @xcite ) .the latter takes into account the potential energy reservoir connected to the ionization of ( mainly ) hydrogen , which breaks the linear dependence of the gas temperature on the internal energy of the gas .this takes care of the thermal energy absorbed by the ionization of neutral gas in a subgrid fashion , but does not significantly affect our models. star formation is modelled using the approach described by @xcite .if a gas particle is in a region of converging gas flow , has a temperature below 15,000 k and a density above a density threshold of 100 amu @xmath14 @xcite , it is potentially converted into a star particle , which is dynamically equal to a dm particle , but has some extra properties attached to it ( metallicity , formation time , ... ) .these star formation criteria together with the multiphase equation of state guarantee that only cold , neutral and collapsing gas can form stars .the probability with which this conversion happens is stochastically sampled to reproduce a schmidt law of the form @xmath15 where @xmath16 and @xmath17 are the density of the stars and the gas , and the dynamical time @xmath18 is given by @xmath19 the star formation efficiency is set by a single parameter , @xmath20 .however , this parameter is degenerate due to the self - regulating character of star formation @xcite .a lower star formation efficiency leads to a larger initial star formation peak , since gas is converted into stars more slowly and more gas reaches the density threshold before stellar feedback shuts down further star formation efficiently .this in turn leads to a higher stellar feedback at later times , which suppresses later star formation .the star formation efficiency parameter is therefore also coupled to another parameter in our system : the stellar feedback efficiency , which sets the fraction of the stellar feedback that is effectively absorbed by the interstellar medium .this parameter is closely related to the numerical restrictions of our feedback model : since we do not resolve the hot gas bubbles around supernova explosions , we do not heat the gas particles in our simulation efficiently enough and most of the energy we put in is radiated away by our cooling model @xcite .because of this , it is unclear how much energy ultimately ends up in the interstellar medium .hence the feedback efficiency parameter . to prevent severe over - cooling , we shut off radiative cooling for all gas particles that receive direct stellar feedback from snii and sw , allowing the feedback to spread out adiabatically . for snia feedbackwe do not shut off radiative cooling , since this feedback typically affects gas that is more hot and diffuse and does not suffer from over - cooling .we also present a model in which we do not switch off cooling to illustrate the effect of over - cooling .due to the degeneracy of the star formation efficiency and feedback efficiency parameters , it suffices to only vary one , while keeping the other fixed .we choose to fix @xmath20 at a value of 0.1 @xcite .since this resolution issue affects all types of feedback , we use the same feedback efficiency parameter to scale all energy injections due to stellar feedback , irrespective of the feedback type .different types of feedback then still differ in relative strength and timing , which will cause them to affect the simulations in distinct ways .star particles in our simulations return three distinct types of feedback to the surrounding gas during their lifetime .immediately after the star particle has formed , massive stars with masses in the range @xmath21 will return high energetic radiation to the surrounding ism in the form of stellar winds ( sw ) .following @xcite , we insert @xmath22 energy due to sw in the surrounding ism , weighted with the fraction of the mass of a single stellar population ( ssp ) that is in the range @xmath21 for a chabrier imf ( @xmath23 ) @xcite and multiplied with the feedback parameter .this energy injection is spread out uniformly over a time interval @xmath24 myr , corresponding to the lifetime of these massive stars .when these massive stars reach the end of their lifetime , they explode in snii explosions , which return an energy of @xmath25 to the ism , again weighted with the ssp fraction and the feedback parameter .this happens in a time interval @xmath26 myr and is also done uniform in time .next to energy , snii also return material to the ism .some of it is in the form of h and he , but part of it is also in the form of heavy metals , that enrich the ism . for every ssp , we return a fraction of 0.191 of its mass to the surrounding ism @xcite . for every gas particle, we also keep track of the mass in fe and mg ( the two metallicity tracers used for the cooling and heating and the advanced gas physics ) . for every ssp, we return a fraction @xmath27 of its mass as fe and a fraction of @xmath28 of its mass as mg .all matter is returned uniformly during the same time interval as the snii energy .stars with masses less than @xmath29 do not give rise to snii explosions .however , less massive stars can form white dwarfs and if they are part of a binary system , mass overflow between a red giant and its companion white dwarf can cause the latter to exceed its chandrasekhar limit , giving rise to a snia explosion .if we assume that all stars in the mass range @xmath30 form white dwarfs , we can estimate the number of possible snia explosions from the imf .this then has to be multiplied with an extra factor to take into account that not all white dwarfs will be part of a binary system and not all binary systems will experience the necessary mass overflow .we use a fixed ratio of snia to snii explosions of 0.15 @xcite .these snia explosions also release an energy of @xmath31 to the surrounding ism .due to the large variety of companion masses in binary systems , snia feedback is spread out over a large interval in time .we use the gaussian model of @xcite and return the total energy using a normal distribution centered on a delay time of @xmath32 gyr and with a standard deviation of @xmath33 @xcite . to limit the computational overhead caused by returning very low feedback values in the tails of the normal distribution , we limit feedback to a @xmath34 time interval around the gaussian peak value .snia also return mass and metals to the ism .we return a fraction of @xmath35 of the mass of the ssp in mass , a fraction of @xmath36 of its mass as fe , and a fraction of @xmath37 as mg @xcite .when energy or mass is transferred between a star particle and the surrounding ism , we will always spread it out across 50 neighbouring sph particles of the star particles . to this end, we associate a smoothing length to every star particle and iteratively look for neighbouring sph particles until 50 neighbours are found within this smoothing length. the energy or mass is then spread out according to the same kernel function that is also used for the sph density calculation and using this smoothing length . to ease the iteration ,a small deviation in the number of neighbours is allowed .we set this deviation to 1 for all our runs . in all our runs ,the gas starts of with zero metallicity and hence we can consider star particles formed out of this zero metallicity gas to be very metal poor `` population three '' or pop iii stars .we can also relax this characterization to include stellar particles with very low , non zero metallicities ( we use the arbitrary upper limit @xmath38 < -5 $ ] ) .pop iii stars are still poorly constrained , mainly because there are no observational data and all our knowledge has to come from numerical simulations @xcite .although most models do suggest that they can obtain very high masses , there is no consensus on the form of the pop iii initial mass function ( imf ) , nor on the energy output of pop iii supernova ( sn ) explosions .@xcite composed tables of energy and metal yields for massive stars , including pop iii stars , up to @xmath39 .these energies are very different from the typical energies released by normal `` population ii '' ( pop ii ) and `` population i '' ( pop i ) supernovae , and hence will probably affect the ism in the simulations in a different way .[ [ model-1 ] ] model 1 + + + + + + + as a first method to include pop iii stars , we just assumed pop iii stars to be star particles with zero metallicity and with masses in the range @xmath40 .the lower limit was chosen arbitrarily and is close to the upper limit we use for the masses of pop ii and pop i stars . as a first approximation, we only look at the energy output of the pop iii supernovae and not their metal output . to this end, we scale up the number of supernova explosions per star particle ( which represents a stellar population ) to match the total energy output of a pop iii stellar population with the same total mass , but we keep the energy and metal output per supernova fixed . even in this simple model ,there are already some uncertainties coming from the uncertainty on the pop iii imf .we can assume the same chabrier imf we use for the pop ii and pop i stars for the pop iii stellar population and extrapolate it out to the upper mass limit @xmath39 , but we can also use a flat imf in the mass range @xmath40 , which seems to be in better agreement with numerically found imfs for pop iii stars @xcite . since the number of massive stars in the latter case will be a lot higher , this will have a major impact on the total energy output of the pop iii stellar population .we will call this model `` pop iii model 1a '' , and parametrize it by the fraction of the total mass of the star particle that consists of high mass stars that will explode as supernovae and each return @xmath41 of energy to the ism . the total energy a star particle pumps into the ism ( over the whole feedback time interval )is then given by multiplying this parameter with the mass of the stellar particle .we will consider a model with fraction @xmath42 ( low feedback ) , @xmath43 ( high feedback ) , and @xmath44 ( middle feedback ) .the high feedback value illustrates the fact that for a flat imf , the total energy emitted by a pop iii sn is a lot higher than for a snii .the time interval for this feedback corresponds to the expected lifetime of the lower and upper mass limit stars : 0.006 to 0.36 myr @xcite .we also investigate the effect of the lower mass limit for pop iii stars on the pop iii energy feedback , by considering a model with lower limit @xmath45 instead of @xmath46 . in this case, all pop iii stars will be extremely short - lived , and will deposit a large amount of energy into the ism over a very short time interval ( 0.006 to 0.043 myr ) .we will call this model `` pop iii model 1b '' and consider two feedback parameters : @xmath47 ( low feedback ) , and @xmath48 ( high feedback ) .[ [ model-2 ] ] model 2 + + + + + + + a second , more advanced pop iii model also takes into account the energy radiated away by the pop iii stars during their stellar lifetime , which is similar to the energy output provided by sw for pop ii and pop i stars , but is potentially a lot higher @xcite .this energy output bridges the very short gap between the birth of the star particle and the onset of pop iii sn feedback and provides a further early energy input into the ism .we consider two versions of this model : a model with high sw energy ( @xmath49 ) and low pop iii sn feedback ( @xmath50 ) , and a model with low sw energy ( @xmath41 ) and high pop iii sn feedback ( @xmath51 ) .[ [ model-3 ] ] model 3 + + + + + + + a third and most advanced pop iii model also takes into account the metal yields provided by @xcite to provide a more realistic metal enrichment of the ism by pop iii supernovae .this model combines the knowledge obtained from the previous two models with a fully consistent treatment of pop iii stars , and is used by @xcite .we will verify this model as part of our parameter study .apart from returning the low pop iii sw feedback and high pop iii sn feedback to the ism , a pop iii star particle returns a fraction of 0.45 of its total mass to the surrounding ism .most of this mass is in the form of h and he , but a fraction of 0.026 is in the form of metals , with a total fe mass fraction of @xmath52 and a mg mass fraction of @xmath53 .these last values are simply 10 per cent of the corresponding value for a normal snii , and were based on the metals yields of @xcite . in total , 263 simulations were run , with a total of 15 different combinations of code versions and parameter values as discussed above .apart from this , we considered initial conditions with 5 different halo masses ( @xmath54 , @xmath55 , @xmath56 , @xmath57 and @xmath58 ) , 3 halo rotation velocities ( no rotation , @xmath59 and @xmath60 ) and 2 resolutions ( @xmath61 and @xmath62 particles ) .we adopt a simple naming convention for our simulations , based on ( a ) the code and parameters with which the simulations was performed ( table [ table_codes ] ) , and ( b ) the parameters of the initial conditions ( table [ table_ics ] ) . the name for a simulation of a @xmath54 model with a low resolution and no rotation with a code with no pop iii stars and a late and low uvb , and with a low stellar feedback parameter e.g. will be c1p1m1r00l .we only have 30 different initial conditions ( ics ) , of which 15 low resolution ics that were used for almost all models . to check the effect of stochastic changes to the simulations due to poisson noise in these ics , we reran 1 model with ics that were generated using a different random seed. this model is called c1p1bis and has code and parameter values equal to those of c1p1 . for the models c1p1 and c3p2 , we ran all simulations using both the low resolution and the high resolution initial conditions , to check the convergence of our simulations . for the other models , we only ran one high resolution simulation , m9r10h , and use this as a general check on the convergence of that specific model .apart from the initial conditions described above , we also discuss one merger simulation , to assess the influence of cosmological effects on our models .this merger simulation is part of a suite of merger simulations that was run as a result of the parameter study performed in this work @xcite . for clarity, we will keep its name from that work : dg10e9-np3 .this model has the same properties as model c3p1 : a full and early uvb , feedback efficiency 0.7 and no pop iii feedback. we will denote it by a @xmath63 symbol . in the end ,7 out of 263 simulations exceeded the 3 month time limit imposed : models c1p1m7r00h , c1p1m9r00h and c1p1m9r05h , model c3p2m9r00h , and models cep1m5r00l , cep1m7r00l and cep1m9r10h .all these models form an excessive number of stars from the beginning of the simulations , leading to a very computationally expensive stellar feedback contribution , which explains the long run time . however , this excessive star formation is in line with other results for the same code and parameter values , so we learn nothing new from these simulations .we will discard them anyway .3 simulations crashed : cep1m1r00l , cep1m3r00l and cep1m9r00l .these simulations did not apply an adiabatic cooling periodic to gas that received stellar feedback and as a result formed so many stars that the program did no longer find any gas to give feedback to and crashed .similar behaviour was found for the other simulations of this model that did successfully run , so that we can discard these simulations .model c1p1m9r00h shows very strange behaviour , in the sense that while for all other models the circular velocity tends to decrease over time , it increases significantly for this model , ending up with an unrealistically high circular velocity of more than 150@xmath64 .the low resolution equivalent c1p1m9r00l has a more realistic circular velocity but ends up with almost no neutral gas , contrary to similar models with the same mass and different rotation parameters or the same rotation parameter and lower masses , which still have neutral gas at the end .simulation c1p1bism9r00l , which has the same parameters but a differently sampled ic shows the same behaviour .we conclude that something is fundamentally wrong with model c1p1m9r00 and will therefore discard these 3 simulations . for very similar reasons , we will also discard simulation c3p2m9r00l .other models with the m9r00l ic will be taken into account . from the 263 simulations , 250 will hence be discussed in the remainder of this paper . [ cols="<,^,<,^,<,^",options="header " , ]since we are interested in the location of our simulations on the btfr , we need to determine ( a ) baryonic masses for all simulations , which are the sum of the stellar masses and neutral gas masses , and ( b ) circular velocities for all simulations . since the btfr is an observational relation , we will try to determine these quantities in the same way observers would do .we of course also have access to precise values for most of these quantities ( some halo properties are shown in fig .[ fig_m_theory ] ) , which will allow us to assess the validity of some observational proxies .apart from the btfr , we will also determine some other global galaxy properties and compare them to observed dwarf galaxies , as a further validation of our models .all analysis was performed using our open source analysis tool hyplot . since showing the results for all 250 simulations in the same plotmakes the plots very hard to read , we also provide online interactive versions of some of our plots . numerically , we can determine the stellar mass by simply adding together the masses of the star particles in the simulation .however , the value obtained in this way will likely be larger than what will be observationally observed .not all star particles reside within the central regions of the galaxy and will hence be observed . and even the star particles that are in the central regions might be too old and weak to be observationally detectable . to obtain mock observational values for our simulations, we fit a srsic profile to the surface brightness profile of our galaxy , cut off at a surface brightness of 30@xmath65 , and use this to estimate a half light radius ( @xmath66 ) and an absolute magnitude in the v- and i - band .surface brightnesses are estimated from the age and metallicity of the star particles using the tables of @xcite .we then use the i - band luminosity and v - i colour to estimate the total stellar mass @xcite .not all our simulations contain enough stars to fit a general srsic profile .we therefore use a simple exponential curve if the srsic profile is visibly a bad fit .all simulations of the c7p6 model have very little stars , which are spread out over huge volumes , so that even an exponential fit is impossible . for these models, we will use the sum of the star particle masses as stellar mass , keeping in mind that these systems in no way resemble real dwarf galaxies .some of the more massive models ( mainly with the m9r00l and m9r00h initial conditions ) have very steep srsic profiles with indices larger than 1.5 .the total magnitudes obtained from these fits are generally much smaller than the total magnitudes obtained by summing the luminosities of the star particles , so that the estimated mass for these systems tends to be wrong by a large factor .we will therefore also fit an exponential curve to the central brightness profile of these models , so that the estimated total magnitude better resembles the summed luminosities . in total , 216 of the 250 simulationswere fitted with a general srsic profile and 18 were fitted using an exponential profile .all 16 simulations of the c7p6 models were discarded . .the dashed line represents the mean ratio of observed and simulated masses and the dotted lines are a 1@xmath67 interval around these values .the full line is a 1:1 relation .model c7p6 ( gray left pointing triangles ) lies on the full line since we set its mock observational mass to be equal to the actual value.[fig_mstar],scaledwidth=50.0% ] in fig .[ fig_mstar ] , we show the actual stellar mass as a function of the mock observational value .we find quite a good correlation between both values , with a mean observed to simulated stellar mass ratio of @xmath68 ( discarding the c7p6 and cep1 models and models with a srsic index larger than 1.5 ) , which means this simple technique indeed works .the fact that we systematically underestimate the stellar mass is due to the differences in assumed imf between the models used to derive luminosities for our star particles @xcite and the mass to light ratio models of @xcite . as described above, we need a detailed model of the ionization state of the gas to correctly describe the hydrodynamics of the gas in our simulation . to this end , we keep track of two metallicity tracers and the temperature , density and redshift of the sph particles and determine the ionization equilibrium by a 5d interpolation on pre - calculated tables @xcite .we can use the same tables used in the simulation to calculate the neutral fraction of the gas for the snapshots of the simulation in post - processing . by multiplying these neutral fractions with the masses of the gas particles ,we can hence very easily determine neutral gas masses .as for the stellar mass , these numerically determined masses might overestimate the observational values , since not all gas will be confined to the central regions of the galaxy. however , observational hi clouds can be significantly larger than the physical size of the galaxy as estimated from @xmath66 .since we consider galaxies in isolation and in the presence of an ionizing uvb , we do not expect neutral gas at large distances from the galactic halo .so rather than introducing an arbitrary cutoff on the gas taken into account for the neutral gas mass , we will always consider all gas . the circular velocity of a halo is a measure for the strength of its gravitational potential .theoretically , one can construct a circular velocity profile of a galaxy based on the gravitational potential inferred from its total mass profile . to characterize this profile ,one then has to either choose a characteristic radius at which to evaluate this profile or take e.g. the maximal value as `` the '' circular velocity .the latter makes sense for halos with a strictly rising circular velocity profile that tends to become flat at larger radii , and for which both methods should obtain the same result if the chosen radius is large enough .observationally , the total mass profile is not accessible and one has to derive the circular velocity using a tracer , e.g. the neutral gas halo or the stellar body , see fig .[ fig_vprof ] .the btfr of @xcite uses the maximal circular velocity as obtained from resolved hi rotation profiles and we will try to use the same definition for the circular velocity of our galaxies whenever possible .alternatively , @xcite also uses circular velocities derived from line widths if a full rotation curve is not available. we will use the same technique and compare it with the circular velocities obtained from rotation curves ., the dashed line represents a linear least - squares fit to the points , and the dotted lines represent a 1@xmath67 interval around this fit.[fig_vstar],scaledwidth=50.0% ] unlike @xcite however , we can not use hi for all our simulations , since not all our simulations contain neutral gas at the end .we therefore will introduce an alternative method to estimate the circular velocity of our halos based on the velocity profile of the stars .to this end , we consider an observationally accessible property of the stellar body , namely the velocity dispersion along a line of sight .for an isothermal sphere , the stellar velocity dispersion can be shown to correlate linearly with the circular velocity @xcite . in fig .[ fig_vstar ] , we show the velocity dispersions in the @xmath69 , @xmath70 and @xmath71 direction of the stars within a sphere with radius @xmath72 for all our simulations as a function of the actual theoretically determined maximal circular velocity . to correct for the fact that not all stars are equally bright and contribute equally to the spectra that can be used to observationally measure velocity dispersions , we have weighted the contributions of the different star particles with the number of red giant branch ( rgb ) stars per particle . apart from a few outliers which do not contain enough stars to fit a srsic profile and for which the whole procedure is in fact meaningless , the values clearly trace out a linear relation .a least - squares fit to the simulation data yields the following relation between the `` observed '' stellar velocity dispersion and the circular velocity of the halo : @xmath73 if no neutral gas is available , we will derive the circular velocity using this relation .we could of course also use the theoretical value , but this is in no way observationally accessible , while the proposed relation in principle is . , the full line represents a 1:1 relation , the dashed line is a linear least - squares fit , while the dotted lines correspond to a 1@xmath74 interval around this fit .only the simulations for which a workable rotation profile was found are shown.[fig_vhi],scaledwidth=50.0% ] if neutral gas is available , we can of course use the rotation of the neutral gas as a tracer for the circular velocity . to this end, we produce a mock rotation curve of the neutral gas and determine its maximum value .we visually checked all rotation curves and only kept the values that were obtained from curves with a clear rotation and enough neutral gas . in fig .[ fig_vhi ] we show the circular velocity derived from the neutral gas as a function of the theoretical value .the agreement is reasonable on the massive end , but gets worse for halos with a small theoretical circular velocity .overall , we find that hi rotation curves underestimate the true circular velocity ., the full line represents a 1:1 relation , the dashed line is a linear least - squares fit and the dotted lines a 1@xmath74 interval around this fit . only the simulations for which a workable mock spectrum could be constructed are shown.[fig_vfwhm],scaledwidth=50.0% ] , the full line represents a 1:1 relation , the dashed line is a linear least - squares fit and the dotted lines a 1@xmath74 interval around this fit . only simulations for which both a workable rotation curve and a workable mock spectrum were constructed are shown.[fig_vfwhm_vs_vhi],scaledwidth=50.0% ] fig .[ fig_vfwhm ] shows the circular velocities obtained by using line widths for the neutral gas , following @xcite , as a function of the theoretical value . to produce these values, we fitted normal distributions to mock spectral lines .the circular velocity is then half @xmath75 , the width of the gaussian bell curve at 20 per cent of its maximum value .we average this value over 8 distinct edge - on lines of sight and again only kept the values for which we could visually confirm a decent fit .we find a slightly better agreement than for the velocities derived from rotation curves , but the theoretical value is still underestimated for most of the simulations . fig .[ fig_vfwhm_vs_vhi ] shows the same velocities , but as a function of the values obtained using the rotation curves .we notice that the values derived from the line widths are systematically higher than the values obtained from the rotation curves .not all simulations have circular velocities determined using both methods ( because of a bad quality rotation curve or a bad gaussian fit to the line widths ) , in which case we set the corresponding velocity to zero . in total ,113 simulations have enough neutral gas to estimate a circular velocity from the gas . for 90 ,we were able to determine the circular velocity using both methods .11 more have circular velocities derived from the rotation curve of the gas , so that in total we will use rotation curve based circular velocities for 101 simulations . for the remaining 12 simulations with enough neutral gas , but for which the rotation curves could not be used , we will use the value derived from line widths instead .it should be noted that observed circular velocities sometimes correct for turbulence by adding correction factors to the quadratic circular velocities , both for those estimated from rotation profiles as from line widths . for an exponential disk ,it is straightforward to determine an asymmetric drift correction .however , our galaxies do not show exponential disks , which makes it much harder to estimate correction factors .moreover , as noted by @xcite , these corrections are typically small for low mass halos .we will therefore quote the circular velocity derived from the neutral gas as a directly observable quantity , without applying any corrections to it . , the stars are the observational values from @xcite and @xcite .the full line is the fit of @xcite , while the dashed lines are least - squares fits to the simulation data.[fig_btfr_all],scaledwidth=50.0% ] fig .[ fig_btfr_all ] shows the btfr for all our simulations .the top panel shows the btfr obtained using circular velocities derived from the gas , while the bottom panel shows those derived from the stellar velocity dispersion .we have also indicated the power law fit of @xcite in both panels .it is clear that ( a ) the btfr with velocities from the gas lies above the observational relation for the bulk of our simulations , and ( b ) there is a clear difference between the relation obtained using gas circular velocities and stellar circular velocities .since we showed above that stellar circular velocities are more closely related to the real theoretical circular velocities , this means that circular velocities derived from neutral gas observations systematically underestimate the circular velocity .notice also that there is a bend in the bottom btfr at stellar circular velocities of @xmath76 , which indicates the transition between a mass regime where all halos keep gas and form stars for the entire lifetime of the simulation , and a mass regime where the halos loose their gas and contain only old stars .this bend is absent in the top panel , since we can not derive circular velocities from the neutral gas for the halos which loose their gas .the observational btfr of @xcite does not show this bend , but has considerably more scatter in this regime than the observed btfr of @xcite . keeping this in mind, we will always show two btfrs when discussing our simulations : the one derived from the gas , which we can directly compare to the observations of @xcite , but which can only be calculated if there is neutral gas , and the one derived from the stellar velocity dispersions , which we can calculate for all our simulations , and which we will compare to the observed btfr of @xcite . to guide the eye, we will compare the latter to a least - squares fit to the btfr in the bottom panel of fig .[ fig_btfr_all ] , which will tell us how the particular model compares to the other models in the set . the final stellar mass and neutral gas mass of a simulated galaxyare completely determined by its star formation history ( sfh ) .differences in these masses between different models will therefore always trace back to differences in the sfh for these models and it is important to have a good way of investigating the sfh .observationally , the sfh is inferred from the colour magnitude diagram ( cmd ) , using a complex fit of stellar population models with different ages and metallicities . since there is a degeneracy in colour between old metal rich and young metal poor stars ,there is a large uncertainty on these observed sfhs , which makes it difficult to compare them with our simulations .moreover , these observations suffer from the same luminosity constraints discussed above and only trace the central galaxy . to explain the differences between our models , we need a more detailed sfh than what is observationally accessible and we need to take into account all stars , not only those that happen to be in the central galaxy at the end of the simulation .we will therefore use a theoretical sfh in our analysis , shown as the effective star formation rate ( sfr ) as a function of time .it is determined by counting the stellar particles that are born in a specific time interval during the simulation , irrespective of their position or intrinsic properties .we note that these sfhs can not be compared to the observed sfhs in e.g. @xcite , which are based on a fit to the observed cmd .determining mock observational sfhs based on a fit to mock cmds falls outside the scope of this paper and will be subject of future work . to check whether our simulated galaxies look like observed galaxies , we look at the so - called _ scaling relations_. these relations describe the correlations of various global observational quantities that were found for observed dwarf galaxies and they loosely define the typical size and luminosity for a dwarf galaxy .@xcite showed that there is a large range of stellar feedback parameters which gives rise to dwarf galaxies well within the range of these scaling relations , so we can not in general use them to constrain our models .however , they do provide a first check on the result , as models that do not lie on the scaling relations can not be considered to represent real dwarf galaxies .the scaling relations are shown in fig .[ scaling_relations ] . the observational data with which we compare consists of early and late type galaxies within the local volume @xcite , including recent additions like leo p @xcite , and pisces a and b @xcite , galaxies within the coma @xcite , virgo @xcite and m81 cluster @xcite , and isolated dwarf galaxies @xcite .model c7p6 is not shown , since we were unable to fit a srsic profile to these simulations ( see section [ subsection_stellar_mass ] ) .simulation capam1r00l is also not shown , since for that simulation the b - band srsic fit was very bad ( leading to an unrealistic @xmath77 colour of -4.96 ) .the large scatter at the high surface brightness end of the @xmath78 colour is similarly caused by bad quality srsic fits to the low stellar mass simulations at this end , but these simulations are shown .all other simulations fall well within the observational scatter .we do notice two important effects : all simulations have @xmath77 colours at the high end of the observational distribution , and all simulations have large half light radii .both effects can be linked to a large initial peak in the galactic star formation , as this will ( a ) produce a predominantly old stellar population , and ( b ) lead to a high stellar feedback at the start of the simulation that will disperse the interstellar gas and lead to an overall larger and shallower halo . to determine stellar metallicities , we follow @xcite , who observed the stellar metallicities for 7 local group dirrs from the optical colours of its red giant branch ( rgb ) . to this end, we use the stellar evolution tracks of @xcite for pop i and pop ii stars and those of @xcite for pop iii stars to estimate the fraction of the ssp represented by a star particle that will reside on the rgb at any given time .we then weigh the metallicities of the star particles within the galaxy with this fraction to calculate the average [ fe / h]-metallicity . we note that this method will yield lower metallicities than for example a luminosity weighted average metallicity as can be obtained from stellar spectroscopy , since it is biased towards the older star particles , that will have significantly higher rgb fractions .this effect is visible for the high metallicity end of fig .[ fig_zlum_vs_zrgb ] .the low metallicity end corresponds to galaxies with a predominantly old stellar population , where both methods yield the same metallicities .we also compare the average @xmath79[fe / h]@xmath80 metallicities with the metallicities of the averaged metal content of the galaxies , [ @xmath79fe / h@xmath80 ] .the former is the quantity discussed in @xcite , while the latter is a more meaningful measure of the total metallicity of the galaxy .it is clear from fig .[ fig_zlum_vs_zrgb ] that @xmath79[fe / h]@xmath80 underestimates the metallicity of the galaxy by giving too much weight to low metallicity rgb stars ., the dashed lines represent a 1:1 relation .simulations which only formed a single generation of pop iii stars ( model c7p6 ) and hence have no metallicity ( [ fe / h ] = -99.0 ) are not shown.[fig_zlum_vs_zrgb],scaledwidth=50.0% ] -band magnitude .the symbols represent the different code and parameter values following table [ table_codes ] , the stars are the observational values from @xcite.[fig_metallicities],scaledwidth=50.0% ] fig .[ fig_metallicities ] shows the average @xmath79[fe / h]@xmath80 metallicities as a function of the @xmath82-band magnitudes for all simulations , compared to the observational data of @xcite .it is clear that all our models lie below the observations and our too metal poor .the reason for this is that most of our simulations form most of their stars early on and then stop forming stars or form very little stars for the rest of the simulation .most stars are hence born from gas that is metal poor . in section [ section_results ]we will discuss ways to reduce the relative strength of this initial star formation peak .it is already worth noting that the models that somewhat succeed in this are closer to the observed metallicities .before we can start comparing different models , we first have to assess the effect of small random changes in the system on the outcome of the simulations .this will help us distinguish between differences caused by stochastic effects and fundamental differences caused by physical ingredients .when discussing differences , we will focus on the btfr and related quantities : stellar and neutral gas masses , circular velocities and star formation histories . .for clarity , simulations that represent the same initial condition have been joined by a full line.[fig_stochasticity_btfr],scaledwidth=50.0% ] fig .[ fig_stochasticity_btfr ] shows the btfr for the same models , but with ics generated with different random seeds ( the c1p1bis model ) .both btfrs are clearly very similar , although there are small differences in final masses between the models with the same initial mass and rotation parameters .however , most models have almost no neutral gas at the end of the simulation , so the agreement stems mostly from very similar final stellar masses . , scaledwidth=50.0% ] when we look at the final neutral gas masses of the different models ( fig .[ fig_stochasticity_gas ] ) , we see the agreement between the models with different ics is a lot worse , especially in the cases with intermediate neutral gas masses , where the relative difference can be up to a factor 2 .this makes sense , since in these cases the galaxy has lost almost all of its gas by the end of the simulation , so the fraction that is left will be largely determined by small stochastic differences between the simulations ., scaledwidth=50.0% ] fig .[ fig_stochasticity_sfr ] shows the sfh for the same model , with two different ics .we see a good qualitative agreement over the entire run time of the simulation .we conclude that stochastic effects do not significantly change the position of our models on the btfr , nor their final stellar masses and sfhs . when discussing the neutral gas masses of different models however , we have to keep in mind the large sensitivity of these masses to stochastic effects , especially when there is neutral gas left but its final mass is low .this poses no real threat to our work , since the ultimate goal is to produce dwarf galaxies with final neutral gas masses that are of the same order of magnitude as the final stellar masses , in which case the outcome is less determined by stochastic effects . andfor other models , our simulations are capable of determining whether or not a galaxy will be able to keep a significant fraction of its neutral gas , which is all that matters for our purposes .since we want to explore a large parameter range , we want to keep the computational cost of the simulations to a minimum .it is therefore important to carefully select the numerical resolution of the simulations .a high resolution will improve the accuracy of the results , but will at the same time increase the run time of the simulation .meanwhile , our sub - grid model is based on the assumption that we can treat individual star particles in our simulations as stellar populations with statistically averaged properties , which limits our highest resolution to a star particle mass of @xmath83 .we therefore opt to use lower resolution simulations with shorter run time for the bulk of our simulations , as long as this resolution is sufficient to resolve the properties of interest . the simulations in e.g. @xcite , @xcite , and @xcite use 200,000 sph particles , which for the lowest mass simulations in our mass range of interest roughly corresponds to five times the minimal star particle mass .the mean run time of these simulations is however too long for a large parameter study .simulations with four times less particles are computationally a lot cheaper . in this subsection, we will compare these low resolution simulations to the high resolution simulations for two distinct uvb models and discuss the convergence of the results .note that there is almost a factor of 10 difference between the masses of the lowest mass and the highest mass ics we employ .this means that there will also be a factor of 10 difference between the effective mass resolution of these simulations .likewise there is a difference in the mass resolutions used for baryons and dark matter , since we use the same number of particles for both components . since the run time of the simulationis set by the actual number of particles and not by their masses , we will use the same number of particles for ics with different masses .we expect the resolution to be worst for the galaxies with the highest total mass .we are interested in global properties of the final dwarf galaxy : total stellar mass , neutral gas mass and the location of the galaxy on the btfr .these properties are mainly set by two mechanisms .on the one hand , the star formation itself will govern the amount of neutral gas converted into stars , which affects both the stellar mass and the neutral gas mass . on the other hand ,feedback from already formed stars will affect the neutral gas mass and through the gravitational interaction between gas and dark matter will affect the galaxy potential and hence the circular velocity of the halo . to obtain a converged result ,we hence need sufficient resolution to ( a ) get the global sfh right , and ( b ) correctly resolve the effect of stellar feedback on the gas surrounding young star particles . in section [ subsection_stochasticity ], we concluded that , for the low resolution models , stochastic effects do not cause significant differences in final stellar mass , circular velocity or being able to keep a significant amount of neutral gas .they do however affect the final neutral gas mass if it is low compared to the stellar mass . since ics with a different number of particles inevitably will suffer from the same stochastic effects , we should keep this in mind when comparing the results for different resolutions ., scaledwidth=50.0% ] fig .[ fig_convergence_sfr ] shows the typical sfh for one of our models and for both the low resolution and the high resolution version of the simulation .qualitatively , the sfh is very similar : there is a large peak in star formation at the start of the simulation , after which the feedback of these initially formed stars strongly suppresses further star formation by dispersing the neutral gas .heating by the uvb prevents this dispersed gas from falling in again , resulting in a very low ongoing star formation for the remainder of the simulation . for models with a halo mass below @xmath84 ,all neutral gas is removed after the initial star formation peak , so that star formation shuts down entirely .this behaviour is reproduced by both the low and high resolution runs and also depends upon the rotation parameter of the halo ( see section [ subsection_uvb ] ) .it is also in agreement with the results of @xcite . and the three different rotation parameters : no rotation ( full lines ) , @xmath59 ( dashed lines ) , @xmath60 ( dotted lines ) .the black lines represent the low resolution version , the grey lines the high resolution version.[fig_convergence_cumsfh],scaledwidth=50.0% ] qualitatively , there is hence no difference between the high and the low resolution runs .both form the bulk of their stellar mass during the first gigayears of the simulations , as can be seen from fig .[ fig_convergence_cumsfh ] .the simulations with a higher rotation parameter build up their stellar mass somewhat slower and this behaviour is reproduced for both the low and the high resolution versions ., scaledwidth=50.0% ] .[fig_convergence_all],scaledwidth=50.0% ] fig .[ fig_convergence_gtfr ] shows the relative difference in final neutral gas mass for the two models between both resolutions .we notice the same pattern as in the case of stochastic differences between the initial conditions : the relative difference is small in the low and high mass regimes , but is significant in the intermediate mass regime . again, this is due to the stochastic nature of feedback in the simulations .if the overall neutral gas mass is low , not much feedback is needed to disperse this gas , so small stochastic differences can lead to significantly different neutral gas masses . as a result , models with relatively more feedback and less neutral gas will have larger differences between the low resolution and high resolution version , as illustrated in fig .[ fig_convergence_all ] .model cep1 is excluded from this figure , as simulation cep1m9r10h exceeded the 3 month time limit .however , both cep1m9r10l and cep1m9r10h have no neutral gas at the latest time a snapshot was written for cep1m9r10h . .for clarity , simulations representing the same initial condition have been joined by a line.[fig_convergence_btfr_c1p1],scaledwidth=50.0% ] . for clarity ,simulations representing the same initial condition have been joined by a line.[fig_convergence_btfr_c3p2],scaledwidth=50.0% ] the resolution has only a minor effect on the location of the simulations on the theoretical btfr for these models , as can be seen from the bottom panels of fig .[ fig_convergence_btfr_c1p1 ] and fig .[ fig_convergence_btfr_c3p2 ] . for the mock observational btfr ,the resemblance is worse , but this is mainly caused by the overall low neutral gas masses , which make it very difficult to obtain reliable circular velocity estimates . for model c1p1 , the mock observational btfr looks better and in this case high and low resolution simulations trace out the same btfr , which is higher than the observed one .we conclude that our low resolution simulations are well converged in terms of stellar mass and theoretical circular velocity .if enough neutral gas is present , the neutral gas mass is also converged and circular velocities derived from the neutral gas are reliable . if the neutral gas mass is low , convergence is less clear , but this is in line with the effect of stochasticity .the low resolution runs are hence sufficient to distinguish between galaxies that are unable to keep any neutral gas and those that keep a significant amount of neutral gas .we can hence study the effect of different parameters on the btfr using the computationally cheaper low resolution simulations and expect our results to be qualitatively correct . for more detailed studies of other properties of the dwarf galaxies , especially related to the small scale structure of the gas halo , we should use the more expensive high resolution runs , but this falls outside the scope of this work . .for clarity , simulations representing the same initial condition have been joined by a line.[fig_uvb_btfr_c4p2],scaledwidth=50.0% ] without a uvb , all our models form too many stars and are able to keep a significant amount of neutral gas , making them trace out a btfr that consistently lies above the observed relation ( fig .[ fig_uvb_btfr_c4p2 ] ) .we need a high stellar feedback parameter ( 1.0 ) to somewhat suppress star formation in these models and even this high value is clearly not sufficient @xcite . , scaledwidth=50.0% ]we already saw in section [ subsection_convergence ] that this behaviour dramatically changes when including a uvb : uvb heating after the initial star formation peak quickly disperses most of the neutral gas that is heated by stellar feedback and prevents the gas from cooling and falling in again .this means that the sfh is qualitatively different from the one in the reference model without uvb ( fig .[ fig_uvb_sfh ] ) . after the initial peak, further star formation is strongly suppressed in the models with uvb and even completely stops in the lowest mass models . .for clarity , simulations representing the same initial condition have been joined by a line.[fig_uvb_btfrs],scaledwidth=50.0% ] when comparing different models for the uvb , the btfr turns out to be very resilient against changes in uvb model .this is illustrated in fig .[ fig_uvb_btfrs ] , where we compare models with respectively a late and low uvb , a late and full uvb , and an early and full uvb ( see section [ subsection_uvb_model ] for the parameters of these models ) .the three models trace out a very similar btfr .the reason for this resilience is twofold .on the one hand , the strength of the uvb heating does not scale linearly with the strength of the uvb .for instance , when the uvb intensity is decreased , the neutral fraction of the gas will go up , which means there is more neutral gas that can absorb uv radiation . as a result , the uvb heating rate will be almost unchanged .if we really want to lower the strength of the heating with a considerable factor , we would need to use unphysically low uvb intensities . on the other hand ,the timing of the onset of the uvb is completely masked by the first star formation peak .although there is a significant difference in redshift between an early and a late uvb , there is only a small difference in time between these events at these high redshifts .this means that the uvb will still kick in early in the simulation , while the gas is collapsing and causing the first large star formation peak .the effect of either uvb on the galaxy will only become noticeable after this first peak , when the gas is dispersed by feedback from the first stars . at this moment , both the early and the late uvb will already have kicked in and their strengths will be comparable ..[fig_merger_sfh],scaledwidth=50.0% ] it is clear that many of the problems with our models are caused by the large star formation peak at the beginning of the simulation . to check whether this initial peak is caused by the idealized initial conditions that neglect cosmological effects , we will investigate the star formation history of the merger simulation dg10e9-np3 .this is shown in fig .[ fig_merger_sfh ] . from fig .[ fig_btfr_all ] , it is already clear that this model lies significantly above the observational btfr .compared to an isolated model with a comparable final halo mass , the merger simulation shows an overall lower star formation rate , so that the final stellar mass is roughly a factor two lower than for the isolated model .the initial star formation peak however still stands out as a clear feature in the star formation history .we conclude that cosmological effects alone are not enough to reduce the initial star formation peak .we do however expect that models that are able to reduce this peak in isolated simulations , will perform even better when cosmological effects are taken into account @xcite .this falls outside the scope of this work , where we focus on the effect of internal feedback mechanisms on the initial star formation peak ..[fig_overcooling_btfr],scaledwidth=50.0% ] before we can investigate the effect of varying the stellar feedback efficiency parameter , we have to address the effect of over - cooling on our simulations .as discussed in section [ subsection_model_stellar_feedback ] , we ad hoc switch off cooling for gas particles that receive feedback from sw and snii , to allow them to go through a phase of adiabatic cooling before radiative cooling starts to radiate away the feedback energy . if we would not do this , most of the feedback energywould be radiated away immediately , leading to very inefficient feedback .the resulting galaxies form way too many stars and are far above the observed btfr , as shown in fig .[ fig_overcooling_btfr ] . , scaledwidth=50.0% ] the excessive star formation is entirely due to the first peak in star formation , which immediately consumes all the gas in the simulation , as illustrated by the sfh in fig .[ fig_overcooling_sfr ] . in section [ subsection_uvb ], we discussed the influence of an external heating mechanism on the btfr , and we showed that , if this external heating is provided by an uvb , this influence is very small . in this part , we will examine the influence of internal heating processes , which are due to feedback from massive stars .for the bulk of our simulations , we used a feedback parameter of 0.7 . to assess the influence of the parameter , we compare it with a model with respectively a feedback parameter of 1.0 and of 2.0 .the latter is strictly speaking unphysical , although due to numerical resolution issues we can not guarantee that all energy that is put into the gas is ultimately used to heat the gas , so in this sense a value of 2.0 might be justified . .for clarity , simulations representing the same initial condition have been joined by a line.[fig_feedback_btfrs],scaledwidth=50.0% ] the results of the comparison are shown in fig .[ fig_feedback_btfrs ] .the effect is considerable : the models with a higher feedback parameter generally have a lower baryonic mass and the more massive models tend to also have lower circular velocities .the models with feedback parameter 2.0 are completely absent on the mock observational btfr , since none of them has any neutral gas mass left at the end of the simulation ., scaledwidth=50.0% ] if we only look at the final neutral gas mass of the models ( fig . [ fig_feedback_gtfrs ] ), we see that the feedback parameter regulates the cut - off mass for which halos are no longer able to keep neutral gas : this cut - off mass is higher for the models with a higher feedback parameter .decreasing the feedback parameter below 0.7 will only shift the btfr to higher baryonic masses , while it is already too high at the high velocity end .although we could expect this to lead to more realistic final neutral gas masses , it will also lead to excessive stellar masses .apart from slightly shifting the total stellar mass and the circular velocity to lower values , increasing the feedback parameter does hence not help to produce galaxies that lie on the observational btfr and even has a detrimental effect on the final neutral gas mass . in section [ subsection_uvb ] and section [ subsection_feedback ] , we discussed the effect of external and internal heating on the resulting btfr for our models .for the case of the external uvb , we considered two aspects of the external feedback : its strength and its timing . both are well constrained and hence leave us little to play with .for the internal stellar feedback , we only discussed the strength .the timing of the feedback is in this case also well constrained : stellar winds and snii feedback occur shortly after the star population was born , while snia feedback occurs with some delay and is more spread out over time .there is no reason why the feedback strength should vary with time , since temporal changes in energy absorption in the ism are handled by the metallicity dependence of our gas cooling model .an exception to this are pop iii stars , which are born from very low metallicity gas early in the simulation and have a different feedback compared to other stars .since they enrich the ism with their feedback , pop iii stars will only be born early on in the simulation , which effectively causes the stellar feedback to change over time .this could provide a solution to the problems caused by the initial large star formation peak , since this gives us a mechanism to reduce this first peak without affecting star formation in later stages of the simulation ..[fig_popiii_stats],scaledwidth=50.0% ] fig .[ fig_popiii_stats ] shows the latest formation time of a pop iii star as a function of the final stellar mass for all simulations including pop iii feedback .it is clear that pop iii stars are indeed only formed at the early stages of the simulations , so that pop iii feedback is limited to these early stages .in section [ subsection_popthree_models ] we introduced three models for pop iii feedback .model 1 only affects the sn feedback from pop iii stellar particles , model 2 also includes a pop iii sw , while model 3 implements more advanced metal yields .we will discuss the effect of the different models separately .model 1 has two variants : variant a assumes a lower limit of 60@xmath85 on the mass of pop iii stars , while variant b has a lower limit of 140@xmath85 and hence returns its energy over a much shorter time interval . . for clarity ,simulations representing the same initial condition have been joined by a line.[fig_popiii_model1a_btfrs],scaledwidth=50.0% ] , scaledwidth=50.0% ] fig .[ fig_popiii_model1a_btfrs ] shows the btfr for the three realizations of model 1a , which have different pop iii feedback energies .it is clear that the realization with a high feedback energy results in gas poor galaxies which have very little stars .the other two realization lead to similar btfrs . in fig .[ fig_popiii_model1a_sfr ] we show the sfh for the same ics and the low energy and middle energy realization of the model .the pop iii feedback effectively reduces the height of the initial star formation peak , while at the same time leaving the further star formation history untouched , apart from stochastic differences .we also notice that the initial star formation peak is not completely reduced . further increasingthe pop iii feedback does not work to completely reduce the peak , as we can conclude from the high feedback model . .for clarity , simulations representing the same initial condition have been joined by a line.[fig_popiii_model1b_btfrs],scaledwidth=50.0% ] , scaledwidth=50.0% ] fig .[ fig_popiii_model1b_btfrs ] and fig .[ fig_popiii_model1b_sfr ] show the btfr and sfr for model 1b , which gives pop iii feedback over a much shorter time interval ( and hence gives stronger feedback over a shorter period of time compared to model 1a ) .we see that this model effectively completely reduces the initial star formation peak .we can conclude that the time dependence of stellar feedback introduced by pop iii feedback significantly changes the strength of the initial star formation peak . . for clarity ,simulations representing the same initial condition have been joined by a line.[fig_popiii_model2_btfrs],scaledwidth=50.0% ] , scaledwidth=50.0% ] to completely suppress the initial star formation peak , we have to assume pop iii feedback over an extremely short time interval , which corresponds to unrealistically high masses for pop iii stars .we can get the same effect if we use a more realistic pop iii mass range and also include pop iii sw .[ fig_popiii_model2_btfrs ] and fig .[ fig_popiii_model2_sfr ] show the btfr and sfh for the two models including pop iii sw .one model has high sw feedback and low sn feedback and vice versa .both show a significant suppression of the initial star formation peak and further star formation and are able to keep realistic amounts of neutral gas for the more massive ics .however , they both form too many stars and have very low circular velocities . .for clarity , simulations representing the same initial condition have been joined by a line.[fig_popiii_model3_btfrs],scaledwidth=50.0% ] in the last model , we also include pop iii metal yields in our feedback model , taking into account the fact that pop iii stars will return less metals to the ism than pop i and pop ii stars .this in turn will affect the cooling and heating of the gas and can influence the further star formation .the results of this model are very similar to those of model 2 , apart from some stochastic differences .they trace out the btfr shown in fig .[ fig_popiii_model3_btfrs ] , which is still a bit higher than the observed relation , but has the same slope .in this paper , we explored a large range of sub - grid models and parameters in order to produce simulated dwarf galaxies that lie on the observed baryonic tully - fisher relation and are gas rich and star forming at late redshifts , like the observed ultra - faint dirrs leo p , leo t and pisces a. we found that the bulk of our simulations experience a large peak in star formation around the time the uvb kicks in , which then leads to an excess in stellar feedback that drives the pre - heated interstellar medium out of the galaxies and leaves them gas poor and dead . as a result ,most simulations form too many stars , do not have enough neutral gas and consequently lie above the observed btfr .they are also too metal poor compared to observed dwarf galaxies .the star formation peak is unaffected by changing the timing or intensity of the uvb within a physically meaningful range due to the self - regulating character of the uvb heating and the fact that uvb heating only starts affecting the galaxies at low redshifts , when the uvb is already strong .changing the stellar feedback strength affects the absolute height of the initial star formation peak and consequently shifts the resulting galaxies on the btfr , but does not help in reducing the relative height of the peak with respect to the consecutive star formation , mainly due to the self - regulating character of the star formation .the only way to reduce the initial star formation peak with respect to the consecutive star formation is by including a time dependence in the stellar feedback , caused by the metallicity dependent feedback of pop iii stars .we explored the properties this pop iii feedback needs to have in order to produce realistic dwarf galaxies and found that in the most successful model , a pop iii stellar population returns 40 times more energy in the form of uv radiation early during its lifetime than a normal stellar population ( which we call stellar wind feedback ) , and 4 times more energy in the form of supernova explosions from massive stars .these values are well within the current constraints from simulations of primordial stars @xcite .furthermore , a pop iii stellar population loses 45 per cent of its mass due to stellar feedback , of which 0.009327 per cent is fe and 0.01514 per cent is mg .we note that the time dependence of the stellar feedback introduced by pop iii feedback is similar to introducing a time dependent star formation efficiency as discussed in @xcite , although it is unclear whether the effect of the latter will be as strong . with our new pop iii feedback prescriptions ,we succeed in reproducing the slope of the observed btfr and we obtain galaxies with more realistic metallicities. our simple model does however not completely solve the problems discussed above , because the simulations still form slightly too much stars and loose too much neutral gas .this is mainly caused by the fact that we still consider galaxies in isolation , while galaxies in a @xmath86cdm universe are formed by subsequent mergers of smaller halos , which has considerable effect on the properties of these galaxies @xcite .@xcite showed that including the merger history of a halo also helps reducing the initial star formation peak , leading to galaxies that have more realistic stellar metal contents .although computationally more efficient than cosmological simulations , these merger simulation are not suited for a parameter study of the size considered for this work .using our advanced sub - grid model including pop iii feedback for a smaller set of merger simulations is subject of verbeke et al .( submitted ) .bv and sdr thank the ghent university special research fund for financial support . rvthanks the interuniversity attraction poles programme initiated by the belgian science policy office ( iap p7/08 charm ) .we thank volker springel for making gadget2 publicly available .our sub - grid model makes use of the chianti atomic database , a collaborative project involving george mason university , the university of michigan ( usa ) and the university of cambridge ( uk ) .

<S> simulating dwarf galaxy halos in a reionizing universe puts severe constraints on the sub - grid model employed in the simulations . using the same sub - grid model that works for simulations without a uv - background ( uvb ) results in gas poor galaxies that stop forming stars very early on , except for halos with high masses </S><S> this is in strong disagreement with observed galaxies , which are gas rich and star forming down to a much lower mass range . to resolve this discrepancy , we ran a large suite of isolated dwarf galaxy simulations to explore a wide variety of sub - grid models and parameters , including timing and strength of the uvb , strength of the stellar feedback , and metallicity dependent pop iii feedback . </S><S> we compared these simulations to observed dwarf galaxies by means of the baryonic tully - fisher relation ( btfr ) , which links the baryonic content of a galaxy to the observationally determined strength of its gravitational potential . </S><S> we found that the results are robust to changes in the uvb . </S><S> the strength of the stellar feedback shifts the results on the btfr , but does not help to form gas rich galaxies at late redshifts . only by including pop iii feedback </S><S> are we able to produce galaxies that lie on the observational btfr and that have neutral gas and ongoing star formation at redshift zero . </S><S> [ firstpage ] galaxies : dwarf galaxies : evolution galaxies : formation methods : numerical </S>

cond-mat0104447

the ground state of the quasi - two dimensional organic conductor has emerged as one of the more important fundamental problems in the area of synthetic metals .this is due to the highly unusual magnetic field dependent behavior of the low temperature ground state which appears below a transition temperature @xmath5 = 8 k. the band calculation and the fermi surface study@xcite suggest a two dimensional fermi surface consisting of a quasi - one dimensional electron section and a closed hole pocket as shown in fig .[ fig1 ] ( a ) .the density wave formation is believed to originate from the instability of the open orbit at low temperatures , which is followed by fermi surface reconstruction@xcite .this in turn leads to the unusual behavior of this compound at low temperatures and high magnetic field@xcite .this ground state is now thought to be related to a charge density wave state where both spin and orbital mechanisms compete , as evidenced by recent studies in tilted magnetic fields@xcite . above 22.5 t an anomaly ( called the `` kink field '' or @xmath6 after its discovery by osada et al ., @xcite which appears in various transport - related measurements ) occurs , and a new high field electronic state emerges .recently the observation of critical state - like behavior in the magnetization@xcite and susceptibility @xcite above @xmath6 has been reported , which indicates an unusual , highly conductive state which is associated with the landau level filling fraction .additional thermodynamic evidence for unusual , hysteretic behavior above @xmath6 , which is related to the landau level filling , has also been reported in high field magnetocaloric measurements @xcite .the purpose of the present work has been to probe this novel ground state by thermoelectric power ( tep ) , which involves measuring the potential difference across a sample in the presence of a thermal gradient , as shown in fig . 1 ( c ) .in general , the resulting signal ( in v / k ) has the form@xcite @xmath7 where @xmath8 is the density of the states , @xmath9 is an average charge velocity , and @xmath10 is the carrier scattering relaxation time .in addition , we have also measured the nernst effect for specific sample configurations .this involves the transverse thermopower which is given in the simple approximation by @xmath11 , where @xmath12 is the nernst coefficient , @xmath13 is the magnetic field , and @xmath14 is the ratio of the distance between the nernst leads and the sample thickness .we note that ideally , for closed orbits , @xmath15 is not sensitive to electron / hole compensation effects , and that it increases linearly with magnetic field .( see discussion below ) .clearly the tep and nernst signals can be a complex mixture of phenomena which can be difficult to interpret , unless specific aspects of the system under consideration dominate and/or change dramatically@xcite .examples include the case of a conventional metal ( @xmath16 ) , a superconductor ( @xmath17 ) , and the quantum hall effect ( @xmath18 0 or a finite value depending on the landau level filling ) .as we will show in the present case , the tep probe for is particularly sensitive to the landau level spectrum , to the gap in the electronic structure , and to the in - plane electronic anisotropy . in light of the discussion above , we state herethe main findings of the present work before the full presentation . , where comparisons can be made with previous studies ( all of which are in - plane , and almost exclusively at low or zero magnetic fields ) , we find that our results agree in the details of the sensitivity of the thermopower to the anisotropic fermi surface topology , and with absolute value of the thermopower signals previously reported .in addition , we extend these measurements below @xmath5 , where the tep reveals details of the transition into the density wave state . , we observe , in the oscillatory metp , quantum oscillations associated with the main closed fermi surface orbit frequency , and also oscillations associated with fermi surface reconstruction and anomalous second harmonic signals which result from the density wave ground state .we believe that this is the first detailed study of quantum oscillations in a quasi - two dimensional organic conductor by thermopower methods . , the background ( non - oscillatory ) metp is sensitive to changes in the density wave ground state with magnetic field .this includes the signatures for three changes in the fermi surface topology , one near 3.75 t , a second above 10 t , and a third above 22.5 t. the first two are associated with magnetic breakdown effects , and the latter is the first order phase transition to the high field state .the temperature - magnetic field phase diagram , based on the metp results , is presented ., we study in detail for the first time the inter - plane thermopower , which gives additional information about the role of the inter - plane transport mechanism of this material ., in the high field state , we describe the phase relationship between the nernst signal and the metp signal , which is consistent with theoretical predictions .to our knowledge , no thermopower measurements have been carried out on the material in very high magnetic fields . in 1950s , a series of calculations were done for high field dependence of magnetotransport properties .it was found that , in the limit of @xmath19 ( where @xmath20 is the cyclotron frequency and @xmath21 is the scattering time ) , asymptotic magnetotranport properties ( electrical conductivity@xcite , thermal conductivity@xcite and thermopower@xcite ) are highly dependent on the fermi surface topology and carrier compensation , while they weakly depend on the scattering time and the energy dispersion relations .it was therefore suggested that those measurements would be useful to reveal the fermi surface topologies . due to the complex nature of thermopower signals in anisotropic materials , it is useful to briefly consider previous studies in organic materials , including those at high magnetic fields where field induced phase transitions occur . for consistency in the anisotropic thermopower parameters reported below , we use the following notation : the measured values can be written as @xmath22 , where @xmath23 , @xmath24 or @xmath25 , signify that the heat current is applied along a geometrical direction @xmath26 and the thermal emf is measured along a direction @xmath27 , and @xmath0 is one of the crystallographic axes along which the heat current is applied .for zero magnetic field , @xmath22 can be written in the reduced form such as @xmath28 , or simply @xmath29 .mori et al .@xcite performed the first in - plane measurements on , where the temperature dependence of @xmath30 and @xmath31 were found to be consistent with the topology of the hole and the electron bands ( see fig .1 ) . here , in general , the tep is positive when the thermal gradient is parallel to the open orbit sheets ( where the closed hole orbits dominate the signal ) , and negative when the thermal gradient is perpendicular to the open orbit sheets ( where the electron orbits are the main contribution ) . our results , shown in fig . 2 ( and to be discussed below ) , are in reasonable agreement with these measurements .this assignment is supported by a simulation of the thermopower , based on boltzmann transport theory , with the bandwidth parameters as input@xcite .another experiment involving in - plane thermopower has been performed on the 10.4 k organic superconductor@xcite @xmath32 - ( bedt - ttf)@xmath1cu[n(cn)@xmath1]br ( see also gartner et al .@xcite and buravov et al .this system also has open and closed orbit fermi surface bands , and again , the sign of the tep in the normal state above @xmath34 follows the general rule that it is positive when the thermal gradient is parallel to the electron sheets , and negative when perpendicular .below @xmath34 , the tep vanished , as expected for a superconductor in zero magnetic field . in studies of quasi - one dimensional ( q1d ) systems based on the tcnq charge transfer salts @xcite and the perylene based compounds@xcite ,various features of tep were exhibited .depending on the species of cations and the temperature range , they show linear temperature dependence , 1/@xmath35 behavior , or temperature independence .the linear decrease of tep was treated by a simple , single carrier in 1d tight binding band model , where the temperature behavior followed a linear @xmath35 dependence , @xmath36 .here @xmath37 is the bandwidth and @xmath38 is the number of carriers per site . below the metal - to - insulator transition ,the low temperature behavior followed the 1/@xmath35 form @xmath39 $ ] where @xmath40 is the gap and @xmath41 is the ratio of the electron to hole mobilities . the temperature independent thermopower with magnitude of @xmath42v / k was attributed to the spin configuration entropy ( @xmath43 ) in the strong coulomb potential limit ( @xmath44 ) in a quarter filled band . in the presence of an applied magnetic field , the transverse thermopower ( the nernst effect )may be measured in addition to the normal longitudinal tep ( the seebeck effect ) .logvenov et al .@xcite have investigated the superconducting states of two @xmath32 - ( bedt - ttf)@xmath1x materials up to 3 t. the results , which showed non - zero tep values even without magnetic fields , were interpreted in terms of flux motion and the magnus force ( in the case of the nernst effect ) .the authors found considerable anisotropy which was attributed to the difference in the electron and hole fermi surface sections .the quasi - one dimensional bechgaard salts , which exhibit spin density wave ( sdw ) formation have been investigated by tep and mtep in some detail . the sdw transition in ( tmtsf)@xmath1asf@xmath45 at ambient pressure( t@xmath46= 12 k ) has been investigated @xcite up to 11.3 t. hysteretic temperature effects in the thermopower were observed as a function of field direction below t@xmath46 , and the authors speculated about pinning , structural changes , and sub - phases associated with the sdw ground state to describe the data . for systems with magnetic fieldinduced spin density wave states ( fisdw ) , two thermopower studies to 30 t have done on the quasi - one dimensional organic conductor series ( tmtsf)@xmath1clo@xmath2 @xcite and ( tmtsf)@xmath1pf@xmath45 under pressure@xcite . in these materialsa fisdw transition occurs from a q1d metallic state to a sequence of sdw states .both the mtep and nernst signals , which bear some relation to the corresponding transport measurements ( i.e. longitudinal resistivity @xmath47 and the hall effect @xmath48 respectively ) , are sensitive to these transitions .as the final fisdw sub - phase is entered at constant field with lower temperatures , the thermopower first rises with an activated ( gap - like ) behavior , followed by a decrease at lower temperatures deeper in the sdw phase .the behavior is interpreted in terms of collective modes within the sdw phases .notably , quantum oscillations ( the so - called `` rapid oscillations '' which are periodic in inverse field ) are observed in the thermopower for both materials .these experiments show the usefulness of thermopower measurements to probe ground states that are induced at high magnetic fields . in this paper , we report thermopower measurement results carried out under high magnetic field and at low temperatures in .the heat current was applied along all three crystallographic axes , and the longitudinal ( seebeck coefficient ) and the transverse ( nernst - ettinghausen effect ) thermopower was measured .the asymptotic behaviors of the seebeck coefficient and the nernst - ettinghausen effect suggest that the magnetic breakdown of the closed orbit in the reconstructed fermi surface are responsible for the observed behavior of magnetothermopower .single crystal samples were obtained from the conventional electrochemical crystallization technique .the orientation of crystallographic axes were determined from polarized ir reflection measurements at room temperature .three different samples from a single batch were used for each three different experimental conditions ( heat current @xmath49 @xmath50-axis , @xmath41-axis or @xmath51-axis ) .the magnetic field was always applied along the least conducting axis ( @xmath51-axis ) . when the heat current was applied in the most conducting plane ( @xmath52-plane ) , off - diagonal term of thermopower tensor ( @xmath53 )was also measured simultaneously .the polarity of magnetic field was switched for the measurement of @xmath53 to remove spurious components due to the misalignment of the voltage leads .( the alignment of voltage leads for @xmath54 measurements was also checked in this manner . )thermopower measurements were carried out by applying sinusoidal heat currents along a crystallographic axis of the single crystal and measuring thermal emf along the same direction and along the direction rotated by 90 degrees .( see fig .[ fig1 ] ( c ) ) the sample was mounted between two quartz blocks , which were heated by sinusoidal heating currents ( with oscillation frequency @xmath55 = 66 mhz ) with @xmath56/2 phase difference .the corresponding temperature gradient ( ) and thermal emf with 2@xmath55 oscillation frequency were measured .the magnetothermopower of au lead wires was in situ calibrated by yba@xmath1cu@xmath57o@xmath58 superconductor . to avoid an ambiguity in determination of under magnetic field, we exploit the reproducibility of , which only weakly depends on magnetic field .the procedures of the magnetothermopower measurement method used in this work are detailed elsewhere@xcite .the temperature dependence of zero field thermopower results for , along each of the principal axes is shown in fig .[ fig2 ] , where the inset shows low temperature expansion near @xmath5 = 8 k. the in - plane results are in general agreement with previous measurements by mori et al.@xcite , as discussed in sect .the negative ( positive ) sign of thermopower along the @xmath59-axis indicates that the fermi surface is electron ( hole)-like along the @xmath59-axis , which corresponds to the fermi surface as shown in fig . 1 : here , for a thermal gradient along the @xmath41-axis direction , the hole pockets should contribute , whereas for the @xmath50-axis direction , the electron - like open orbits should contribute as well .we note however , that the temperature dependence of the tep shown in fig .[ fig2 ] is non - monotonic .there is a minimum in @xmath60 and @xmath61 around t = 120 to 150 k .this behavior can be explained by applying the two - dimensional tight binding parameters for this material to the boltzmann transport equation@xcite,@xcite .the inter - plane tep ( @xmath62 ) , perpendicular to the conducting planes , which has not been previously measured , shows a monotonic decrease until 35 k , where it changes sign .the positive sign of @xmath62 above @xmath35 = 35 k indicates that the hole - like carriers are dominant for the inter - layer transport above @xmath35 = 35 k and the electron - like carriers dominate below that temperature .clearly , the inter - plane transport shows a distinctly different behavior to that of the in - plane tep . from the linear dependence of @xmath60 and @xmath62 at higher temperature, the density of states per carriers are derived to be 5.3 states / ev for @xmath60 and 7.6 states / ev for @xmath62 .the deviation from the linear dependence in the intermediate temperature range has been treated@xcite by mori et al ., as mentioned in sect .ii above , as has the peak in @xmath60 near 25 k .the peak in @xmath60 may arise from the energy - dependent scattering term of eq .[ mott ] , or from the phonon drag effect .the tep associated with phonon drag results from the electron - phonon coupling .the temperature dependence changes from a @xmath63 to a @xmath64 behavior @xcite with increasing temperature .especially , peak structures at low temperature ( typically between @xmath65/10 and @xmath65/5 , where @xmath65 is the debye temperature ) are usually attributed to the phonon drag effect , whose temperature dependence in noble metals is quite similar to that of @xmath60 .thermopower measurements on the other members of this compound ( @xmath0-(bedt - ttf)@xmath1mhg(scn)@xmath2 , where m = tl , rb , nh@xmath2 ) also showed peak structures at lower temperatures ( 15 @xmath66 k ) , which were attributed to the phonon drag effect@xcite . of noteis the superconductivity of the nh@xmath2 salt , where there is an enhancement of density of states and effective mass ( m * @xmath67 for @xmath32-(bedt - ttf)@xmath1nh@xmath2hg(scn)@xmath2 over that of non - superconducting ( m * @xmath68 ) ) , as pointed out by mori et al.@xcite however , the origin of enhancement of m * , which may be due to electron - electron interaction and/or electron - phonon interaction , is not yet clear .when the heat current is applied perpendicular to the most conducting planes , the 2d band model will not be appropriate without considering the energy dispersion between conducting planes .since simple mixing of two contributions of @xmath61 and @xmath60 can not produce temperature dependence and magnitude of @xmath62 , it is necessary to find a proper model for inter - plane transport .a straight - forward application of the @xmath69 relationship produces unrealistic values of the @xmath51-axis bandwidth ( of order 250 mev ) , which is known from fermiology studies to be at least 100 times smaller . at presentthere is no satisfactory treatment of either the magnetoresistance or the thermopower for the @xmath51-axis transport behavior .the behavior of the thermopower through the density wave transition ( @xmath5= 8 k ) has also not been previously reported . from the inset of fig .[ fig2 ] , jumps of thermopower for @xmath61 and @xmath60 are observed near the metal - density wave transition temperature ( @xmath5 ) . in the resistance measurements , the change of slope near@xmath5 may be attributed to the vanishing conductivity of the open orbit band ( @xmath70 ) in parallel with the remaining finite conductivity of closed orbit band ( @xmath71)@xcite .in addition , for thermopower measurements , metallic carriers give a linear @xmath35-dependence while activation over a band gap gives a 1/@xmath35 dependence .therefore for one type of carrier ( either from open orbit or closed orbit ) exclusively , @xmath61 will diverge below @xmath5 .experimentally , there is always the possibility of random diffusion of heat along the sample due to misalignment between heat currents and crystal axes , and also , the total thermopower may involve a mixing of contributions from several bands ( especially below @xmath5 where the fermi surface is reconstructed ) .nevertheless , we believe that dominant behavior of thermopower in the present case reflects the influence of the heat currents with respect to the topology and orientation of the specific orbits .when two types of carriers are involved , the total thermopower can be written as@xcite , @xmath72 above @xmath5 , both orbits have metallic conductivity and thermopower ; hence the total thermopower shows a linear temperature dependence . but below @xmath5 , a band gap opens for @xmath73 band , therefore @xmath70 goes to zero exponentially as the temperature decreases , and @xmath74 diverges as 1/@xmath35 .however , below this temperature , a modified closed orbit band remains , which still gives a metallic thermopower contribution .consequently , just below the transition temperature , there will be a jump of thermopower , @xmath75 , from the divergence of @xmath74 , but this contribution quickly disappears as @xmath70 goes to zero .we may further compare the magnitude of jump of thermopower ( @xmath75 ) and that of electronic heat capacity ( @xmath76 ) .since thermopower measures heat carried by a carrier divided by the carrier charge and temperature , @xmath75 equals @xmath76/@xmath77 . adding both contributions from @xmath61 and @xmath60 , @xmath75 is about 1 @xmath78v / k , which corresponds to @xmath79 @xmath80 j/ k@xmath81carrier .the reported @xmath76 is about 0.1 j / mol@xmath81k@xcite and 0.25 j / mol@xmath81k@xcite , which are in fair agreement with @xmath75 .a corresponding change in @xmath82 at @xmath5 is unclear , if it exists at all ( in fig .this indicates that the phase transition around @xmath5 = 8 k occurs predominantly within the in - plane electronic structure . in fig .[ fig3 ] , we present the magnetic field dependence of the tep ( i.e. mtep ) for all three crystallographic directions at different , fixed temperatures . to show the field dependence more clearly , some traces are offset from zero as indicated by the dashed lines .the mtep signal contains two components , the background mtep which is sensitive to the fermi surface topology , and the oscillatory mtep which is a manifestation of the landau quantization of the closed orbit fermi surface . to separate the two signals , we show in fig .[ fig4 ] the background mtep data where the oscillatory mtep has been removed by filtering the total mtep signal .finally , the nernst - ettinghausen effect at different temperatures is shown in fig .[ fig5 ] . in all casesthe magnetic field is applied along the @xmath51-axis , perpendicular to the @xmath83 conducting plane .a similarity between the magnetic field dependence of the thermopower and the resistance can be expected from the boltzmann transport equations .indeed we find that the general field - dependence of the mtep is similar in some respects to that of the magnetoresistance@xcite : ( 1 ) below @xmath5 , the in - plane mtep shows a broad peak at @xmath84 10 t. ( 2 ) above the `` kink field '' @xmath6 at 22.5 t , the field dependence is weak .( 3 ) quantum oscillations are observed in the mtep signal associated with the de haas - van alphen ( dhva ) and shubnikov - de haas ( sdh ) effects ( i.e. , from eqn .1 , both thermodynamic and transport properties are involved in the thermopower ) .the oscillation amplitudes are largest for @xmath85 , and the oscillations are also observed in @xmath86 .( 4 ) the mtep also shows hysteresis for up and down field sweeps . beyond the general comparison between resistivity and mtep , unique field - dependent features are observed in the mtep and nernst signals : ( 1 ) for @xmath87 and @xmath85 , the mtep exhibits a minimum at @xmath88 after which it rapidly increases up to maximum at @xmath89 .@xmath90 is also temperature dependent , and decreases with increasing temperature .( 2)for the @xmath50-axis behavior , the nernst signal @xmath91 shows a similar field dependence to the mtep signal @xmath87 in general , but @xmath91 is asymptotic to zero above @xmath6 for all temperatures measured . in contrast, @xmath87 is asymptotic to non - zero , temperature dependent values above @xmath6 .\(3 ) for the @xmath41-axis behavior , @xmath92 shows linear field dependence at higher field , which is distinct from the other thermopower coefficients .\(4 ) for the @xmath51-axis , @xmath86 exhibits some aspects of @xmath87 and @xmath85 mentioned above , which may involve some mixing of the in - plane components .[ fig6 ] shows the temperature dependence of magnetothermopower for fixed magnetic field . as the magnetic fieldis increased , the transition temperature @xmath5 , represented as a jump or shoulder in the data , shifts to lower temperature .( for @xmath86 and @xmath85 at @xmath13=27 t however , @xmath5 could not be resolved . ) the temperature - field phase diagram , based exclusively on the field and temperature dependent thermopower data ( @xmath93 , and @xmath94 ) , is shown in fig .the phase diagram is very similar ( with the exception of the new @xmath90 line ) to previous determinations by transport@xcite , nmr@xcite , magnetization@xcite , and specific heat@xcite . to interpret the origin of the field dependence of the mtep , it is useful to examine the magnetic field dependence of the thermopower for an anisotropic system .when thermopower is measured under magnetic field , tensor elements of thermopower should be considered since both the temperature gradient and the magnetic field are vectorial . for our experimental conditions ( zero electric current and zero transverse heat current ) ,the kinetic equations for thermopower and heat currents can be written as , @xmath95 where @xmath96 is the electric field produced by the thermal emf , @xmath0 is the thermopower tensor under magnetic field , @xmath97 is the heat current , and @xmath98 is the thermal conductivity tensor .we have two different experimental situations .\(a ) @xmath99 @xmath49 @xmath100 axis @xmath49@xmath97 @xmath49 @xmath101 ( @xmath86 in this paper ) : in this case , no lorentz force is applied so that @xmath102= @xmath103 = 0 , and above equations have simple scalar form , @xmath104 the measured value @xmath86 is obtained from @xmath105/@xmath106 .\(b ) @xmath99 @xmath49 @xmath100 axis @xmath107 @xmath97 ( @xmath87 , @xmath91 , @xmath85 and @xmath92 in this paper ) : in this case , is not parallel with @xmath97 , and has a non - zero @xmath24 component given by -@xmath102@xmath108/@xmath109 . if we rewrite eq .[ kinetic ] for this case , @xmath110 in terms of measured values , eq . [ sxy ] can be expressed as , @xmath111 @xmath112 the asymptotic behavior of thermoelectric tensor @xmath113@xcite and thermal conductivity tensor @xmath114@xcite were calculated at low temperature and high magnetic field .they show different behavior for the three different cases , i.e. , ( a ) closed and compensated orbits ( b ) closed and uncompensated orbits and , ( c ) open orbits along the @xmath115direction . for each case , the asymptotic behaviors of @xmath116 can be calculated and the results are summarized in table [ table1 ] . the saturation values are temperature dependent for each case , and they should be determined by considering the scattering and the dispersion relations .we now compare the thermopower data with the theoretical field dependence described above .we focus our attention on the behavior of the @xmath60 signal shown in fig .3 and 4 , since it is significantly larger than the other signals , and it is predominantly hole - like ( i.e. positive , and closed orbit - sensitive ) .we may use as a guide the detailed work by uji et al . on the fermiology study of the quantum oscillations@xcite , due to closed orbits in the magnetic breakdown network , which have been proposed to describe the fermi surface topology below @xmath5 . in this modelit is assumed that only small closed orbit pockets exist in the reconstructed fermi surface below @xmath5 near zero magnetic field . as the magnetic fieldis increased , larger and larger fermi surface sections become involved in the sdh oscillations as magnetic breakdown becomes more probable .these effects appear experimentally in the sdh measurements : in figs . 8 and 9 of ref .[ 45 ] , the onset of the sdh oscillations associated with the various magnetic breakdown orbits ( @xmath0 , @xmath117 , etc . ) occurs above 3.5 t. at higher fields , above 7 t , the @xmath118 orbit associated with the brillouin zone appears , i.e. magnetic breakdown occurs over most of the fermi surface . finally , above the kink field ,the original fermi surface topology ( as in fig .1 ) re - emerges . in light of the above, our first point of comparison is the behavior of @xmath90 . for @xmath60 , and indeed in all mtep directions , there is a distinct change in the signal above @xmath90 from a relatively flat response to a pronounced increase in slope ( approximately linear ) with field .this behavior is entirely consistent with the theoretical model above for a set of closed , compensated orbits . at higher fields , above @xmath119 , the compensated closed orbit behavior of @xmath60 changes dramatically , and this indicates that the magnetic breakdown network of closed orbits is undergoing modification with increasing field . by comparison with the theory above, we may speculate that this is the onset of uncompensated closed orbit behavior , which exhibits less ( eventually no ) field dependence .finally , above the kink field , the original , uncompensated closed orbit fs is realized , and the field dependence disappears , as is evident from the @xmath60 data . above @xmath120, we may compare the temperature dependence of @xmath60 with those at zero field above @xmath5 ( see fig .[ fig6 ] ) . except for fielddependent offsets , the thermopower above @xmath94 for the three directions is similar , save for the case of @xmath62 which is anomalous , even in the absence of field . beyond this simple comparison of the most dominant mtep signal with previous sdh studies and theoretical expectations , the assignment the mechanisms responsible for the magnetic field dependence of @xmath61 and @xmath62 are more speculativethis is due to the apparent mixing of terms in the experimental data , and to sort them out is beyond the scope of the present work .however , the nernst data shows significant difference for the @xmath50 and @xmath41 axes ( @xmath91 and @xmath92 ) .this is most apparent above @xmath6 , where the original fs is expected to be recovered . here@xmath91 is seen to be asymptotic to zero ( i.e. @xmath121 dependence ) , whilst @xmath92 clearly assumes a linear ( i.e. h@xmath122 ) dependence .this behavior is consistent with expectations based on the theory represented in table 1 . from eq .1 , it is clear that for a quasi - two dimensional fermi surface , the magnetic field dependence of the landau level spectrum@xcite will produce quantum oscillations in the thermopower , which will originate from both thermodynamic and electron transport factors . the latter has been treated by zilberman@xcite and long et al . @xcite where scattering between states of the landau level closest to the fermi surface and states of other occupied are considered .we find in our results a clear indication of the landau level spectrum in the mtep and nernst data , the periodicity of which is in good agreement with previous sdh and dhva measurements@xcite .in particular , the oscillation frequency obtained from the fast fourier transform ( fft ) analysis of @xmath85 and @xmath92 at 0.7 k ( see fig . [ fig8 ] ) is about 667 t , which agrees with results obtained from other measurements@xcite .this comparison also includes the observation of higher harmonics of the fundamental @xmath0 orbit , and even the @xmath117 orbit ( 180 t ) which is expected to be a manifestation of the magnetic breakdown topology below @xmath6 .( see for instance uji et al.@xcite ) .an important feature of our results , most evident above @xmath6 , is a phase difference between the metp and nernst quantum oscillations .this is shown in fig .[ fig9 ] for two different cases , the @xmath50- , and @xmath41-axes . indeed , zilbermans model@xcite predicts a phase difference of @xmath56 between the mtep and nernst oscillations .we show this more explicitly for @xmath85 and@xmath92 at @xmath35=0.7 k in fig .[ fig10 ] , with the angle ( @xmath123 ) defined as the ratio of the electric fields due to transverse and longitudinal thermopower ( @xmath123=arctan ( @xmath124)= arctan(@xmath125 ) .the phase difference is @xmath56 , which supports zilberman s model@xcite .zilberman s model also predicts similar field and temperature dependence of oscillation amplitude with that of the lifschitz - kosevich formula for de haas - van alphen oscillations .the amplitude of thermopower oscillation increases when @xmath126 , but it is almost field independent above @xmath6 .we note that in this range of temperature , the quantum oscillation amplitude is complicated due to the anomalous high field state , where the dhva and sdh behave in a significantly different manner , and a comparison based on standard lk theory is not applicable. a more systematic study of the mtep and nernst effects above @xmath6 vs temperature , and also vs. field orientation , will be necessary to provide a complete picture .we have applied the method of magnetothermopower to investigate the low temperature , magnetic field dependent phase of the anisotropic organic conductor .our results show that the sensitivity of thermopower to anisotropic fermi surface topologies may be extended to the study of systems where temperature and magnetic field alter these topologies .the most significant results of the present investigation include the determination of the low temperature phase diagram of , based purely on thermopower measurements , the study of the onset of magnetic breakdown effects via the magnetothermopower , and the measurement of quantum oscillations in the magnetothermopower and nernst effect , which are in accord with theoretical expectations .this work was carried out at the national high magnetic field laboratory , which is funded by a cooperative agreement between the national science foundation and the state of florida through nsf - dmr 00 - 84173 .jsb acknowledges support from nsf - dmr-99 - 71474 for this work , and esc acknowledges post - doctoral support from kosef .s. tanaka , m. oshima , g. saito , t. mori , y. maruyama , and h. inokuchi , bull .jpn . , * 63 * , 2183 ( 1990 ) .kartsovnik , a.e .kovalev and n.d .kushsch , j. phys .i france * 3 * , 1187 ( 1993 ) .kartsovnik and v.n .laukin , j. phys .i france * 6 * , 1753 ( 1996 ) .t. sasaki , a.g .lebed , t. fukase , and n. toyota , phys .b * 54 * , 12969 ( 1996 ) . j. s. qualls , l. balicas , j. s. brooks , n. harrison , l. k. montgomery and m. tokumoto , phys .b * 62 * , 10008 ( 2000 ) . c. proust , a. audouard , a. kovalev , d. vignolles , m. kartsovnik , l. brossard , and n. kushch , phys . rev .b * 62 * , 2388 ( 2000 ) .t. osada , r. yagi , a. kawasumi , n. miura , m. oshima and g. saito , phys .b * 41 * , 5428 ( 1990 ) .n. harrison , l. balicas , j.s .brooks , m. tokumoto , phys .b * 62 * , 14212 ( 2000 ). n. harrison , c. h. mielke , a.d .christianson , j. s. brooks , m. tokumoto , phys . rev . lett ., * 86 * , 1586 ( 2001 ) . n. a. fortune , m.a. elben , s. uji , h. aoki , j. yamada , s. tanaka , s. make , s. nakatsuji , h. anzai , synth . met . *103 * , 2078 - 2079 ( 1999 ) .d. k. c. mcdonald , _ thermoelectricity _ , john wiley & sons , new york , 1962 .p.m. chaikin , in _ organic superconductivity _kresin and w.a .little , eds . , plenum press , new york , pp101 - 115 , 1990 . i. m. lifshitz , m. ya .azbel and m. i. kaganov , soviet physics jetp * 4 * , 41 ( 1957 ) . m. ya .azbel , m. i. kaganov and i. m. lifshitz , soviet physics jetp * 5 * , 967 ( 1957 ) .bychkov , l. @xmath127 .gurevich and g. m. nedlin , soviet physics jetp * 37 * , 377 ( 1960 ) .t. mori , h. inokuchi , h. mori , s. tanaka , m. oshima and g. saito , j. phys .. jpn . * 59 * , 2624 - 2626 ( 1990 ) .t. mori and h. inokuchi , j. phys .. jpn . * 57 * 3674 ( 1988 ) .r. c. yu , j. m. williams , h. h. wang , j. e. thompson , a. m. kini , k. d. carlson , j. ren , m .- h .whangbo , and p. m. chaikin , phys .b * 44 * , 6932 - 6936 ( 1991 ) .s. gartner , e. gogu , l. heinen , h.j .keller , t. klutz , and d. schweitzer , solid state ., * 65 * , 1531 ( 1988 ) .buravov , l.i . ; kushch , n.d . ;merzhanov , v.a . ; osherov , m.v . ; khomenko , a.g . ; yagubskii , e.b . journal de physique i * 2 * 1257 ( 1992 ) .p. m. chaikin , r. l. greene , s. etemad and e. engler , phys .b * 13 * , 1627 ( 1976 ) , p. m. chaikin , j. f. kwak and a. j. epstein , phys .lett . , * 42 * , 1178 ( 1979 ) . k. bender , d. schweitzer , and h.j .keller , journal de physique colloque c3 , * 44 * 1433 - 1436 ( 1983 ) .logvenov , m.v .karsovnik , h. ito , and t. ishiguro , synt . met . *86 * , 2023 - 2024 ( 1997 ) .bondarenko , s. kagoshima , m. maesato , t. hasegawa , n. miura , and t. yamaguchi , solid state commun . * 107 * 477 ( 1998 ) .yu , l. y. chiang , r. upasani and p. m. chaikin , phys .lett . * 65 * , 2458 ( 1992 ) .w. kang , s. t. hannahs , l. y. chiang , r. upasani and p. m. chaikin , phys .b * 45 * , 13566 ( 1992 ) .e. s. choi , j. s. brooks , j. s. qualls and y. s. song , rev .june ( 2001 ) ; los alamos preprint library : physics/0009054. s. v. demishev , m.v .kondrin , v.v .glushkov , n.e .sluchanko , and n.a .samarin , jetp * 86 * , 182 ( 1998)[zh.eksp.teor . fiz . *113 * , 323 ( 1998 ) ] .brooks , g. j. athas , s. j. klepper , x. chen , c. e. campos , s. valfells , y. tanaka , t. kinoshita , n. kinoshita , m. tokumoto and h. anzai , physica b * 201 * , 449 ( 1994 ) .a. kovalev , h. mueller and m. v. kartsovnik , jetp * 86 * , 578 ( 1998 ) .p. f. henning , j. s. brooks , j. e. crow , y. tanaka , t. kinoshita , n. kinoshita m. tokumoto and h. anzai , sol . stat .comm . * 95 * , 691 ( 1995 ) .n. biskup , j. a. a. j. perenboom , j. s. brooks and j. s. qualls , solid state commun .* 107 * , 503 ( 1998 ) .k. miyagawa , a. kawamoto and k. kanoda , phys .b 56 , r8487(1997 ) p. kuhns , j. s. brooks , t. caldwell , w. g. moulton , a. p. reyes , n. biskup , a. m. kini , j. a. schlueter , h. h. wang , u. geiser and j. m. williams , solid state commun .109 , 637 ( 1999 ) .a. kovalev , h. mueller , m.v .kartsovnik , jetp 113 , 1058(1998 ) .kartsovnik , w. biberacher , e. steep , p. christ , k. andres , a.g.m jansen , and h. mueller , synthetic metals , 86 1933 ( 1997 ) .r. mckenzie , g. j. athas , j. s. brooks , r. g. clark , a. s. dzurak , r. newbury , r. p. starrett , a. skougarevsky , m. tokumoto , n. kinoshita , t. kinoshita and y. tanaka , phys .b * 54 * , r8289 ( 1996 ) .t. sasaki , s. endo and n. toyota , phys .b * 48 * , 1928 ( 1993 ) . v. n. topnikov , s. i. pesotskii and v. n. laukhin , jetp lett . *59 * , 374 ( 1994 ) .i. m. lifshitz and a. m. kosevich , soviet physics jetp * 2 * , 636 ( 1957 ) . g. e. zilberman , soviet physics , jetp * 2 * , 650 ( 1956 ) . j. r. long , c. g. grenier and j. m. reynolds , phys .rev . * 140 * , a187 ( 1965 ) .j. wosnitza , _ fermi surfaces of low - dimensional organic metals and superconductors _ , springer verlag , berlin - heidelberg 1996 .s. uji , t. terashima , h. aoki , j. s. brooks , m. tokumoto , n. kinoshita , t. kinoshita , y. tanaka and h. anzai , phys .b * 54 * , 9332 ( 1996 ) .cccc & closed&closed&open along the + & compensated&uncompensated&@xmath115direction + + @xmath128&@xmath129&@xmath130&@xmath121 + @xmath131&@xmath130&@xmath129&@xmath129 + @xmath132&@xmath130&@xmath121&@xmath130 + @xmath133&@xmath130&@xmath121&@xmath121 + @xmath134&@xmath129&@xmath129&@xmath121 + + @xmath54&@xmath130&@xmath129&@xmath129 + @xmath53&@xmath130&@xmath130&@xmath121 + @xmath135&@xmath129&@xmath129&@xmath129 +

<S> we have used a low - frequency magneto - thermopower ( mtep ) method to probe the high magnetic field ground state behavior of @xmath0-(bedt - ttf)@xmath1khg(scn)@xmath2 along all three principal crystallographic axes at low temperatures . </S><S> the thermopower tensor coefficients ( @xmath3 and @xmath4 ) have been measured to 30 t , beyond the anomalous low temperature , field - induced transition at 22.5 t. we find a significant anisotropy in the mtep signal , and also observe large quantum oscillations associated with the de haas - van alphen effect . </S><S> the anisotropy indicates that the ground state properties are clearly driven by mechanisms that occur along specific directions for the in - plane electronic structure . </S><S> both transverse and longitudinal magnetothermopower show asymptotic behavior in field , which can be explained in terms of magnetic breakdown of compensated closed orbits . </S>

nucl-ex0702035

the study of heavy flavors in relativistic nuclear collisions follows two different approaches : ( i ) the direct reconstruction of heavy flavored mesons and ( ii ) the study of semi - leptonic decays of such mesons . in this sectionwe describe these methods , used by star and phenix , and discuss their advantages and limitations .direct reconstruction of heavy - flavor mesons is being performed by the star collaboration using the decay channel @xmath0 ( and c.c . ) with branching ratio of 3.83% in _ _d__+au and au+aucollisions . because of the small branching ratio and the lack of dedicated detector triggers , the direct reconstruction of @xmath1-mesons requires the analysis of large amount of data .the available statistics limits the study of such mesons to the low- region ( @xmath2 gev / c ) .kaons and pions are identified using the star time projection chamber ( tpc ) @xmath3 .the resulting invariant mass spectrum of kaon - pion pairs contains a substantial amount of background from random combinatorics that can be subtracted using mixed event methods .details of the analysis can be found in ref @xcite .figure [ fig_reco]-a shows an invariant mass distribution after event mixing subtraction where a clear @xmath4 peak can be seen .despite the limitation on and statistics , direct reconstruction of @xmath1-mesons is the cleanest probe to investigate heavy quarks in relativistic nuclear collisions . direct reconstruction of @xmath1-mesons can be done only by star in a limited range and still difficult to perform with high efficiency in high multiplicity events . in this casesemi - leptonic decays of such mesons ( ex : @xmath5 ) provide a more efficient measurement of charm and bottom production . at rhic two methodsare utilized to measure heavy flavor production via semileptonic decays : ( i ) identification of electrons from @xmath1 and @xmath6-meson decays and ( ii ) identification of muons from @xmath1-meson decays .the analysis of non - photonic electrons ( the excess of electrons after subtracting all possible sources of background , such as photon conversions and dalitz decays ) is a technique used by star and phenix .electron identification in star is done using the @xmath3 and momentum ( @xmath7 ) information from the tpc and the time of flight ( tof ) data for low- ( @xmath8 gev / c ) electrons @xcite or energy ( @xmath9 ) and shower shape in the electromagnetic calorimeter ( emc ) for high- ( @xmath10 gev / c ) electrons @xcite ( see figure [ fig_reco]-b ) . electron identification in phenix @xcite is largely based on the ring imaging cherenkov detector ( rich ) in conjunction with a highly granular emc .the momentum is derived from drift and pad chambers .a major difficulty in the electron analysis for both experiments is the fact that there are many sources of electrons other than semi - leptonic decays of heavy - flavor mesons .the main sources of background are photon conversion in the detector material ( less significant in phenix due to the reduced amount of material when compared to star ) and @xmath11 and @xmath12 dalitz decays .other sources of background , such as @xmath13 , @xmath14 and @xmath15 decays are also taken into account , although their contribution is small compared to the sources mentioned above .these background sources are commonly called photonic electrons . background subtraction in the phenix experiment is performed by two different methods : ( i ) the converter method , in which a well defined amount of material is added to the detector to increase the number of background electrons . by comparing the electron spectra with and without the converter it is possible to measure directly a significant part of the background .( ii ) the cocktail method , with which phenix measures the spectra of the main background sources ( photons , @xmath11 and @xmath12 ) .both methods agree to each other very well .due to its the large acceptance and tracking efficiency , star can directly reconstruct a considerable fraction of the photonic background by performing invariant mass reconstruction of @xmath16 pairs with high efficiency . for photon conversion , @xmath11 and@xmath12 dalitz decays the invariant mass spectrum shows a peak near zero mass .other background sources are evaluated by simulations and account for a very small fraction of the total background in this case .muon identification at low- ( @xmath17 gev / c ) plays an important role because this range imposes a significant constrain on the measurement of charm cross section @xcite .this measurement is being done by star using @xmath3 and momentum information from the tpc and mass reconstruction from the tof detector .a clear muon signal can be reconstructed , as seen in figure [ fig_reco]-c .background from @xmath18 and @xmath19 decays can be subtracted by looking at the dca ( distance of closest approach ) between the muon and the primary vertex of the collision .the determination of the total charm cross section via non - photonic electrons requires precise measurements down to very low- .this is an experimental challenge : the lower the , the higher the amount of photonic electrons that contaminates the measurement , resulting in large systematic uncertainties . in order to overcomethis limitation star performs three independent measurements : direct reconstruction of d - mesons , @xmath20 and @xmath21 .the total charm cross section is obtained from a combined fit of these measurements @xcite as illustrated in figure [ fig_xs]-a .phenix , on the other hand , uses its @xmath21 measurements to extract total cross sections @xcite . to overcome the large background at low- phenix is improving analysis techniques in order to reduce systematic uncertainties due to photonic background .this allowed a decrease in the minimum that can be measured by phenix from @xmath22 in the earlier datasets to @xmath23 gev / c ( figure [ fig_xs]-b ) .[ t ] , @xmath20 and electron data used in a combined fit from star .( b ) electron data used by phenix . curvesare fonll prediction used by phenix to extrapolate the electron spectra to @xmath24.,scaledwidth=80.0% ] [ fig_xs ] figure [ figsummary]-a summarizes all charm cross section measurements from to central au+au collisions at @xmath25 gev at rhic @xcite .shown is the total cross section per binary collision .figure [ figsummary]-a makes evident a factor of @xmath26 between star and phenix , this difference being larger than the combined systematic uncertainties in the case of central au+au collisions .the dashed line in this figure depicts the average prediction from fonll calculations @xcite for charm production at rhic energies and the yellow band its uncertainty determined by independent variation of quark masses , renormalization and factorization scales .the dash - dotted lines correspond to the average values for both experiments . despite the differences between experimentsthe results , for each experiment , suggest that charm production follows a binary scaling from to central au+aucollisions .this is a strong indication that charm are predominantly produced in the early stages of the collision evolution and other processes , such as thermal production in the qgp , are not significant .figure [ figsummary]-b shows the ratio of star @xcite and phenix @xcite measurements to fonll calculations for high- electrons in collisions at rhic .the dash - dotted lines correspond to the ratio of the experimental to fonll cross section .this makes evident that the factor of @xmath26 discrepancy in the cross section extends up to large values .note that , despite this normalization discrepancy fonll describes the shape of the measured spectra well in both cases , suggesting that the differences between the measurements may be related to an experimental normalization effect .future rhic measurements will address this experimental discrepancy .star is planning a run without its inner tracking detectors ( svt and ssd ) in the next years thus reducing the amount of photon conversion and will address in detail some of the systematic uncertainties in the background removal of the electron measurements . added to that , both star and phenix are developing detector upgrades providing drastic improvements in secondary vertex reconstruction which will allow the use of displaced vertex techniques to direct measure @xmath1 and @xmath6 mesons with high precision and efficiency .putting aside the discrepancies in the absolute cross sections measurements between star and phenix we now investigate how heavy quarks interact with the qgp formed in au+au collisions at rhic . over the last few yearsrhic is providing interesting information on how the partons behave in such hot and dense medium .the study of flavor dependence of these interactions will further expand our knowledge about the properties of the nuclear matter under such extreme conditions .nuclear effects in non - photonic electron production are measured through the comparison of spectra from _ _d__+au and au+au collisions to the equivalent spectrum in : the relevant quantity is the ratio @xmath27 , where @xmath28 is the differential yield in au+au ( _ _ d__+au ) and @xmath29 the corresponding yield in collisions .@xmath30 is the nuclear overlap integral , derived from glauber calculations @xcite . in the absence of nuclear effects , such as shadowing , cronin effect , or gluon saturation , hard processes are expected to scale with the number of binary collisions and hence , ( ) = 1 .figure [ fig_raa]-a shows the average for high- non - photonic electrons as a function of for star and phenix data . despite the cross section discrepancies between star and phenix , results are consistent with each other . for non - photonic electronsshows an increasing suppression from peripheral to central au+au collisions , indicating an unexpectedly energy loss of heavy quarks in the medium .this suppression is similar to the one observed for light - quark hadrons , indicated by the shaded area in the figure .for the 5% most central collisions , non - photonic electron production for @xmath31 is suppressed by a factor @xmath32 .[ t ] as a function of for minimum bias collisions.,title="fig:",scaledwidth=36.0% ] as a function of for minimum bias collisions.,title="fig:",scaledwidth=36.0% ] as a function of for minimum bias collisions.,title="fig:",scaledwidth=36.0% ] the suppression of non - photonic electrons in central au+aucollisions can be , to some degree , explained in terms of radiative energy loss in the medium .figure [ fig_raa]-b also shows different theoretical predictions for the non - photonic electron suppression in central au+au collisions considering different energy loss mechanisms . curves ( i ) and ( ii ) correspond to the expected heavy quark suppression when the energy loss mechanism is induced gluon radiation , considering electrons from @xmath1 and @xmath6 mesons decays . curve ( i ) corresponds to the average value of dvgl radiative energy loss calculation in which the medium gluon density is @xmath33 @xcite . in this case , the radiative energy loss does not account for the observed suppression , although the uncertainties , both in the data and theory , are large . on the other hand , curve ( ii )@xcite shows a larger suppression than curve ( i ) . in this case , however , the sensitivity of to the time averaged transport coefficient @xmath34 becomes smaller as it increases .in fact , the variation in for @xmath35 ( gev@xmath36/fm ) @xmath37 at @xmath38 gev / c is only @xmath39 @xcite .this saturation of the suppression for high values of @xmath34 can be attributed to the highly opacity of the medium , biasing the particle production towards its surface .although gluon radiation is still expected to be a significant energy loss mechanism other processes may become important to describe the suppression observed in central au+au collisions . curve ( iii ) is an average prediction for electrons from heavy - quark mesons decays and includes both dvgl radiative and elastic energy loss , as well as jet path length fluctuations @xcite .in contrast with light quarks , the elastic energy loss is comparable to the radiative for heavy quarks @xcite and the effect on is significant . in fact , theoretical predictions for elastic rescattering of partons in the medium that can create resonant d and b meson states via quark coalescence @xcite in the medium can lead to a significant suppression at moderate as seen in curve ( iv ) . in this case, the amount of suppression depends on the resonance widths .other processes , such as in - medium fragmentation @xcite , shown in curve ( vi ) may also contribute to the observed suppression in central au+au collisions .it is clear that the full understanding of the energy loss mechanisms is a fundamental milestone for the characterization of the medium properties .heavy flavors provide an important tool to investigate these mechanisms .it has been argued that the matter created in heavy - ion collisions at rhic is sufficiently hot and dense that charm quarks might thermalize in the medium @xcite . the most promising method to studythis hypothesis is the measurement of charm elliptic flow .large elliptic flow for @xmath1-mesons would imply in a large number of interactions , enough to suggest thermalization .recent phenix results @xcite for non - photonic electron @xmath40 , shown in figure [ fig_raa]-c , suggest that charm quarks carry a substantial amount of @xmath40 .non - photonic electron @xmath40 at low- is compatible with rescattering+resonances calculations suggesting a reduction in thermalization times of heavy quarks in the medium , when compared to pqcd scattering @xcite . at high-we notice a tendency of dropping in the non - photonic electron @xmath40 , although the statistical uncertainties are very large .this could be an indication of increasing dominance of electrons from bottom productions since bottom is not assumed to show a pronounced @xmath40 . more precise data and the independent measurement of @xmath1-meson @xmath40 is fundamental to understand @xmath40 in this region .generally , all current models overpredict at high- .it is important to note that in all calculations charm quarks are substantially more quenched than bottom quarks . curve ( v ) in figure [ fig_raa]-b , which is based only on electrons from @xmath1 decays describes the data best .it is the dominance of electrons from @xmath6 decays for @xmath41 that pushes the predicted to higher values .all theoretical calculations use the relative contribution of @xmath1 and @xmath6 as predicted by pqcd calculations @xcite .the understanding of the observed suppression as well the observed @xmath40 of non - photonic electrons require independent measurements of @xmath1 and @xmath6 mesons at high- .@xmath42 @xmath43 correlations in collisions is a promising tool to measure this relative contribution . in general ,heavy flavors have harder fragmentation functions than gluons and light quarks , making the near side correlation ( @xmath44 ) more sensitive to the decay kinematics .figure [ figcorrel ] shows preliminary data from star on the relative contribution of charm and bottom to the non - photonic electron data .the relative contribution is obtained from the fit of the near side correlation with predictions for @xmath42 correlations for @xmath1 and @xmath6 mesons from pythia .experimental uncertainties contain the ones from photonic background removal ( dominant part ) and the choice of fit method .details of this analysis can be found in ref @xcite .the lines show fonll predictions for charm and bottom .the yellow band reflects the uncertainty in the calculation . within errors ,the experimental data is compatible with fonll calculations .this result , if confirmed , would imply that bottom may be more suppressed in central au+au collisions than predicted by current energy loss calculations . in order to further investigate the relative contribution of charm and bottom to non - photonic electrons one may look for @xmath45-meson correlations at @xmath46 in collisions . because of the decay channels of @xmath1 and @xmath6 mesons the study of the charge dependence ( @xmath47 and @xmath48 ) of this correlations can help on the separation of electrons from @xmath49 and @xmath50 pairs .rhic is providing challenging data on heavy flavor production and its interplay with the medium .recent results show that this interaction with the medium is stronger than what we expected resulting in a large energy loss in central au+au collisions . in order to take the next step it is imperative to independently measure nuclear modification factors and @xmath40 of @xmath1 and @xmath6 mesons. this will be possible in the near future when detector upgrades are available .detailed and systematic measurements are required to address existing discrepancies and are fundamental to our understanding of heavy flavor production at rhic .the author thanks the star and phenix collaborations for providing preliminary data used in this work .the author also thanks r. averbeck , m. djordjevic , j. dunlop , h. van hees , x. lin , r. rapp , t. ullrich , i. vitev , r. vogt and z. xu and y.f .zhang for valuable discussion on this subject over the last months .this work is partially supported by fapesp and cnpq of brazil .

<S> the study of heavy flavor production in relativistic heavy ion collisions is an extreme experimental challenge but provides important information on the properties of the quark - gluon plasma ( qgp ) created in au+au collisions at rhic . </S><S> heavy - quarks are believed to be produced in the initial stages of the collision , and are essential on the understanding of parton energy loss in the dense medium created in such environment . moreover , heavy - quarks can help to investigate fundamental properties of qcd in elementary p+p collisions . in this work we review recent results on heavy flavor production and their interaction with the hot and dense medium at rhic . </S><S> high energy nuclear collisions at the relativistic heavy ion collider ( rhic ) @xcite have opened a new domain in the exploration of strongly interacting matter at very high energy density . </S><S> high temperatures and densities are generated in central nuclear collisions , creating the conditions in which a phase of deconfined quarks and gluons exists ( qgp ) @xcite . </S><S> the measurements at rhic in conjunction with theoretical calculations suggest that a dense , equilibrated system has been generated in the collision with properties similar to that of an ideal hydrodynamic fluid . </S><S> the strong suppression phenomena observed for high - hadrons @xcite suggest that the system early in its evolution is extremely dense and dissipative . </S><S> heavy quark ( charm and bottom ) measurements further expand the knowledge about the matter produced in nuclear collisions at rhic . </S><S> because of their large masses , their production can be calculated by pertubative qcd ( pqcd ) @xcite . in particular , comparative measurements in , _ _ d__+au and au+au are sensitive to the initial state gluon densities in these systems @xcite . in au+aucollisions , </S><S> medium effects such as heavy quark energy loss can be studied through comparison of distributions of bottom and charm production with those observed for inclusive hadrons . </S><S> the suppression of small angle gluon radiation ( dead cone effect ) predicted for heavy quarks would decrease the amount of energy loss @xcite though gluon emission and , therefore , the suppression of heavy quark mesons at high- is expected to be smaller than that one observed for light quark hadrons at rhic . in this case , </S><S> smaller energy loss allows heavy quarks to probe deeper into the medium @xcite . </S><S> it also opens new possibilities to investigate other interaction mechanisms such as collisional energy loss @xcite and in medium fragmentation @xcite . </S><S> moreover , measuring open charm and bottom production at rhic provides essential reference data for studies of color screening via quarkonium suppression @xcite . </S>

0807.4909

at low energy , below 1.4 gev , the cross section is largely dominated by the non - resonant processes @xmath10 . however , in the region around @xmath2 , a contribution from the ozi and g - parity violating decay @xmath11 is expected .this strongly suppressed decay ( -5 , ) can be observed via interference with the non - resonant process , showing up as a dip in the total cross section dependence from .+ the cross section as a function of is parametrized as @xcite : @xmath12 where @xmath13 is the bare cross section for the non - resonant process , @xmath14 is the interference parameter ( i.e. the ratio between the @xmath9 decay and the non - resonant process amplitudes ) , while @xmath15 , @xmath16 and @xmath17 are the mass , the width and the inverse propagator of the @xmath9 meson respectively .the non - resonant cross section , in the energy range of interest , increases linearly with .a model independent parametrization is used to describe the non resonant part : @xmath18 . in this workwe study two different final states : and , corresponding to the @xmath19 decay in @xmath20 ad @xmath21 , respectively . from the analysis we have measured the z parameter that has been used to determine the @xmath11 branching ratio ( br ) . in the case ofwe expect contributions also from @xmath22 and @xmath23 intermediate states , being @xmath24 a scalar meson . in @xcitewe have shown that at @xmath25 the interference between @xmath23 and @xmath26 events , evaluated by fitting the - dalitz plot , is small .therefore a fit to the cross section interference pattern for final state will provide information about the @xmath27 process and the resonant decays @xmath11 and @xmath28 .moreover , combining the results we extract the ratio @xmath29 .the kloe experiment operates at @xcite , the frascati @xmath9-factory .da@xmath30ne is an @xmath8 collider running at a center of mass energy of @xmath31 mev , the mass of the @xmath9 meson .equal energy positron and electron beams collide at an angle of @xmath32 - 25 mrad , producing @xmath9 mesons nearly at rest .the kloe detector consists of a large cylindrical drift chamber , dc , surrounded by a lead - scintillating fiber electromagnetic calorimeter , emc . a superconducting coil around the emc provides a 0.52 t field .the drift chamber @xcite , 4 m in diameter and 3.3 m long , has 12,582 all - stereo tungsten sense wires and 37,746 aluminum field wires .the chamber shell is made of carbon fiber - epoxy composite and the gas used is a 90% helium , 10% isobutane mixture .these features maximize transparency to photons and reduce @xmath33 regeneration and multiple scattering .the position resolutions are @xmath34150 @xmath35 m and @xmath36 2 mm .the momentum resolution is @xmath37 .vertices are reconstructed with a spatial resolution of 3 mm .the calorimeter @xcite is divided into a barrel and two endcaps , for a total of 88 modules , and covers 98% of the solid angle .the modules are read out at both ends by photo - multipliers , both in amplitude and time .the readout granularity is ( 4.44.4 ) @xmath38 , for a total of 2440 cells arranged in five layers .the energy deposits are obtained from the signal amplitude while the arrival times and the particles positions are obtained from the time differences .cells close in time and space are grouped into calorimeter clusters .the cluster energy @xmath39 is the sum of the cell energies .the cluster time @xmath40 and position @xmath41 are energy - weighed averages .energy and time resolutions are @xmath42 and @xmath43 , respectively .the kloe trigger @xcite uses both calorimeter and chamber information . in this analysisthe events are selected by the calorimeter trigger , requiring two energy deposits with @xmath44 mev for the barrel and @xmath45 mev for the endcaps .a cosmic veto rejects events with at least two energy deposits above 30 mev in the outermost calorimeter layer .data are then analyzed by an event classification filter @xcite , which streams various categories of events in different output files .all available statistics collected at the @xmath9 peak during 20012002 data - taking periods , corresponding to 450 pb@xmath1 , have been analyzed . moreoverfour scan points ( at 1010 , 1018 , 1023 and 1030 mev ) of @xmath010 pb@xmath1 each and the off - peak run at 1000 mev of @xmath0100 pb@xmath1 acquired in 2005 - 2006 have been included in this analysis .the luminosity is measured with 0.5% absolute precision counting large angle bhabha scattering events @xcite .data taken at the @xmath9 peak are grouped in center of mass energy bins of 100 kev width . for all the other points , close values are grouped together and the average center of mass energy is evaluated by weighting with luminosity . in the following ,the visible cross section for a given final state ( _ _ j__=@xmath46 , ) is defined @xmath47 , where @xmath48 , @xmath49 and , are the number of signal events , the analysis efficiency and the luminosity , respectively .+ the visible cross section is related to the bare cross section through the radiator function h ( see sec . [par : fitresults ] ) : @xmath50 in the analysis , data are filtered by selecting events with the expected final state signature : two tracks with opposite curvature connected to a vertex inside a small cylindrical fiducial volume ( @xmath52 cm and @xmath53 cm ) around the interaction point ( ip ) and four neutral clusters in the prompt time window ( tw ) , defined as @xmath54 . to minimize contamination from machine background ,we require for the clusters a minimal energy of 10 mev and an angle with respect to the beam line in the interval @xmath55 . with this selection , at @xmath9 resonance ,the main background contributions come from @xmath56 and @xmath57 , with @xmath58 , which have the same final state signature .other two resonant background components ( @xmath59 with @xmath60 , and @xmath61 ) mimic the final state signature because of additional clusters due to accidental coincidence of machine background events and/or shower fragments ( cluster splitting ) . an additional non - resonant background contribution of the order of few percent from @xmath62 , dominated by the @xmath63 intermediate state , is also expected for all values @xcite . a global kinematic fit ( @xmath64 ) , imposing total four - momentum conservation and proper time of flight ( tof ) for photons coming from the charged vertex , improves both the signal / background separation and the determination of the photon energies .the resulting @xmath65 ( @xmath66 ) is used to select signal enriched ( @xmath67 ) , s@xmath68 , and background dominated ( @xmath69 ) , b@xmath68 , samples . in s@xmath68 samplethe overall contamination from resonant background at @xmath70 is about @xmath71 while becomes negligible outside @xmath9 resonance .the contribution of the @xmath63 background is about 4% for all values .the signal selection efficiency in the s@xmath68 sample is evaluated by monte carlo ( mc ) and corrected with data / mc ratios for tracking , vertexing and clustering .the resulting value @xmath72 is dominated by the selection requirements and shows a small dependence on , which is taken into account in the evaluation of @xmath73 .the signal counting is performed for each bin by fitting the recoil mass ( @xmath74 ) distribution for both , s@xmath68 and b@xmath68 samples with mc signal and resonant , non - resonant background shapes .the fit procedure is based on a likelihood function which takes into account both data and mc statistics . in fig .[ fig : datamc_wpc].a - d , data - mc comparisons for events in the most populated bin of are shown . in fig . [fig : datamc_wpc].e - f , the @xmath74 distribution is also shown for the two outermost center of mass energies .mev : @xmath65 of kinematic fit for s@xmath68 ( a ) and b@xmath68 ( b ) samples , angle between charged pions in the @xmath19 rest frame ( c ) and recoil mass ( d ) for s@xmath68 events .the last distribution is shown also at @xmath75 mev ( e ) and @xmath76 mev ( f ) . ]the results are summarized in tab .[ tab : counting ] , where the signal counting ( @xmath77 ) and the visible cross section ( @xmath73 ) are reported for all bins .errors on @xmath73 include statistics , background subtraction and a systematic error of 0.75% , dominated by the luminosity measurement ( 0.5% ) .the other systematics being due to cosmic ray veto , event counting and final state radiation .we have checked the stability of the result by repeating the whole analysis chain with a wide variation of the selection criteria on : ( a ) minimum cluster energy and angle in the tw definition ; ( b ) value of @xmath66 cut used to select s@xmath68 and b@xmath68 samples ; ( c ) distribution used for the b@xmath68 sample when performing signal counting ; ( d ) data / mc efficiency correction curves .the resulting @xmath73 values are then used to estimate the systematic error of the measurement .the fit to the visible cross section as a function of described in sec .[ par : fitresults ] is repeated for all of these variations .the quoted systematics uncertainty is calculated as the quadratic sum of rms s obtained from variations ( a)-(d ) .the selection for @xmath79 starts requiring five neutral clusters in the prompt time window with @xmath80 mev and polar angle @xmath81 .after a first kinematic fit ( fit1 , @xmath82 ) imposing total 4-momentum conservation and time of flight , photons are paired to s by minimizing a @xmath65 function , built using the invariant mass of the two @xmath83 pairs . a second kinematic fit ( fit2 , @xmath84 )imposes also the constraint on masses .events taken at @xmath85 mev : ( a ) normalized @xmath65 of the second kinematic fit after acceptance cuts ( fit2 ) , ( b ) normalized two - photon invariant mass after @xmath86 cut , ( c ) @xmath21 invariant mass after @xmath87 cut , ( d ) @xmath88 distribution after all analysis cuts ; @xmath89 distribution at @xmath75 mev ( e ) and @xmath76 mev ( f ) . ]the background is rejected by requiring @xmath90 and @xmath91 and @xmath92 are evaluated using the photon momenta from fit1 .after these cuts the remaining sample is dominated by @xmath93 and @xmath94 events .signal is then selected neglecting the interference between these two processes and cutting on the intermediate state mass . defining @xmath89 as the closest mass to @xmath95 of the two @xmath21 combinations ,only events satisfying the requirement @xmath96 mev are retained . the residual background contamination( @xmath97 at the @xmath9 peak ) comes predominantly from @xmath98 events , where two photons are lost or merged . in fig .[ fig : datamc_wpn].a - d , data - mc comparison for events in the most populated bin are shown .the @xmath99 variable is the minimum angle between the photon and the s in the di - pion rest frame .a good agreement is observed both after acceptance selection and after applying analysis cuts .the comparison for the @xmath89 distribution is also shown for the two outermost center of mass energies , where the background contribution is practically negligible ( fig .[ fig : datamc_wpn].e - f ) .the overall signal selection efficiency is evaluated by applying the whole analysis chain to signal mc events : @xmath100 , almost independent from .the value obtained for each bin , together with the corresponding integrated luminosity , has been applied to the signal counting to obtain the visible cross section ( @xmath101 ) .results are summarized in tab .[ tab : counting ] .errors on @xmath101 include statistics , background subtraction and a systematic error of 0.6% , dominated by the luminosity measurement ( 0.5% ) .the other systematic effects , related to cosmic ray veto and to the event classification filter , have been evaluated with downscaled data samples .the stability of the results has been checked by repeating the fit with : ( a ) a variation of the selection criteria on the tw definition and on @xmath86 and @xmath89 cuts ; ( b ) a rescaling of the background components according to the value obtained by fitting background - enriched distributions on data with the corresponding mc components ; ( c ) a subtraction to the visible cross section of the interference effect between @xmath23 and the signal .this interference contribution is extracted from a fit to the dalitz plot of the @xmath79 final state @xcite at the @xmath9 peak , which provides the parameters describing the two processes .these values are then used to evaluate for each the interference contribution , which is subtracted to the corresponding @xmath102 .the largest contribution is 1.4% ..signal counting , visible cross section and radiative correction for and events .the radiative correction values are calculated using for the bare cross section parameters in tab .[ tab : fitres ] . [ cols="^,>,^,^,>,^,^",options="header " , ] after removing common systematics on the luminosity ( 0.5% ) , from the two measurements we obtain : @xmath103 taking into account the phase space of the two decays @xcite , the ratio of the partial widths is : @xmath104 since these two final states account for @xmath105 of the @xmath19 total width , we use the @xmath29 ratio and the sum of rarer br s @xcite to obtain from a fit : @xmath106 with a correlation of -71% .comparison between our evaluation and the values in @xcite is shown in fig .[ fig : wbrfit ] .our result for @xmath107 is less than the pdg value by three standard deviations .it is in good agreement with the recent prediction in @xcite .decay channels .the black square is the kloe fit result , while the black dot is the constrained fit result in @xcite .the gray ellipses are the 68% c.l .the measured @xmath109 and @xmath110 parameters of the final state are related to the br(@xmath11 ) through the relation : @xmath111 where @xmath112 is the total cross section of the @xmath113 process and @xmath114 is the peak value of the bare cross section for the resonance . using the parameters obtained from the analysis , the @xmath115 measurement from kloe @xcite for the evaluation of @xmath114 , and our value for @xmath116 we extract : @xmath117the error is reduced by a factor of two with respect to the best previous measurement from the snd experiment @xcite , which is in agreement with our result .using a sample of 600 pb@xmath1 collected at center of mass energy between 1000 and 1030 mev , we have measured the cross section parameters for the two processes @xmath11 and @xmath11 , obtaining the ratio @xmath29 with an accuracy of 1.8% .this ratio , together with the unitarity relation and the br measurements on the other @xmath19 decay channels , substantially improves the accuracy on the dominant @xmath19 branching fractions giving a value for br(@xmath118 ) which is three standard deviations lower than the particle data group fit @xcite .moreover , the parameters describing the reaction around @xmath2 are used to extract the most precise br measurement of the ozi and g - parity violating decay @xmath11 .we thank the dafne team for their efforts in maintaining low background running conditions and their collaboration during all data - taking . we want to thank our technical staff : g.f.fortugno and f.sborzacchi for their dedicated work to ensure an efficient operation of the kloe computing center ; m.anelli for his continuous support to the gas system and the safety of the detector ; a.balla , m.gatta , g.corradi and g.papalino for the maintenance of the electronics ; m.santoni , g.paoluzzi and r.rosellini for the general support to the detector ; c.piscitelli for his help during major maintenance periods . this work was supported in part by eurodaphne , contract fmrx - ct98 - 0169 ; by the german federal ministry of education and research ( bmbf ) contract 06-ka-957 ; by the german research foundation ( dfg),emmy noether programme , contracts de839/1 - 4 ; and by the eu integrated infrastructure initiative hadronphysics project under contract number rii3-ct-2004 - 506078 .v. m. aulchenko , j. exp .phys . 90 ( 2000 ) 927 .kloe collaboration , f. ambrosino , eur .j. c 49 ( 2007 ) 473 .s. guiducci , proc . of the 2001 particle accelerator conference ( chicago ,illinois , usa ) , p. lucas s. webber eds . , 2001 , 353 .kloe collaboration , m. adinolfi , nucl .inst . and meth .a 488 ( 2002 ) 51 .kloe collaboration , m. adinolfi , nucl .inst . and meth .a 482 ( 2002 ) 363 .kloe collaboration , m. adinolfi , nucl .inst . and meth .a 492 ( 2002 ) 134 .kloe collaboration , f.ambrosino , nucl .inst . and meth .a 534 ( 2004 ) 403 .kloe collaboration , f.ambrosino , european physiscal journal c47 ( 2006 ) 589 .r. r. akhmetshin , phys .b 466 ( 1999 ) 392 .m. greco , phys .b 318 ( 1993 ) 635 .w. m. yao , j. phys .g 33 ( 2006 ) and 2007 partial update for 2008 edition ( _ http://pdg.web.cern.ch/pdg _ ) m. benayoun , european physiscal journal c55 ( 2008 ) 199 .kloe collaboration , f. ambrosino , phys .b 608 ( 2005 ) 199 .

<S> we have studied the cross section in the interval 1000 - 1030 mev using the and final states with a sample of @xmath0 600 pb@xmath1 collected with the kloe detector at . by fitting the observed interference pattern around @xmath2 for both final states , </S><S> we extract the ratio of the decay widths @xmath3 and derive the branching fractions @xmath4 , @xmath5 . </S><S> the parameters describing the reaction around @xmath2 are also used to extract the branching fraction for the ozi and g - parity violating @xmath6 decay : @xmath7 . , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , @xmath8 collisions , rare @xmath9 decays , vmd , ozi violation , isospin violation </S>

astro-ph0105392

the problem of how spiral structures form in galaxies is one that has often been studied ( for a general overview of the subject , see @xcite ) , and can be divided into two aspects .the first is that of temporary structures most likely caused by gravitational perturbations from passing galaxies or an asymmetric halo , or disk material having an initial velocity relative to the local standard of rest from formation processes . herethe phenomenon seems to be related to density waves @xcite , quasi - stable modes in the gravitational potential of the disk which offer a good description of grand design spirals , where the arms are well - defined and has a high degree of symmetry .however , this paper is devoted to the second part of the problem , namely , that of spiral structures in `` isolated '' galaxies , i.e. where we ignore influences from outside the disk and consider only internal ( and recurring ) mechanisms .these galaxies can be represented by flocculent spirals , so called because of the fleecy appearance of their many short and asymmetric spiral arms . because these spirals are seen in blue light , but not red @xcite suggesting the arms are composed of younger , bluer stars , while the older , redder stars are more evenly distributed across the disk they must be an artifact of star formation itself , and are not primarily density waves .our attitude here is to consider structure formed by the processes of star formation and neglect gravitational influences .this should apply equally to all spirals , although for grand design spirals , we would expect it to act with density waves arising from other sources .most of the theoretical work done in star formation processes deals with the solar neighborhood ( those stars in the vicinity of our sun ) as opposed to structures on a galactic scale @xcite .other models used to study these aspects of star formation and spiral structure , such as those built on propagating star formation @xcite , either greatly simplify the physics involved or need finely tuned parameters to match observations . in this paper , we give a model of spiral structure in isolated galaxies based on the idea of star formation as part of a network of self - regulated and autocatalyzed reactions @xcite .recently , there has been a great deal of both theoretical and experimental work studying how non - equilibrium systems in chemistry and biology produce patterns in both time and space .this has included looking at both the organic such as bacterial colonies @xcite , the differentiation of cell types @xcite , and the formation of embryonic structure in multicellular organisms @xcite as well as the inorganic the bz chemical reaction @xcite , diffusion limited aggregation @xcite , and self - organized critical systems @xcite .there have been many successes in reproducing patterns in the laboratory , and these models typically share both partial differential equations and discrete elements such as cellular automata .however , most of this work in non - equilibrium systems has been after the main work on the formation of structures in astronomy .thus one might seek to apply this work to patterns seen in spiral structures . because we are studying the process in isolated galaxies , we know that the formation must be caused by events within the disk , rather than by the actions of outside players .the isolated galactic spiral is far from equilibrium there is differentiation of material into stars and clouds of gases whose distribution varies over space and time . in addition , star formation happens at a constant rate , as averaged across the disk .this is a clue that the process is regulated by a feedback loop to maintain this constancy ( for evidence of this mechanism , see @xcite ) .these characteristics are shared by other types of non - equilibrium systems .below , we list the predominant features that these kinds of networks of reactions have in common , along with examples of the same behavior in galactic disks . * * steady state system * there is a slow ( relative to the dynamical time scales ) and steady flow of energy , and perhaps matter running through the disk . in spiral galaxies ,star formation proceeds at a constant rate , averaged over the disk , for time scales on the order of 10@xmath0 years . the fact that this is greater than the time scales of the actual star formation process ( 10@xmath1 years ) implies that the slow and steady rate is regulated by feedback mechanisms . * * non - equilibrium system * the steady state is far from thermodynamic equilibrium , and there is a coexistence of several species or phases of matter , which exchange matter and energy among themselves through closed cycles .the galactic disk is not a uniformly dense clump of material at thermal equilibrium , but instead is divided into gases at different temperatures and stars of various mass .these species exchange matter : for example , massive stars supernova to form warm gas , which can cool and then condense into new stars . * * feedback mechanisms * the rates at which material flows around these cycles are governed by feedback loops that have arisen during the organization of the system into the steady state .an example of this is suggested by parravano and collaborators @xcite which explains how the average pressure in the interstellar medium ( ism ) is maintained .they argue that there are two phases , the warm gas of the ambient phase , and the cooler gas of the condensed phase , with a phase boundary in the pressure - temperature plane .ultraviolet radiation from the supernovae of massive stars heats the gas , which prevents the condensation of newer stars , so the supernova rate goes down , allowing new massive stars to form ( and so the supernova rate will increase again ) .this feedback mechanism keeps the gas on the phase boundary . ** autocatalytic reaction networks * any substances that serve as catalysts or repressors of reactions in the network are themselves produced by reactions inside the network .suppose we look at the condensation of giant molecular clouds ( gmcs ) .this is catalyzed by dust grains produced by cool giant stars , shielding the clouds and providing sites for molecular binding , and carbon and oxygen , which may cool the clouds by radiation from the rotational modes of co molecules .the condensation is inhibited by ultraviolet radiation from massive stars , as described by the parravano process mentioned above .* * separation in space * there may be spatial segregation of the different phases or materials in the cycles .this occurs when the inhibitory and catalytic influences propagate over different distance scales . at the smallest scale, this means the production of certain substances may be subject to refractory periods once production has occurred in a local region , it will not be repeated there for some period of time . for the influences in the process of gmc condensation , dust grains , carbon and oxygen propagate only over distances of about 100 parsecs ( how far supernovae and massive stars can spread their products ) while uv radiation can travel over much of the galactic disk .as can be seen , there is evidence that we can think of spiral structure in isolated galaxies to be a product of a self - organized , autocatalyzed network of reactions in the star formation process . given a system with the characteristics listed above, there are models which can describe the spatial structure , the most typical of which is the reaction - diffusion model @xcite , and we develop a model along these lines for spiral structure in isolated galaxies .we briefly outline the processes necessary to start star formation .the cold clouds of the gmcs condense out of the interstellar medium ( ism ) , forming distributions of gas and dust that are apparently scale - invariant . as mentioned above ,the condensation is helped along by the actions of dust , carbon and oxygen , while it can be impeded by uv radiation from massive stars .the parravano process may place some limitations on the amount of condensation .typical time scales for the inhibition are about @xmath2 years , the average life span of massive stars , after which the supernova ( sn ) rate and the uv radiation flux will die off . once the gmcs start to condense , then their cores may collapse to form stars .this collapsing is brought on by shockwaves from supernovae or hii regions ( we neglect any impact that density waves may have in these collapses ) , and so have the same length scale as the propagation of dust by supernovae , or about 100 parsecs .once the stars are formed , they can inhibit the infall of gas by the stellar wind or uv radiation produced by the star @xcite .these effects occur on short length scales , about the size of one cloud complex , and reduce the star forming efficiency of the clouds down to a few percent .we must take the processes occurring in the galactic disk and abstract them to produce a viable mathematical model .one way we can simplify the system is to take the continuous spectrum of star types and break it up to those massive stars that can supernova ( and thus provide matter and radiation back into the system ) and those lighter stars that can not .we will be neglecting the fact that these lighter stars can return matter to the ism , so in our model , they will simply act as a matter sink .there is also matter exchange between the cold gas of the gmcs and the warmer ambient gas , due to heating and cooling .we can summarize the flows of material in fig . 1 , andthe various material and energy components are given in table [ table1 ] .now we give a brief description of the physical processes we include in the model .* cloud destruction @xmath3 : because of the presence of massive stars and their stellar winds and sne produced shocks , there will be a mechanism of cloud distruption . some of this pressure will trigger star formation @xmath4 , but the eventual result will be the disruption of the cloud into warm gas @xmath5 . typically , the efficiency of star formation is around a few percent .* cloud - cloud collisions @xmath6 : another source of pressure is the collision of clouds , which will have the same types of effects as the star - induced cloud destruction mentioned above .again , the rate of star formation is a few percent . *mass infall to the disk @xmath7 : galaxies are believed to be formed from the condensation of matter from a spherical halo into a disk , and so there is certainly the possibility that there is a continuing flow of matter .it is believed that the rate is enough to replenish the material in the disk in a time span of billions of years ( see section 4.3 of larson @xcite ) .* direct cloud destruction by stars @xmath8 : this represents the effects of stars , such as stellar winds , which come directly from the massive stars , as opposed to radiation and shockwaves , which might travel some distance .the main physical action behind this term is the ionization and champagne flows of stars formed inside the cloud @xcite . *uv radiation , shockwave production @xmath9 : the sources of uv radiation and shockwaves are from sn events from massive stars .these effects are more long range , although shocks will travel only about 100 pc , while radiation can traverse the entire galaxy . * damping terms @xmath10 : because the energy carried by uv radiation and shockwaves will be dissipated by the interaction with matter both warm and cold we include a damping term .* cooling term @xmath11 : uv radiation will act as a thermostat , since warm gas is less likely to cool in an environment with a high radiation density . in the results presented here , we consider cooling which is inversely proportional to the radiation density . * cloud destruction by uv radiation and shockwaves @xmath12 : these effects are more long - range than the direct cloud destruction used previously .note that ultraviolet radiation will ionize the clouds into warm gas , as does the destruction induced by massive stars , but the pressure due to shockwaves will initiate new star formation .now , we take this model of the material flows in the galactic disk , and write them in a system of equations using the law of mass action that is , the change in time of the output quantity is given by the product of the densities of the inputs , with a constant parameter giving the rate of the reaction .thus , given the processes outlined above , we have [ eqns - one - zone ] @xmath13 where the parameters of the equations @xmath14 and their ranges are given in table [ table2 ] . to choose the parameters , we use the lifetime of a typical massive star , @xmath2 years , as the dimension of time , and the units of mass and energy to be those appropriate for each components for example , we choose the mass unit for warm gas @xmath15 to be one hydrogen atom per cubic centimeter .then each parameter is the rate of the reaction at the mass and energy densities we select , e.g. since the mass flow into the galaxy is estimated to completely replace the current material in @xmath16 years , we choose the mass inflow constant @xmath17 . for a fuller discussion of the choices made , see @xcite .finally , note that conversation of matter implies @xmath18 and @xmath19 . as mentioned above ,only massive stars are adding material to the ism , through the mechanism of supernovae we are ignoring the fact that light stars add material via evaporation .there are a few other simplifications which have been used to arrive at these equations , such as neglecting the impact of catalyzers to gmc condensation such as dust and carbon to avoid parameters that depend on metal concentrations . also , we noted previously that our cooling term is inversely proportional to the radiation density .this gives the simple result that cooling is faster when there is less radiation , but this is certainly not the only possibility ( although it is the one we have examined more closely ) .another choice would be to include something like a step function once the uv radiation is lower than a certain amount , the rate of cooling increases greatly , but above the cutoff , it is negligible .this idea of a critical density is essentially the process advocated by parravano and collaborators .however , it can be shown @xcite that there is little variation in the average values of the components as the functional dependence of the cooling is altered .in addition , the use of a step function in the cooling function can lead to oscillatory behavior in the components that is almost discontinuous and is therefore undesirable .the system of equations @xmath20 can be useful in understanding such things as the chemical evolution in a galaxy , but there already exists a substantial literature on one - zone models in galactic evolution .our goal here is to work with a model , similar to those in chemical and biological non - equilibrium systems , where spatial and temporal patterns are generated . in these systems , there are different length scales : l@xmath21 is the scale of the whole disk , the distance that uv radiation can travel ; l@xmath22 is the scale of distances between cloud complexes , and the distance core collapse is induced by supernovae ; and l@xmath23 is the scale of a single cloud , the distance new stars evaporate the gmc from which they condensed .note that the reactions characterized by l@xmath22 are catalytic , whereas those of l@xmath21 and l@xmath23 are inhibitive . to incorporate the effects of these length scales and the inhomogeneitiesthey produce , we introduce diffusion terms into the system of equations .although diffusion is clearly appropriate for such things as the movement of material such as stars and clouds through the disk , it might seem more realistic to use a wave equation to represent the effects of shockwaves and radiation .however , as is well - known , there are wavelike solutions to reaction - diffusion equations , such as the prototype fisher equation @xcite .also , we can see from the success the cellular automata of gerola , seiden and schulman @xcite and the model of elmegreen and thomasson @xcite that this is a valid approach . because the scale height of the milky way is much smaller than that of the radius of the disk, we consider the model on a two - dimensional annulus , leaving out the galactic bulge since it has little influence on the star formation in the disk .therefore the component functions we considered in sec .[ one - zone ] are now functions of the radius @xmath24 and the angle @xmath25 , in addition to time @xmath26 : @xmath27 note that the radius and angle , along with the angular velocity @xmath28 in the disk , will be the only greek letters that are not parameters of the model . being in a rotating system , with angular velocity @xmath28 , we must use the convective derivative , so that the evolution of the components is given by [ eqns - react - diff ] @xmath29 where the parameters have the same meaning as before , in the one - zone model .now that we have our model , we want to know if there are any structures of the right size that will arise . to do this , we consider a linearization of the model , and see if there are any instabilities .we assume that the galaxy is two - dimensional to simplify the analysis , and expand equations @xmath30 to linear order around the steady state ( i.e. @xmath31 , and similarly for the other functions ) .we assume instabilities with a wave vector @xmath32 growing with time scale @xmath33 , so that , for example , @xmath34 . note that we can find eigenfunctions of the differential operator made up of the laplacian and the convective derivative @xcite .then these equations give us a matrix equation of the form @xmath35 , where @xmath36 is a column vector made up of the component functions we can solve this matrix equation for the eigenvalues @xmath33 , and find which modes of instability are likely to grow exponentially as a function of the parameters . to carry out the linearization analysis, we pick an arbitrary set of parameters lying within the physical ranges , as described in sec .[ react - diff ] .if we look at the maximum real eigenvalue ( mre ) , we get a sense of which unstable modes will grow at the fastest rate . in figure 2, we see a graph of the mre of the matrix @xmath38 as a function of the logarithm of the wavenumber @xmath39 , using the values @xmath40 @xmath41 and the three diffusion constants @xmath42 since we are using @xmath43 parsecs ) as our length scale , then we note several things about the value of the mre .it is always negative all modes are stable and will eventually decay to equilibrium , although some decay slower .these modes can possibly allow the formation of some spiral structure in the disk , even if it is not a permanent pattern .note that the slowly decaying modes have pattern sizes of between @xmath44 to @xmath45 ( 1 to @xmath46 parsecs ) .unfortunately , this seems to be generic within the model , although there is the possibility that there exist parameter sets with positive mres that we have not found .once we had a set of parameters that , at least for some time , would form structures of the right size , the equations @xmath30 were numerically simulated using finite differencing , with the time evolution given by operator splitting .because the diffusion constant @xmath47 used in the linearized analysis is so much larger than @xmath48 and @xmath49 , it was decided to use the mean field approximation for the uv radiation : the radiation was spread instantaneously across the galactic disk between time steps , instead of diffusing . because the time increment used in the simulation was on the order of the light crossing time of the disk ( about @xmath50 years ) , this is not too unphysical a proposition .to take into account the effects of gravity , a constant linear velocity @xmath51 was given to all the material , approximating the situation in the milky way .this was done by using @xmath52 in the convective derivative .we present a picture of our results in figure 3 .the initial data is given on an annulus with @xmath53 parsecs ) and @xmath54 parsecs ) , and is just a gaussian `` blip '' for one component ( @xmath55 ) , at an arbitrarily choosen location .the disk is rotating with a constant linear speed @xmath56 , which is the velocity of the sun in the milky way .the boundary conditions on the inner and outer edge of the annulus is such that the radial derivatives of all the components are zero , i.e. there is no flux through the edges .this comes from the rationale that there are no clouds or stars infalling from the outside , and the central bulge does not interact much with the annulus .the run covered a time of about @xmath57 , a timespan of two billion years or approximately eight revolutions of the galaxy .as one can see from the figures , spiral structure indeed is formed .the spirals develop and grow in length as they rotate with the galactic disk .one could imagine with several such initial gaussian `` blips '' , several spiral arms would form , thus resembling a real , flocculant , spiral galaxy .however , after about @xmath58 or one billion years , the ends of a spiral arm , begin to interact ( diffuse ) with each other and eventually ( at about @xmath59 billion years ) merge forming a ring like structure .this appears to be a generic limitation of the model .several initial data and parameter sets were tried with the same results .it seems that only a modification of the model will be able to save the spirals from this eventual , unfortunate fate !work on this issue is currently underway and shall be presented elsewhere .even with this limitation , this model competes well with other current models . beyond this issue, there are a number of future directions that this work can be taken .this run only used one specific set of initial data ; one can hope that spiral formation does not depend strongly on initial conditions .we have made several runs with differing initial data , but we have not further explored this area .we are also lacking a detailed study of the effects of changes in the values of the parameters ; because of the large number of parameters and components , however , it is difficult to see how an exhaustive search for a more fruitful choice could be made .modifying the boundary conditions is another avenue to consider for example , allowing radiation to flow in from outside the galaxy , or to have interaction between the disk and the central bulge .also , we might alter the role of the uv radiation in the model .one consideration is that we have not given the radiation any type of radial fall - off as it leaves the source , but instead used a mean field approximation .this , unfortunately , heats the entire disk evenly .if , instead , there is a great deal of radiation in the space near the source , and little far away , this might give a more realistic feedback mechanism to inhibit star formation near the source but allow it further away .there are , of course , other ways that the model can be changed to make it more realistic , such as adding in more reaction mechanisms .the authors would like to thank lee smolin for introducing us to this subject , as well as many helpful discussions .also , we appreciate the help of christopher beetle with numerical methods , olaf dreyer with mathematica and sameer gupta with latex .most of this work was done when the authors were at the center for gravitational physics and geometry at the pennsylvania state university .h. gerola and p. e. seiden , astrophys .j. * 223 * , 129 ( 1978 ) ; h. gerola , p. e. seiden , and l. s. schulman , astrophys .j. * 232 * , 702 ( 1979 ) ; p. e. seiden and h. gerola , fundamentals of cosmic physics * 7 * , 241 ( 1982 ) ; l. s. schulman and p. e. seiden , science * 233 * , 425 ( 1986 ) ; l. s. schulman and p. e. seiden , adv .phys . * 39 * , 1 ( 1990 ) ; schulman , l. s. 1993 , in _ cellular automata : prospects in astrophysical applications _ , edited by a. lejeune and j. perdang ( world scientific , singapore , 1993 ) , p. 294a. burkert and g. hensler , astrophys .space sci . * 171 * , 149 ( 1990 ) ; d. p. cox and j. franco , astrophys .j. * 273 * , 243 ( 1983 ) ; j. franco and s. n. shore , astrophys . j.*285 * , 813 ( 1984 ) ; a. dopita , astrophys . j. * 295 * , l5 ( 1985 ) ; j. silk and r. f. g. wyse , astrophys . j. * 296 * , l1 ( 1985 ) ; c. f. mckee , astrophys .j. * 345 * , 782 ( 1989 ) a. parravano , astron .astrophys . * 205 * , 71 ( 1988 ) ; a. parravano , astrophys . j. * 347 * , 812 ( 1989 ) ; a. parravano , p. rosenzweig , and m. teran , astrophys .j. * 350 * , 100 ( 1990 ) ; j. mantilla ch . , and a. parravano , astrophys .j. * 250 * , 70 ( 1991 )

<S> the presence of spiral structure in isolated galaxies is a problem that has only been partially explained by theoretical models . because the rate and pattern of star formation in the disk must depend only on mechanisms internal to the disk </S><S> , we may think of the spiral galaxy as a self - regulated system far from equilibrium . </S><S> this paper uses this idea to look at a reaction - diffusion model for the formation of spiral structures in certain types of galaxies . in numerical runs of the model , spiral structure forms and persists over several revolutions of the disk , but eventually dies out . </S>

math0306439

many interesting examples of cyclic branched coverings of knots in @xmath3 admitting cyclic presentations for the fundamental groups have recently been found ( see @xcite ) . in order to investigate these relations , m. j. dunwoody introduced in @xcite a class of 3-manifolds , depending on six integer parameters , with cyclically presented fundamental groups .it has been shown in @xcite that all these manifolds turn out to be strongly - cyclic branched coverings of @xmath4-knots in lens spaces ( possibly @xmath3 ) . moreover , the explicit dunwoody representation for all cyclic branched coverings of 2-bridge knots has been obtained . in this paperwe give a similar result for a wide class of torus knots , which are , together with 2-bridge knots , the most important examples of @xmath4-knots in @xmath3 .the dunwoody parameters are obtained for all cyclic branched coverings of torus knots of type @xmath5 , with @xmath1 and @xmath2 , thus including all torus knots with bridge number @xmath6 .these manifolds have been considered , from a different point of view , in @xcite and @xcite .we refer to @xcite for details on knot theory and cyclic branched coverings of knots and to @xcite for details on cyclic presentation of groups .a knot @xmath7 in a 3-manifold @xmath8 is called a @xmath4-_knot _ if there exists a heegaard splitting of genus one @xmath9 where @xmath10 and @xmath11 are solid tori , @xmath12 and @xmath13 are properly embedded trivial arcs , and @xmath14 is an attaching homeomorphism ( see figure [ fig . 1 ] ) .obviously , @xmath8 turns out to be a lens space @xmath15 ( including @xmath16 ) .-decomposition . ]it is well known that the family of @xmath4-knots contains all torus knots and all two - bridge knots in @xmath3 .several topological properties of @xmath4-knots have recently been investigated ( see references in @xcite ) .an algebraic representation of @xmath4-knots has been developed in @xcite and @xcite , where it is shown that there is a natural surjective map @xmath17 from the pure mapping class group of the twice punctured torus @xmath18 to the class @xmath19 of all @xmath4-knots .the fundamental group of the exterior of a @xmath4-knot @xmath20 can be explicitly obtained using its representation @xmath21 .let @xmath22 be the loops depicted in figure [ fig . 0 ] .they represent a set of free generators for @xmath23 , while @xmath24 and @xmath25 freely generate @xmath26 .a straightforward application of seifert - van kampen theorem gives : [ fundamental1 ] the fundamental group of the exterior of a @xmath4-knot @xmath27 admits the presentation @xmath28 where @xmath29 is the homotopy class of @xmath30 , being @xmath31 the inclusion map . as a consequence ,all @xmath4-knots in @xmath3 are prime @xcite , and can therefore be classified , up to mirror image , by their fundamental groups ( see @xcite ) .many results concerning the connections between cyclically presented groups and strongly - cyclic branched coverings of @xmath4-knots ( see definition in @xcite ) have been obtained .it has been proved in @xcite that every @xmath32-fold strongly - cyclic branched covering of a @xmath4-knot admits a heegaard diagram of genus @xmath32 which encodes a cyclic presentation for the fundamental group .this result has been improved in @xcite , where a constructive algorithm which explicitly gives the cyclic presentations is obtained . the family of dunwoody manifolds has been introduced in @xcite by a class of trivalent regular planar graphs ( called _ dunwoody diagrams _ ) with cyclic symmetry , depending on six integers @xmath33 , such that @xmath34 , @xmath35 and @xmath36 . for certain values of the parameters , called _admissible _ , the dunwoody diagrams @xmath37 turn out to be heegaard diagrams , so defining a wide class of closed , orientable 3-manifolds @xmath38 with cyclically presented fundamental groups .more precisely , an admissible dunwoody diagram @xmath37 is an open heegaard diagram of genus @xmath32 ( see figure [ fig . 13 ] ) , which contains @xmath32 upper cycles @xmath39 , and @xmath32 lower cycles @xmath40 , each having @xmath41 vertices . for every @xmath42 , the cycle @xmath43 ( resp .@xmath44 ) is connected to the cycle @xmath45 ( resp .@xmath46 ) by @xmath47 parallel arcs , to the cycle @xmath48 by @xmath49 parallel arcs and to the cycle @xmath46 by @xmath50 parallel arcs .we denote by @xmath51 the set of all these arcs .the cycle @xmath43 is glued to the cycle @xmath52 ( subscripts mod @xmath32 ) so that equally labelled vertices are identified together .observe that the parameters @xmath53 and @xmath54 can be considered mod @xmath55 and @xmath32 respectively .since the identification rule and the diagram are invariant with respect to a cyclic action of order @xmath32 , the dunwoody manifolds admit a cyclic symmetry of order @xmath32 . obviously , the dunwoody manifold @xmath56 is homeomorphic to a lens space ( possibly @xmath3 ) , since it admits a heegaard splitting of genus one .a characterization of all dunwoody manifolds as strongly - cyclic branched coverings of @xmath4-knots is given by the following : [ strongly - cyclic ] the dunwoody manifold @xmath38 is the @xmath32-fold strongly - cyclic covering of the lens space @xmath56 ( possibly @xmath3 ) , branched over a @xmath4-knot @xmath57 only depending on the integers @xmath58 .the converse of this result is also true , as proved in @xcite .so the class of the dunwoody manifolds coincides with the class of the strongly - cyclic branched coverings of @xmath4-knots .the @xmath4-knots @xmath57 occurring in proposition [ strongly - cyclic ] admit a natural @xmath4-decomposition @xmath59 depicted in figure [ fig . 4 ] , where the arcs of @xmath51 constitute the curve @xmath60 , @xmath21 being the element of @xmath18 correspondent to @xmath61 .the fundamental group of the exterior of @xmath57 can be directly read in the dunwoody diagram of @xmath62 .the relation @xmath29 of the presentation of lemma [ fundamental1 ] is obtained by walking along the arcs of @xmath51 , following a fixed orientation : associate to each arc a word in @xmath24 and @xmath25 representing its homotopy class in the fundamental group of @xmath63 ( see figure [ fig .4 ] ) , where a properly embedded disk with boundary @xmath64 is considered squeezed to the base point @xmath65 . ]an interesting problem is to find the dunwoody parameters of the cyclic branched coverings of important classes of @xmath4-knots , in particular when the knot lies in @xmath3 .this type of result has been obtained in ( * ? ? ?* theorem 8) for all 2-bridge knots ._ let @xmath67 .figure [ fig .11 ] depicts the dunwoody diagram @xmath68 .the number of vertices of each cycle is @xmath69 .starting from the vertex of @xmath70 with label @xmath55 , we pass along all the arcs of @xmath51 with the described orientation , obtaining the word @xmath71 . therefore , since the conditions of ( * ? ? ?* corollary 4 ) are satisfied , @xmath72 is admissible .moreover , the fundamental group of @xmath73 is @xmath74 , which is trivial .so @xmath75 .moreover , @xmath76 . since @xmath77, we have @xmath78 , with @xmath79 .obviously this group is isomorphic to the group @xmath80 , which is the group of the torus knot @xmath81 . asa consequence @xmath7 is precisely @xmath81 .the proof is analogous to the previous one .let @xmath82 .figure [ fig .12 ] depicts the dunwoody diagram @xmath83 .the number of vertices of each cycle is @xmath84 ( note that the labelling is considered mod @xmath55 ) . starting from the vertex of @xmath70 with label @xmath55 , we pass along all the arcs of @xmath51 , obtaining the word @xmath85 .so @xmath72 is admissible .moreover , @xmath86 is homeomorphic to @xmath3 , since its fundamental group is trivial . since @xmath87 , the group @xmath88 is isomorphic to the group @xmath89 .therefore , @xmath7 is the torus knot @xmath90 . from proposition[ strongly - cyclic ] , we only need to find the sixth parameter @xmath54 of the 6-tuples .the relation @xmath91 mod @xmath32 must be satisfied for all @xmath32 , where @xmath92 is the number of arcs of @xmath51 oriented from @xmath93 to @xmath70 minus the number of the ones oriented from the @xmath70 to @xmath93 and @xmath94 is the number of arcs of @xmath51 oriented from the right to the left minus the number of the ones oriented from the left to the right in the dunwoody diagram ( see @xcite ) .in other words , @xmath95 is the total exponent of @xmath24 in @xmath96 and @xmath97 is the total exponent of @xmath25 in @xmath96 . in the first case @xmath98 and @xmath99 ; therefore @xmath100 . in the second case @xmath98 and @xmath101 ; therefore @xmath102 .an extension of the previous results to the whole class of torus knots by using the same proof technique seems to be very complicated , although some partial results have been obtained. a possible alternative method could be to work by induction on the number of steps of the euclidean division algorithm needed to obtain the greatest common division of the parameters of the torus knot . following this approach, the results of this article could represent the first inductive step .* acknowledgements .* work partially developed during a visit by the third named author to the department of mathematics of atatrk university , erzurum ( turkey ) , supported by tokten / unistar funds , and performed under the auspices of the g.n.s.a.g.a . of i.n.d.a.m .( italy ) and the university of bologna funds for selected research topics .

<S> we obtain an explicit representation , as dunwoody manifolds , of all cyclic branched coverings of torus knots of type @xmath0 , with @xmath1 and @xmath2 . </S><S> + + _ mathematics subject classification 2000 : _ primary 57m05 , 20f38 ; secondary 57m12 , 57m25 . </S><S> + _ keywords : _ torus knots , heegaard splittings , dunwoody manifolds . </S>

0905.2071

a complete description of complex polymers , such as proteins , includes information about their structure and their dynamics .in particular it is of utmost importance to answer the following questions : what are the structural conformations possible ?is there any relevant hierarchy among these conformers ?what are the transition paths between them ?these and other questions can be addressed by analyzing in an efficient way the free energy landscape of the system . with this knowledge, several problems about biomolecular reactions ( such as enzymatic activity , protein folding , protein deposition diseases , etc ) can be tackled . in this articlewe show how to efficiently describe the free energy landscape for small and large peptides . by mapping the trajectories of molecular dynamics simulations into a graph ( the conformational markov network ) , and unveiling its structural organization ,we obtain a coarse grained description of the protein dynamics across the free energy landscape in terms of the relevant kinetic magnitudes of the system .therefore , we show the way to bridge the gap between the microscopic dynamics to the macroscopic kinetics by means of a mesoscopic description of the associated conformational markov network . along this paththe compromise between the physical nature of the process and the magnitudes that characterize the network is carefully kept to assure the reliability of the results shown .polymers and , more specifically , proteins , show complex behavior at the cellular system level , _ e.g. _ in protein - protein interaction networks @xcite , and also at the individual level , where proteins show a large degree of multistability : a single protein can fold in different conformational states @xcite . as a complex system @xcite, the dynamics of a protein can not be understood by studying its parts in isolation , instead , the system must be analyzed as a whole .tools able to represent and handle the information of the entire picture of a complex system are thus necessary .complex network theory @xcite has proved to be a powerful tool used in seemingly different biologically - related fields such as the study of metabolic reactions , ecological and food webs , genetic regulatory systems and the study of protein dynamics @xcite . in this latter context, diverse studies have analyzed the conformational space of polymers and proteins making use of network representations @xcite , where nodes account of polymer conformations .additionally , some studies have tried to determine the common and general properties of these conformational networks @xcite looking at magnitudes such as clustering coefficient , cyclomatic number , connectivity , etc .recently , trying to decompose the network in modules corresponding to the free energy basins , the use of community algorithms over these conformational networks have been proposed @xcite .although this approach has opened a promising path for the analysis of free energy landscapes ( fel ) , the community based description of the network leads to multiple characterizations of the fel and thus it is difficult to establish a clear map from the communities found to the basins of the fel .a similar approach , commonly used to analyze the complex dynamics , is the construction of markovian models .markovian state models let us treat the information of one or several trajectories of molecular dynamics ( md ) as a set of conformations with certain transition probabilities among them @xcite .therefore , the time - continuous trajectory turns into a transition matrix , offering global observables as relaxation times and modes . in @xcite the use of markovian models is proposed with the aim of detecting fel meta - stable states .however , the above approaches to analyze fels of peptides involves extremely large computational cost : either general community algorithms or large transition matrices . finally , other strategies to characterize the fel that have successfully helped to understand the physics of biopolymers , are based on the study of the potential energy surface ( pes ) @xcite .the classical transition - state theory @xcite allows us to project the behavior of the system at certain temperature from the knowledge of the minima and transition states of the pes .this approach entails some feasible approximations , such as harmonic approximation to the pes , limit of high damping , assumption of high barriers , etc .these approximations could be avoided working directly from the md data . in this articlewe make a novel study of the fel capturing its mesoscopic structure and hence characterizing conformational states and the transitions between them .inspired by the approaches presented in @xcite and @xcite , we translate a dynamical trajectory obtained by md simulations into a conformational markov network .we show how to efficiently handle the graph to obtain , through its topology , the main features of the landscape : conformers and their basins of attraction , dwell times , rate constants between the conformational states detected and a coarse - grained picture of the fel .the framework is shown and validated analyzing a synthetic funnel - like potential .after this , the terminally blocked alanine peptide ( ace - ala - nme ) is studied unveiling the main characteristics of its fel .in this section we show the round way of the fel analysis : the map of microscopic data of a md into a conformational markov network ( cmn ) and , by unveiling its mesoscopic structure , the description of the fel structure in terms of macroscopic observables .first , we encode a trajectory of a stochastic md simulation into a network : the cmn .this map will allow us to use the tools introduced henceforth to analyze a specific dynamics of complex systems such as biopolymers .the cmn has been proven to be a useful representation of large stochastic trajectories @xcite .this coarse grained picture is usually constructed by discretizing the conformational space explored by the dynamical system and considering the hops between the different configurations as dictated by the md simulation . in this way , the nodes of a cmn are the subsets of configurations defined by the conformational space discretization and the links between nodes account for the observed transitions between them .the information of the stochastic trajectory allows to assign probabilities for the occupation of a node and for the transitions between two different configurations .defined as above , a cmn is thus a weighted and directed graph .our cmn is constructed as follows .the conformational space is divided into @xmath0 cells of equal volume , therefore every node @xmath1 ( @xmath2 ) of the cmn contains the same number of possible configurations .next , by evolving a stochastic trajectory enough time steps ( of length @xmath3 ) to assure the ergodicity condition we can define the final cmn set up .we assign to each node a weight , @xmath4 , that accounts for the fraction of trajectory that the system has visited any of the configurations contained in node @xmath1 ( the following normalization @xmath5 holds ) .second , a value @xmath6 is assigned to each directional link , accounting for the number of hops from node @xmath7 to node @xmath1 .note that transitions between configurations contained in the same node are also considered by @xmath8 , _i.e. _ the network can also contain self - loops . finally , the weights of the outgoing links from a node @xmath7 , @xmath9 , are conveniently normalized so that @xmath10 .the cmn constructed in this way , is described by a single matrix @xmath11 and a vector whose components are the occupation probabilities @xmath12 .hence , the matrix @xmath13 is the transition probability matrix of the following markov chain , @xmath14 where @xmath15 is the instant probability distribution of the system at time @xmath16 .since the matrix @xmath13 is ergodic and time invariant , one can compute the stationary distribution associated to the markov chain , @xmath17 , that satisfies @xmath18 .the latter stationary distribution has to be identical to the computed weights of the network nodes , @xmath19 ( @xmath2 ) , provided the stochastic trajectory is long enough .moreover , the detailed balance condition , @xmath20 must hold thus relating the elements of matrix @xmath13 to the stationary probability distribution .therefore , the transition matrix @xmath13 appears to be the minimal descriptor of the stochastic trajectory and , as consequence , of the cmn .provided the md trajectory is long enough to consider the sample in equilibrium , the weight - distribution of nodes in the cmn will be the stationary solution of eq .( [ evolves ] ) and detailed balance condition ( [ balance ] ) will be fulfilled @xcite . however , this property is not enough to consider the model markovian : although the continuous trajectory will be produced using langevin dynamics ( and therefore inherently markovian in the phase space @xcite ) the discrete representation of the cmn and the integration of momenta defies the markovian character of our model @xcite .several methods are proposed in the literature to validate markov models @xcite . in order to obtain a reliable description ,specially about those magnitudes related to the time evolution of the system ( see subsection temporal hierarchy of basins ) , the time step @xmath3 must be large enough to avoid memory effects @xcite. a detailed check and discussion about the markovian character of the networks shown in this article can be found in the _ supporting information text ( si)_. to illustrate the cmn approach and the methods presented below , we introduce here a synthetic potential energy function , that serve us as a toy model where results can be easily interpreted .this potential energy is reminiscent of that funnel surfaces recurrently found when the fel of proteins are studied @xcite . in particular , we have considered a two - dimensional system where a particle moves in contact with a thermal reservoir and whose potential energy is given by , @xmath21 where we have set @xmath22 , @xmath23 , @xmath24 , @xmath25 , @xmath26 , @xmath27 and @xmath28 . as shown in figure [ funnel]a the above potential energy confines the movement of the stochastic trajectory inside a finite region of the conformational space .however , thermal fluctuations allow the particle to jump between several basins of attraction .a stochastic trajectory has been simulated using an overdamped langevin dynamics and the equations of motion have been integrated with a fourth order stochastic runge - kutta method @xcite .figure [ funnel]c shows the region of the conformational space visited by the particle .we have conveniently discretized the two - dimensional space into pixels of equal area and computed their corresponding occupation probabilities . thereby , with the transition probabilities between pixels , the trajectory is represented as the cmn shown in figure [ funnel]b .the question now is : can we recover the topology of the fel ( derived from eq .( [ eq : funnel ] ) ) from the cmn representation ? up to now , we have illustrated the conversion of molecular dynamics data into a graph ( the cmn ) .now , we show how to efficiently obtain the thermo - statistical data from the mesoscopic description of the cmn . inspired by the deterministic steepest descent algorithm to locate minima in a potential energy surface we propose a _stochastic steepest descent _ ( ssd )algorithm to define basins on the discretized fel . dealing with the nodes and links as we describe below, the proper structure of the cmn is unveiled to call the modules obtained conformational macro - states or basins .picking at random one node of the cmn , say @xmath29 , and an initial probability distribution @xmath30 ( @xmath2 ) , the markov process relaxes according to @xmath31 .the whole probability concentrated in node @xmath29 at time @xmath32 , in a single time step @xmath33 , evolves driving the maximum amount of probability down hill over the fel .the next node @xmath34 in the descendent pathway from @xmath29 is taken by following the link that carries maximum probability flux .focusing again all the probability in node @xmath34 , @xmath35 we continue the pathway from @xmath29 towards a local fel minimum by identifying the next node @xmath36 for which the probability current @xmath37 is maximal . iterating this operation for each node of the cmn, we obtain a set of disconnected descent pathways that help us to define the basins of attraction .we establish formally the above procedure assisted by a vector @xmath38 ( with @xmath2 ) that label the nodes : * we start by assigning @xmath39 .* select at random a node @xmath40 with @xmath41 ( _ i.e. _ , @xmath40 has not been labeled yet ) and start to write an auxiliary list @xmath42 of nodes adding @xmath40 as the first entry in this list . * search , within the neighbors of the node @xmath40 , a node @xmath43 so that @xmath44@xcite and check which of the following options is fulfilled : 1 . if @xmath45 and @xmath46 : add @xmath43 to the list @xmath42 and go again to _( iii ) _ taking @xmath43 in the place of @xmath40 .2 . if @xmath45 and @xmath47 then write the labels of all the nodes in the list @xmath42 as @xmath48 @xmath49 . the process continues going to step _ ( ii)_. 3 . if @xmath50 the link @xmath51 is removed from the graph .the process returns to step _ ( iii ) _ with the next exception : since this step has been iterated @xmath52 times for the same node @xmath40 ( being @xmath53 the number of coordinates discretized to construct the cmn ) , @xmath40 is stated as local minimum and @xmath54 . in this case @xmath55 for those nodes @xmath56 and the process comes back to step _( ii)_. the whole procedure ends when no nodes unlabeled remain in the cmn , @xmath57 @xmath58 . the restriction introduced in step ( _ iii_.c ) with the dimensionality@xmath53 avoids a transition from a local minimal energy configuration to any other node of the same basin or to a deeper local minimum of a different basin .when every node of the cmn has been visited , the conformational space is completely characterized and we have thus traced all the maximum descent pathways from any node to the local fel minima . finally , all those nodes with the same label @xmath59 belong to the same fel basin and therefore they are associated to the same conformational state . the result of the procedure is the partition of the cmn in a set of modules which correspond to basins of attraction of the discretized conformational space . to illustrate the basin decomposition of a cmn, the ssd algorithm has been applied to the funnel - like potential .the result is the detection of six basins in agreement with the number of local minima in its fel ( figure [ funnel]b and [ funnel]c ) . with the aim of studying biomolecules and systems with high degree of dimensionality , the way to detect these fel basinsmust be computationally efficient .the method described above takes a computational time @xmath60 , once the @xmath52 largest hooping probabilities @xmath61 are computed for all the nodes in the network .additionally , the method is deterministic providing with a unique partition of the cmn into different modules .these two characteristics make this analysis faster and more straightforward than any other partitioning method @xcite .these advantages come from the knowledge of the physical meaning of links and nodes of cmns . in the _si _ other community algorithms ( newman s modularity and markov clustering algorithm ) are tested over our toy model system .none of the algorithms reported in the _ si _ give a satisfactory result mapping the modules obtained with the free energy basins . to get a more comprehensible representation of the fel studied , a new cmn network can be built by taking the basins as nodes .the occupation probabilities of these nodes as well as the transition probabilities among them can be obtained from those of the original cmn as @xmath62 @xmath63 where @xmath1 and @xmath7 are indexes relative to the nodes of the original cmn and @xmath64 and @xmath65 are indexes for the basins ( new nodes ) .note that the new coarse - grained cmn has its weights normalized and fulfills the detailed balance condition eq .( [ balance ] ) .figure [ funnel]d shows the corresponding coarse - grained for the funnel - like potential .the weighted nodes and links have a clear physical meaning @xcite .considering the transition state @xmath66 and assuming local `` intra - basin '' equilibrium , the rate constant of this transition is @xmath67 ( where @xmath33 is the time interval between snapshots used to make the original cmn ) .the relative free energy of the basin @xmath64 , taking basin @xmath65 as reference , is @xmath68 . besides , the expected waiting time to escape from @xmath64 to any adjacent basin is @xmath69 @xcite .other magnitudes , such as first - passage time for inter - basins transitions and other rate constants relaxing the local equilibrium condition @xcite can also be computed from the original cmn . the ability to define the proper regions of the conformational space in an efficient way let us compute physical magnitudes of relevance .for instance , the coarse - grained cmn is nothing but a graphical representation of a kinetic model with @xmath70 ( the number of basins ) coupled differential equations : @xmath71 ) .the original basins merge in function of time .both hierarchies reveals a coarse - grained behavior of two macro - states : @xmath72 and @xmath73.,scaledwidth=60.0% ] the first hierarchy aims to answer the following question : what is the structure of the cmn when nodes with lower weight than a certain threshold are removed together with their links ?let us take the control parameter @xmath74 as the threshold to restrict the existence of the nodes in the original cmn .where @xmath75 is the _`` adimensional free energy '' _ of a node @xmath1 relative to the most weighted node @xmath76 . with the above definitions we start a cmn reconstruction by smoothly increasing the threshold from its zero valueat each step of this process , we obtain a network composed of those nodes with free energy lower than the current threshold value . as the free - energy cut - off increases , new nodes emerge together with their links .these new nodes may be attached to any of the nodes already present in the network or they can emerge as a disconnected component . at a certain value of @xmath74 , some components of the network become connected by the links of a new node incorporated at this step .a hoshen - kopelman like algorithm @xcite is used to detect the disconnected components of the network at each value of the threshold used : from zero until all the @xmath0 nodes of the cmn were already attached .this bottom - up network reconstruction provides us with a hierarchical emergence of nodes along with the way they join together .this picture can be better described by a process of basins emergence and linking that is easily represented by means of a basin dendogram .this representation let us guess at first glance the hierarchical relationship of the conformational macro - states and the height of the barriers between them .let us remark that the transition times can not be deduced from these qualitative barriers since the entropic contribution or the volume of the basin are not reflected in this diagram .the basins family - tree obtained for the funnel - like ( see figure [ dendo]a ) reveals that , despite of having a two - dimensional potential with the shape of a funnel , one can not describe it as a sequence of metastable conformations that drive the system to the global minimum .moreover , the diagram shows a roughly similar behavior as for a double asymmetric well , composed by two sets of basins : @xmath77 and @xmath73 .the cmn representation of a md simulation provides with another hierarchical relationship that is meaningful to understand the behavior of the biological systems .the links of the original cmn have been weighted according to the stochastic matrix @xmath11 .taking into account the markovian character of the process , we can make use of the chapman - kolmogorov equation to generate new transition matrices at times @xmath78 , @xmath79 , etc ... formally , the markov chain at sampling time @xmath80 is defined by the matrix : @xmath81^{m}\;.\ ] ] for each value of @xmath43 a new cmn is defined .this family of cmns have different weighted links but the same weights @xmath4 for the nodes as the original one ( @xmath82 ) .it is worth to discuss the behavior of the matrix @xmath83 . in the limit @xmath84we have @xmath85 independently of the initial state @xmath86 .this means that any node is connected to a given node of the network with the same weight , regardless of the initial source .therefore , only one basin would be detected by the ssd algorithm since every node is connected with the most weighted link to the most weighted node in only one step .from the original @xmath33-description of the fel to the integrated ( @xmath87 ) one , we can devise another algorithm to establish a second hierarchy of the basins by performing the next two operations : first , for each value of @xmath43 a new cmn is generated by constructing the matrix @xmath88^{m}$ ] and second , the ssd algorithm is applied to this new cmn .the process finishes when only a basin ( the whole network ) is detected ( for large enough values of @xmath43 ) . by using this techniquewe can observe how basins merge with others at different time scales ( labeled by the integer @xmath43 ) .the result of this procedure performed for the funnel - like potential is shown in figure [ dendo]b . at @xmath89 only two basinsare observed : @xmath90 and @xmath73 , being the largest plateau observed for any of the nontrivial arrangement of basins found .therefore , the macroscopic description in time is in agreement with the free energy hierarchy described previously .it is clear that the number of basins decrease as @xmath43 increases. one should be aware that the concept of basin depends dramatically on the time resolution at which the cmn is built , and this time limits also the resolution in the fel structure .note also that this procedure provides with useful information similar to the _ structural decorrelation time _@xcite . and @xmath91 .( b ) plot of the cmn generated .the @xmath92 sets of nodes ( corresponding to different colors ) are the result of the ssd algorithm .( c ) left : ramachandran plot with the probability of occupation of the cells used to build the cmn .the boundaries of the free energy basins are shown in white .right : the @xmath92 basins represented as regions of different color .( color code : orange = @xmath93 , red = @xmath94 , yellow = @xmath95 , pink = @xmath96 , green = @xmath97 and blue = @xmath98).,scaledwidth=60.0% ]the alanine dipeptide , or terminally blocked alanine peptide ( ace - ala - nme , figure [ dipeptido]a ) , is the most simple " biological molecule that exhibits the common features shown by larger biomolecules . despite of its simplicity, this system has more than one long - life conformational state with different transition pathways . since the first attempt by rossky and karplus @xcite to model this dipeptide solvated ,this system has been widely studied in theoretical works @xcite .the alanine dipeptide has been also the appropriate molecule to test tools to explore the fel @xcite and , specifically , to study reaction coordinates @xcite .the alanine dipeptide has two slow degrees of freedom , the rotatable bonds @xmath99 ( c - n - c@xmath100-c ) and @xmath91 ( n - c@xmath100-c - n ) ( see figure [ dipeptido]a ) .the fel projected onto these dihedral angles let us identify the conformational states that characterize the geometry of biopolymers , namely : alpha helix right - handed ( @xmath93 ) , alpha helix left - handed ( @xmath94 ) , beta strands ( @xmath96,@xmath97 ) , etc .the number of local minima in the ( @xmath99 , @xmath91 ) space depends on the effective potential model used to simulate the system . up to date, electronic structure calculations have identified a total of nine different conformers @xcite . regarding md simulations different conformational stateshave been observed : _( i ) _ using classical force fields with explicit solvent up to six conformers are detected @xcite , _ ( ii ) _ at least four stable states by using implicit solvent @xcite , and _( iii ) _ two stable conformers in vacuum conditions @xcite . on the other hand , since the angles @xmath99 and @xmath91 seem appropriate to distinguish the metastable states , the kinetics between them is not accurately described with this choice of reaction coordinates , the solvent coordinates and/or other internal degrees of freedom must be taken into account @xcite .we have used _ ssd _ algorithm to detect the local minima and their corresponding basins for this molecule in the @xmath99-@xmath91 space . for this purpose ,a langevin md simulation of @xmath101 ns has been performed at a temperature of @xmath102 k ( see the _si _ for further details ) .additionally the cmn has been built dividing the @xmath103 ramachandran plot into cells of surface @xmath104 ( @xmath105 ) and taking dialanine conformations at time intervals of @xmath106 ps .the resulting cmn have a total of @xmath107 nodes and @xmath108 directed links .the @xmath109 algorithm applied to the cmn network reveals @xmath92 basins .figure [ dipeptido]b shows the resulting network where nodes belonging to the same basin take the same color . bringing back this information to the ramachandran map , these @xmath92 sets of nodes define @xmath92 regions represented in figure [ dipeptido]c . to better illustrate this division , other representation , where each region has a different color , is shown . by comparing with previous studies on this molecule , we identify the regions in orange , red , yellow and pink with conformers @xmath93 , @xmath94 , @xmath95 and @xmath96 respectively @xcite .besides , region green corresponds to conformer @xmath97 and the blue one to @xmath98 @xcite .remarkably , the basin @xmath94 ( one of the less populated state ) has been visited @xmath110 times with a mean stay time of @xmath111 ps .we now look at the coarse - grained picture of the fel by describing the properties of the @xmath92 basins detected .the different weights of the basins are related to the free energy of the corresponding conformational macro - states . in table[ mst ] these energy differences @xmath112 are shown , taking the most populated basin as the energy reference @xmath113 .the lowest free energy basins correspond to configurations with @xmath114 ( @xmath96 , @xmath93 , @xmath97 and @xmath98 ) , whereas the two other conformers located in the region @xmath115 have the highest free energy but the largest dwell time .moreover , we have also analyzed the trapping efficiency of each basin by reporting the mean escape time ( @xmath116 ) as well in table [ mst ] .chodera jd , singhal n , pande vs , dill ka , swope wc ( 2007 ) automatic discovery of metastable states for the construction of markov models of macromolecular conformational dynamics .j chem phys 126 : 155101 - 1155101 - 17 .

<S> the knowledge of the free energy landscape topology is the essential key to understand many biochemical processes . </S><S> the determination of the conformers of a protein and their basins of attraction takes a central role for studying molecular isomerization reactions . in this work </S><S> , we present a novel framework to unveil the features of a free energy landscape answering questions such as how many meta - stable conformers are , how the hierarchical relationship among them is , or what the structure and kinetics of the transition paths are . exploring the landscape by molecular dynamics simulations , </S><S> the microscopic data of the trajectory are encoded into a conformational markov network . </S><S> the structure of this graph reveals the regions of the conformational space corresponding to the basins of attraction . </S><S> in addition , handling the conformational markov network , relevant kinetic magnitudes as dwell times and rate constants , or hierarchical relationships among basins , complete the global picture of the landscape . </S><S> we show the power of the analysis studying a toy model of a funnel - like potential and computing efficiently the conformers of a short peptide , the dialanine , paving the way to a systematic study of the free energy landscape in large peptides . </S>

astro-ph9902244

it has been proposed that molecular outflows , at least from low and intermediate mass young stars , may be driven by highly collimated jets ( see , for example , padman , bence & richer @xcite ) .although the likely mechanism by which such jets transfer their momentum to the ambient medium remains unknown , a number of ideas have been put forward ( for a review of models the reader is referred to cabrit , raga & gueth @xcite ) . of thesethe most promising seems to be the so - called `` prompt entrainment '' mechanism . according to this model the bulk of the molecular outflowis accelerated ambient gas near the head of the jet or more precisely along the wings of its associated bow shock .observational support for prompt entrainment comes from the spatial coincidence of shocked molecular hydrogen bows with peaks in the co outflow emission ( e.g. davis & eislffel ) .smith , suttner & yorke ( @xcite ) and suttner et al .( @xcite ) have carried out a number of 3-d simulations of dense molecular jets propagating into a dense medium in order to test the prompt entrainment hypothesis .these authors found that their simulations reproduced many of the observational characteristics of molecular flows including the so - called ` hubble law ' ( see , e.g. padman et al .@xcite ) and strong forward , as opposed to sideways , motion ( lada & fich @xcite ) .while such results are encouraging for jet - driven models , it is still fair to say that no individual model has yet been able to plausibly account for all the observations ( lada & fich @xcite ) .moreover , alternatives to the jet model may be better at explaining the observational characteristics of some molecular flows ( e.g. padman et al .@xcite and cabrit et al .@xcite ) . the limited resolution of the 3-d jet simulations of smith et al . ( @xcite ) , and suttner et al .( @xcite ) , along with the high densities used by these authors , meant that they could not resolve the post - shock cooling regions in the flow .in addition it was not possible to explore parameter space as only a few such simulations could be performed .here we take a somewhat different approach by assuming low density atomic / molecular jet mixtures , a low density ambient medium and cylindrical symmetry .although such an approach obviously has it limitations , it does allow us to explore parameter space more fully and to resolve post - shock cooling regions ( this might be important , for example , if one is to gauge the importance of certain instabilities ) .the primary goal of this work is to investigate the efficiency of yso jets in accelerating ambient molecular gas _ without causing dissociation _ of its molecules .an additional question we address in this paper is whether velocity variations ( pulsing ) of the jet might enhance transfer of momentum from the jet to its surroundings and thus help to accelerate ambient gas .pulsing induces internal shocks which can squeeze jet gas sideways ( raga et al .this gas does not interact with the ambient medium directly , but is instead squirted into the cocoon of processed ( post - shock ) jet gas , which separates the jet from the `` shroud '' of post - shock ambient gas .chernin & masson ( ) however argue that , through the cocoon , momentum from the jet may be continuously coupled to the ambient flow .the properties of the simulated systems in which we are interested are as follows : * how much momentum is transferred to the ambient molecules ?* is there a power - law relationship predicted between mass in the molecular flow and velocity ?* what are the proper motions of the molecular ` knots ' , and how does their emission behave with time ?* is the so - called ` hubble law ' of molecular outflows reproduced under reasonable conditions ?* is there extra entrainment of ambient gas along the jet due to velocity variations? we will discuss each of these points in turn when presenting our results .our numerical model is presented in 2 and our results in 3 .conclusions from this work are presented in 4 and a simple way of overcoming the numerical problem of negative pressures while still maintaining overall energy conservation is given in the appendix .the equations solved are @xmath2 \label{mom}\\ { \frac{\partiale}{\partialt } } & = & - \vec{\nabla}\cdot\left[\left(e+p\right)\vec{u}\right]-l \label{energy } \\ { \frac{\partialn_{\rm h } x}{\partialt } } & = & -\vec{\nabla}\cdot\left[n_{\rm h } x \vec{u}\right]+ j(x , n_{\rm h},t ) \label{ifrac } \\ { \frac{\partialn_{\rm h_2}}{\partialt } } & = & - \vec{\nabla } \cdot ( n_{\rm h_2 } \vec{u } )- n_{\rm h_2}n_{\rm h}k(t ) \label{h2 } \\ { \frac{\partialn_{\rm h}}{\partialt } } & = & - \vec{\nabla } \cdot ( n_{\rm h } \vec{u } ) + 2n_{\rm h_2 } n_{\rm h } k(t ) \label{h } \\ { \frac{\partial\rho \tau}{\partialt } } & = & - \vec{\nabla } \cdot ( \rho \tau \vec{u } ) \label{tracer}\end{aligned}\ ] ] where @xmath3 , @xmath4 , @xmath5 , @xmath6 and @xmath7 are the mass density , velocity , pressure , total energy density and identity matrix respectively . @xmath8 and @xmath9 are the number densities of atomic and molecular hydrogen , @xmath10 is the ionization fraction of atomic hydrogen , @xmath11 is the temperature , @xmath12 is the ionization / recombination rate of atomic hydrogen , @xmath13 is the dissociation coefficient of molecular hydrogen , and @xmath14 is a passive scalar which is used to track the jet gas .we also have the definitions @xmath15 where @xmath16 is the specific heat at constant volume , @xmath17 is boltzmann s constant , @xmath18 is the ionization energy of hydrogen and @xmath19 is the dissociation energy of h@xmath20 .so @xmath21 is a function which denotes the energy loss and gain due to radiative and chemical processes .@xmath22 is the loss due to radiative transitions and is made up of a function for losses due to atomic transitions ( sutherland & dopita ) , and one for losses due to molecular transitions ( lepp & shull ) . the second term in @xmath21 is the energy dumped into ionization of h , and the third is that dumped into dissociation of h@xmath20 .the dissociation coefficient @xmath13 is obtained from dove & mandy ( ) and the ionization rate , @xmath23 , is that used by falle & raga ( ) .these equations are solved in a 2d cylindrically symmetric geometry using a temporally and spatially second order accurate muscl scheme ( van leer @xcite ; falle @xcite ) . the code uses a linear riemann solver except where the resolved pressure differs from either the left or right state at the cell interface by greater than 10% where it uses a non - linear solver ( following falle 1996 , private communication ) .non - linear riemann solvers allow correct treatment of shocks and rarefactions without artificial viscosity or entropy fixes . applying them only in non - smooth regions of the flowmeans that , while the benefits are the same , the computational overhead is minimised .this code is an updated version of that described in downes & ray ( ) .sometimes negative pressures are predicted by simulations involving radiative cooling .typically these are overcome by simply resetting the calculated pressure to an arbitrary , but small , positive value .however , this involves injecting internal energy into the system and this is undesirable . a fairly reliable way of overcoming this problem is discussed in appendix a. initially the ambient density and pressure on the grid are uniform and defined so that the ambient temperature on the grid is @xmath24 k. the jet temperature is set to @xmath25 k. the function @xmath21 is set to zero below this latter temperature as the data used in the cooling functions becomes unreliable and cooling below this temperature is not dynamically significant anyway .in most cases the ratio of jet density to ambient density ( @xmath26 ) is set to 1 ( see table [ list_sims ] ) . the ratio @xmath27 both inside and outside the jet , unless otherwise indicated ( again , see table [ list_sims ] ) . in all casesthe gas is assumed to be one of solar abundances .the boundary conditions are reflecting on @xmath28 ( i.e. the jet axis ) and on @xmath29 except where the jet enters , and gradient zero on every other boundary .the computational domain measures @xmath30 cells ( but larger in the @xmath31 simulations ) , with a spacing of @xmath32 cm .we find that the efficiency of momentum transfer is sensitive to the grid spacing .we performed a number of simulations with different spacings and concluded that this is the absolute minimum necessary to get reliable results .this length should be reduced with increasing density .incidentally this means that examining this property at the densities used by , for example , smith et al .( @xcite ) is impractical .the jet enters the grid at @xmath29 and @xmath33 and the boundary conditions are set to force inflow with the jet parameters .@xmath34 is set at @xmath35 cm or 50 grid cells .the jet velocity is given by @xmath36 with @xmath37 km s@xmath38 corresponding to a mach number of 65 and @xmath39 are chosen so that the corresponding periods are 5 , 10 , 20 and 50 yrs .here @xmath40 is effectively an amplitude for the velocity variations where present .the jet is initially given a small shear layer of about 5 cells ( @xmath41 cm ) in order to avoid numerical problems at the boundary between the jet and ambient medium .in this layer the velocity decays linearly to zero .nine simulations were run with varying values of density and velocity perturbation .these are listed in table [ list_sims ] which also gives the key we will use to refer to the simulations . in addition simulations of a purely atomic jet , and of a jet with a wide shear layer of almost 25 cells ( @xmath42 cm ) were run .the velocity of the jet with the wide shear layer is given by @xmath43 where @xmath44 is given by eq .[ jet_noshear ] .each system was simulated to an age of 300 yrs .although this is very young compared to the observed age of stellar jets , it was felt that the qualitative behaviour of the system at longer times could reliably be inferred from these results .the densities chosen are rather low to ensure adequate resolution of the system as described above .unfortunately this precludes the use of the data of mckee et al .( @xcite ) for calculations of the emissions from co as their calculations are only valid for gases of much higher densities . as a resultwe present our findings in terms of the mass of molecular gas rather than its luminosity . .[list_sims ] a list of the simulations performed in this work . unless otherwise stated @xmath45 [ cols="^,^,^,^",options="header " , ] we do find a power - law relationship between mass of molecular gas and velocity .if we write @xmath46 then we find that @xmath1 lies between 1.58 and 3.75 and that @xmath1 tends to increase with time , in agreement with smith et al .( @xcite ) .table [ fractions ] shows the values of @xmath1 at @xmath47 yrs for all the models .these values are consistent with observations ( e.g. davis et al .@xcite ) , and also with the analytical model presented in the appendix of smith et al . ( @xcite ) for the variations of mass with velocity . however , it is important to emphasise that what is actually observed is a variation in co line intensity with velocity .co line intensity is directly proportional to mass , in the relevant velocity channel , providing we are in the optically thin regime and the temperature of the gas is higher than the excitation temperature of the line ( see , e.g. mckee et al.@xcite ) .note that smith et al .( @xcite ) incorrectly state that the channel line brightness scales with @xmath48 .[ m - v ] shows a sample plot of the molecular mass versus velocity for the jet moving at an angle of @xmath49 to the plane of the sky .we do not see the jet contribution in the velocity range chosen here .the molecular fraction in the jet has a marked influence on the value of @xmath1 predicted by these models as we can see by comparing the results for simulations g and g2 .in fact , @xmath1 increases with decreasing molecular abundance in the jet .this is due to the reduction in strength of the high velocity jet component .it appears that @xmath1 does not depend in a systematic way on the amplitude of the velocity variations .note also that the introduction of a wide shear layer dramatically reduces @xmath1 .this is due to the fact that more gas is ejected out of the jet beam ( because of the more strongly paraboloid shape of the internal working surfaces ) and this accelerates the cocoon gas , leading to a stronger high velocity component .in addition , a wide shear layer causes the bow shock to be more blunt .it can be seen from the analytic model of smith et al.(@xcite ) that this also leads to a lower value of @xmath1 .it is interesting to speculate that lower values of gamma , which may be more common in molecular outflows from lower luminosity sources ( see davis et al .@xcite ) could result from such flows having a higher molecular fraction in their jets and perhaps a wide shear layer .the behaviour of @xmath1 with viewing angle is the same as that noted by smith et al .( @xcite ) .the actual values of @xmath1 obtained by these authors are somewhat lower than those obtained here . however , since our initial conditions are so different , and since @xmath1 is dependent on the shape of the bow shock , this discrepancy is not disturbing .we measured the apparent motion of the emission from the internal working surfaces .near the axis of the jet this emission moves with the average jet speed ( i.e. @xmath50 ) , as would be expected from momentum balance arguments .however , there are knots of emission arising from the bow shock itself and these move much more slowly ( @xmath51515% of the average jet speed ) with the faster moving knots being closer to the apex of the bow .this is in agreement with the observations of micono et al .( @xcite ) . fig . [ h2_em ] shows the emission from the s(1)10 line of h@xmath20 for model g. there is little emission from the cocoon since the cocoon gas has been strongly shocked in the jet shock and so is mostly atomic .we can also see that the emission becomes more intense as we move away from the apex of the bow shock , as reported by many authors ( e.g. eislffel et al .it is also clear that the internal working surfaces in the jet are giving rise to emission in this line .we can see that the emission begins to die away as we move away from the jet source .this is in agreement with observations of , for example , hh 46/47 ( eislffel@xcite ) where the emission from the knots appears close to the jet source and then fades away .this decrease in emission happens for two reasons .the first is that the shocks in the jet become weaker as they move away from the source simply because the velocity variations , which give rise to the shocks in the first place , are smoothed out by the shocks ( see , for example whitham ( @xcite ) .in addition , the mass flux through an individual shock decreases with time due to the divergent nature of the flow ahead of each internal working surface .this means that the emission will decrease because there is less gas being heated by the shock .we have found that the so - called ` hubble law ' ( e.g. lada & fich @xcite ) is reproduced in these simulations .[ hub_law ] shows a position velocity diagram calculated from simulation g assuming that the jet makes an angle of @xmath49 to the plane of the sky .this diagram is based on the mass of h@xmath20 rather than intensity of co emission .there is a gradual , virtually monotonic , rise in the maximum velocity .it is also worth noting that near the apex of the bow shock the rise in the maximum velocity present becomes steeper .these properties are related to the shape of the bow shock as gas near the apex of the shock is moving away from the jet axis at higher speed than that far from the apex . as a very basic model of this , suppose we represent the contact discontinuity between the post - shock jet and ambient gas to be an impermeable body moving with velocity @xmath52 through a fluid whose streamlines will follow the surface of the body . see fig.[schematic ] for a schematic diagram of the system .let this surface be described by the equation @xmath53 where @xmath54 and @xmath55 is the position of the apex of the bow shock on the @xmath56 axis ( see , e.g. , smith et al .since the contact discontinuity is a streamline of the flow we get that the ratio of the @xmath56-component to the @xmath57-component of the velocity is simply @xmath58^{\frac{s-1}{s}}\ ] ] note that this is the negative of the slope of the bow shock .this is because of our choice of the bow shock pointing to the right , and hence the @xmath56 component of the velocity will be negative . if we assume the post - shock velocity to be @xmath59 ( related to @xmath52 by the shock jump conditions ) , it is simple to show that @xmath60 finally , after some simple algebra , we can write down the velocity along the line of sight as a function of @xmath56 by @xmath61 where @xmath62 is the angle the bow shock makes to the plane of the sky .in fact we can write down @xmath59 if we assume that the bow shock is a strong shock everywhere , thus yielding a compression ratio of 4 ( from the shock conditions ) .it is easy to derive that @xmath63 this formula yields a shape for the position - velocity diagram which suggests the hubble - law and is similar to diagrams generated from these simulations assuming that the flow is in the plane of the sky .this indicates that the ` hubble - law ' effect is , at least partly , an artifact of the geometry of the bow shock . as is clear from fig .[ tau_mol ] there is not much extra entrainment of ambient gas resulting from the velocity variations .if the velocity variations were to involve the jet ` switching off ' for a time comparable to the sound crossing time of the cocoon then we would expect ambient gas to move toward the jet axis and probably be driven into the cocoon when the jet switches on again .this does not happen in these simulations where the maximum period of the variations is 50 yrs and the amplitude is at most 60% of the jet velocity .however , there is a small amount of acceleration of molecular gas at the left - hand boundary of the grid .this effect is quite small , but may grow over time .it is also possible , however , that this effect is due simply to the reflecting boundary conditions .these simulations show very little mixing between jet and ambient gas except very close to the apex of the bow shock .this means that the high velocity component of co outflows often observed along the main lobe axis ( bachiller @xcite ) is difficult to explain without invoking the presence of co gas in the jet beam itself just after collimation .it is now generally agreed that low velocity molecular outflows represent ambient gas that is somehow accelerated by highly collimated and partly ionized jets , at least in the case of low and intermediate mass ysos .most of this acceleration is thought to be achieved at the head of the jet through the leading bow shock ( the so - called prompt entrainment mechanism ) . in this paperwe have examined through many axially symmetric simulations the efficiency of the prompt entrainment mechanism as a means of transferring momentum to ambient molecular gas without causing dissociation .it is found , as one would expect , that the fraction of jet momentum transferred to the ambient environment depends on the jet / ambient density ratio .more importantly , we see that cooling , which is particularly important at higher densities , decreases the fractional jet momentum that goes into _ ambient molecules_. it would seem on the basis of the simulations presented here that both heavy and equal density ( with respect to the environment ) jets with radiative cooling have very low efficiencies at accelerating ambient molecules without causing dissociation . in partthis is because cooled jets have more aerodynamic bow shocks than the corresponding adiabatic ones ( i.e. they present a smaller cross sectional area to the ambient medium ) .the actual shape , however , of the bow shock also seems to to be important as the decrease in momentum transferred to the ambient medium seems to affect the acceleration of molecules more so than atoms / ions .we have also tested whether pulsed jets are more efficient at transferring momentum to the ambient medium than the corresponding steady jet with the same average velocity .somewhat surprisingly we found that even relatively large velocity variations do not give rise to significant changes in the amount of momentum being deposited in ambient gas .fundamentally this is because in high mach number jets , even with cooling , there is very little coupling between the jet s cocoon and the `` sheath '' ( i.e. the post - shock ambient gas ) .this lack of coupling is also the reason why turbulent entrainment is not significant in yso jets .our simulations were also used to model the expected variation of mass with velocity in molecular flows .we found relatively large @xmath1 values , i.e. mass should decline steeply with velocity , in line with observations .interestingly we found that a wide shear layer and an increasing molecular component in the jet _ reduced _ @xmath1 . if , as one might expect , such conditions are common among outflows from low luminosity ysos , this could explain their observed lower values for @xmath1 .finally we have shown that the so - called hubble law for molecular outflows is almost certainly a local effect in the vicinity of a bow shock .as noted in [ num_method ] it is common for conservative numerical codes to predict negative pressures under certain conditions .flows giving rise to such problems are usually highly supersonic , and diverging .the difficulties occur because the ratio of internal to total energy goes like @xmath64 , where @xmath65 is the mach number of the flow .therefore , in a highly supersonic flow , the fractional error required to predict a negative pressure is rather small .the introduction of energy losses exacerbates this problem .this is due to the fact that relatively weakly diverging flows , for example , can become supersonically diverging when the system is cooled because the sound speed is reduced .schemes which are second order accurate in space tend to produce more negative pressures than first order ones .this is because first order schemes dissipate strong features quickly so that strongly diverging flows rarely occur .it seems reasonable , then , to invoke a scheme which is first order in space whenever a negative pressure is produced as this will introduce extra dissipation .this extra dissipation may eliminate the negative pressure by moving some extra energy from neighbouring cells into the problem one . the scheme used by the code in this workcan be summarised as follows : @xmath66\ ] ] where @xmath67 and @xmath68 are the state vector and second order flux calculated at @xmath69 and @xmath70 respectively , and @xmath71 .if this scheme produces a negative pressure at @xmath72 then simply apply eq .[ scheme ] using the first order fluxes @xmath73 instead of the second order ones .note that it is necessary to adjust the neighbouring cells also so that overall conservation is maintained .while this fix does not work for all problems , it was found to be rather effective in the simulations presented here where very strong rarefactions are produced both at the edge of the jet at the @xmath29 boundary , and within the jet itself where the enforced velocity variations can cause problems .this fix can be implemented very easily by ensuring that the flux vectors around problem cells are stored .it has the advantage that it does not involve losing conservation of any of the physically conserved quantities .reducing the scheme to first order in certain regions of the grid is not a significant problem because this happens around shocks anyway in order to maintain monotonicity .we would like to thank c. davis , a. gibb and d. shepard for interesting discussions on the mass - velocity relationship in molecular outflows .we would also like to thank luke drury for his assistance with the development of the code and t. downes wishes to acknowledge the support of the nwo through the priority program in massively parallel computing .our simulations were carried out on the beowulf cluster at the dublin institute for advanced studies .finally we wish to thank the referee , michael smith for his very helpful comments .bachiller , r. , 1996 , ara&a , 34 , 111 cabrit , s. , raga , a. , gueth , f. 1997 , in _ herbig - haro outflows and the birth of low mass stars _ ,iau symposium no .182 , eds .b. reipurth & c. bertout , kluwer academic publishers , 163 chernin , l.m . , masson , c.r .1995 , apj 455 , 182 davis , c.j . ,eislffel , j. 1995 , a&a 300 , 851 davis , c.j . , moriarty - schieven , g. , eislffel , j. , hoare , m.g . ,ray , t.p ., 1998 , apj 115 , 1118 dgani , r. , van buren , d. , and noriega - crespo , a. 1996 , apj 461 , 927 dove , j.e . , mandy , m.e .1986 , apj 311 , l93 downes , t.p . 1996 , phd thesis ( university of dublin , ireland ) downes , t.p . , ray , t.p .1998 , a&a 331 , 1130 eislffel , j. , davis , c.j . , ray , t.p ., mundt , r. , 1994 , apj 422 , l91 falle , s.a.e.g .1991 , mnras 250 , 581 falle , s.a.e.g . , raga , a.c .1995 , mnras 272 , 785 lada , c.j . , fich , m. 1996 , apj 459 , 638 lepp , s. , shull , m.j .1983 , apj 270 , 578 mckee , c.f . , storey , j.w.v . ,watson , d.m . ,green , s. 1982 , a&a 113 , 285 micono , m. , davis , c.j . , ray , t.p ., eislffel , j. , shetrone , m.d . , 1998 ,apj , 494 , l227 padman , r. , bence , s. , richer , j. 1997 , in _ herbig - haro outflows and the birth of low mass stars _ , iau symposium no . 182 , eds . b. reipurth & c. bertout , kluwer academic publishers , 123 raga , a.c . , cant , j. , calvet , n. , rodrguez , l.f . , torrelles , j. 1993 , a&a 276 , 539 smith , m.d . , suttner , g. , yorke , h.w .1997 a&a 323 , 223 sutherland , r.s . , dopita , m.a .1993 , apjs 88 , 253 suttner , g. , smith , m.d . ,yorke , h.w . , zinnecker , h. 1997 , a&a 318 , 595 van leer , b. 1977 , j. comp .23 , 276 whitham , g.b .1974 , in _ linear and nonlinear waves _ , john wiley & sons , 50 .

<S> while it is generally thought that molecular outflows from young stellar objects ( ysos ) are accelerated by underlying stellar winds or highly collimated jets , the actual mechanism of acceleration remains uncertain . </S><S> the most favoured model , at least for low and intermediate mass stars , is that the molecules are accelerated at jet - driven bow shocks . </S><S> here we investigate , through high resolution numerical simulations , the efficiency of this mechanism in accelerating ambient molecular gas _ without causing dissociation_. the efficiency of the mechanism is found to be surprisingly low suggesting that more momentum may be present in the underlying jet than previously thought . </S><S> we also compare the momentum transferring efficiencies of pulsed versus steady jets . </S><S> we find that pulsed jets , and the corresponding steady jet with the same average velocity , transfer virtually the same momentum to the ambient gas . </S><S> the additional momentum ejected sideways from the jet beam in the case of the pulsed jet only serves to accelerate post - shock _ jet _ gas which forms a , largely atomic , sheath around the jet beam . for both the steady and pulsing jets , </S><S> we find a power law relationship between mass and velocity ( @xmath0 ) which is similar to what is observed . </S><S> we also find that increasing the molecular fraction in the jet decreases @xmath1 as one might expect . </S><S> we reproduce the so - called hubble law for molecular outflows and show that it is almost certainly a local effect in the presence of a bow shock </S><S> . finally , we present a simple way of overcoming the numerical problem of negative pressures while still maintaining overall conservation of energy . </S>

hep-ph9912231

it is widely recognized that the combined sources of evidence for neutrino flavor transitions coming from the solar neutrino problem @xcite and from the atmospheric neutrino anomaly@xcite demand an approach in terms of three - flavor oscillations among massive neutrinos @xmath7 @xcite .the three - flavor @xmath8 parameter space is then spanned by six variables : @xmath9 \ , \\\phi & = & \theta_{13}\in [ 0,\pi/2 ] \ , \\\psi & = & \theta_{23}\in [ 0,\pi/2 ] \ , \\\delta & = & { \rm cp\ violation\ phase } \ , \end{aligned}\ ] ] [ e1 ] where the @xmath10 rotations are conventionally ordered as for the quark mixing matrix @xcite . in the phenomenologically interesting limit @xmath11 , the two eigenstates closest in mass @xmath3 drive solar neutrino oscillations , while the `` lone '' eigenstate @xmath12 drives atmospheric neutrino oscillationsin such a limit ( see @xcite and refs . therein ) : @xmath13 the phase @xmath14 becomes unobservable ; @xmath15 the atmospheric parameter space is spanned by @xmath16 ; and @xmath17 the solar neutrino parameter space is spanned by @xmath18 ., the atmospheric and solar parameter spaces are decoupled into the two - family oscillation spaces @xmath19 and @xmath20 . ]building on a previous work @xcite , we perform a thorough analysis of the available solar neutrino data in the @xmath18 variables , in the context of the mikheyev - smirnov - wolfenstein ( msw ) oscillation mechanism @xcite .there are several motivations to revisit , update and improve the analysis in @xcite , the most important being the need to include the high - statistics super - kamiokande ( sk ) observations of neutrino events ( total rate , @xcite , energy spectrum @xcite , and day - night difference @xcite ) .indeed , the few post - sk papers on three - flavor msw oscillations we are aware of @xcite include neither the spectrum information nor day - night variations .this contrasts with the more familiar two - family msw analyses , which have been regularly updated with state - of - the - art global fits at different detector lifetime days ; see , e.g. , the sk official fits ( @xcite and refs . therein ) , as well as various analyses by independent research groups : @xcite ( 100 days ) , @xcite ( 300 days ) , @xcite ( 500 days ) , and @xcite ( 800 days ) .after the work @xcite , there have been also other relevant experimental and theoretical improvements ( implemented in the present analysis ) : updated measurements of the experimental ( total ) rates in the chlorine ( cl ) @xcite and kamiokande ( k ) @xcite experiments and in the gallium ( ga ) detectors sage @xcite and gallex @xcite ; new standard solar model ( ssm ) estimates for the neutrino fluxes and their uncertainties , and for the neutrino production regions in the sun @xcite ; updated calculations of the @xmath21b neutrino spectrum @xcite , of the @xmath22 scattering cross section on electrons @xcite , and of the @xmath23 neutrino absorption cross section in chlorine @xcite and in gallium @xcite ; analytical methods to compute the ( time - averaged ) @xmath23 survival probability in the earth @xcite .in addition , some relatively old topics , such as the role played by a possibly large _ hep _ neutrino flux @xcite , or by solar neutrino mixing at and beyond `` maximal '' values @xmath24 @xcite , are currently being revisited ( see @xcite and @xcite , respectively ) and demand an updated discussion . in particular , the case of maximal solar neutrino mixing @xmath25 appears now more `` natural '' and interesting in light of the super - kamiokande results on atmospheric neutrinos , which favor maximal @xmath26 mixing ( @xmath27 ) .msw oscillations at @xmath28 are nontrivial when earth matter effects are included @xcite .this fact was not considered in refs .@xcite , where only matter effects in the sun were studied . ]therefore , it is important to go beyond the usual analyses in terms of the solar @xmath8 mixing parameter @xmath29 ( equivalent to take @xmath30 ) , which miss a potentially interesting region of the msw solutions at @xmath31 , as we shall see later ( see also @xcite for earlier discussions ) .concerning the parameter @xmath32 , we do not include _ a priori _ the constraints coming from the chooz reactor experiment @xcite , which , together with the atmospheric neutrino data @xcite , imply that @xmath33 few % @xcite ( the exact upper limit depending on the value of @xmath34 , on the confidence level chosen , and on the number of degrees of freedom in the oscillation model ) .in fact , we think that it is instructive to study the constraints on @xmath32 coming from solar neutrino data _ alone _ , as it has been done similarly for atmospheric neutrino data @xcite . in any case ,observable effects on the msw solutions can be generated even by small ( few % ) values of @xmath35 , which induce a fractional suppression in the solar neutrino survival probability @xmath36 approximately equal to @xmath37 .in fact , if @xmath38 is the electron density profile , the @xmath39 and @xmath40 expressions for @xmath36 are related by the expression @xmath41 which , for small values of @xmath32 , gives roughly @xmath42 even within the stringent chooz limits on @xmath43 , such variations of @xmath36 can become as large as the uncertainties on the neutrino event rates , and thus can not be neglected .the above motivations warrant the present work , which is organized as follows . in sec .[ s2 ] we present the theoretical and experimental ingredients of the analysis . in sec .[ s3 ] we update the familiar two - family oscillation fit ( @xmath44 ) , and discuss in detail the features of the @xmath40 msw solutions . in sec .[ s4 ] we extend the analysis to three - flavor oscillations , and discuss how the @xmath39 msw solutions change for increasing values of @xmath32 .we summarize our results in sec .the details of the statistical analysis are given in the appendix .in the present analysis , we use the so - called bp98 standard solar model @xcite for the electron density in the sun and for the input neutrino parameters ( @xmath23 fluxes , spectra , and production regions ) , and compare the predictions to the experimental data for the following observables : total neutrino event rates , sk energy spectrum , and sk day - night asymmetry .table [ tablerates ] shows the latest results ( and standard expectations ) for the total neutrino event rates measured at homestake @xcite , kamiokande @xcite , sage @xcite , gallex @xcite , and super - kamiokande ( 825 live days ) @xcite .since the sage and gallex detectors measure exactly the same quantity ( up to a negligible difference in the detector latitude ) , their results are combined in a single ( ga ) rate of @xmath45 snu . on the other hand ,the kamiokande and super - kamiokande data are treated separately ( rather than combined in a single datum ) , since the two experiments , although based on the same @xmath8-@xmath46 scattering detection technique , have rather different energy thresholds and resolution functions .fit to the total rates for the @xmath39 msw case [ three data ( cl , ga , sk ) minus three free parameters @xmath47 .in such a situation , the value of @xmath48 would not have a well - defined likelihood . ]the sk electron recoil energy spectrum and its statistical and systematic uncertainties ( 825 lifetime day , @xmath49 mev ) are graphically reduced from the 18-bin histograms shown by sk members in recent summer 99 conferences @xcite .the corresponding numerical values have been already reported in detail in ref .@xcite and are not repeated here .our theoretical calculation of the binned spectrum properly takes into account energy threshold and resolution effects ( see , e.g. , the appendix of ref .standard @xmath21b @xcite and _ hep _ @xcite neutrino spectra and fluxes are used , unless otherwise noted . concerning the sk day - night asymmetry of the event rates, we use the latest measurement @xcite : @xmath50 in the presence of @xmath40 or @xmath39 oscillations , the msw effect in the sun is computed as described in ref .the additional earth matter effects are treated as in ref .the @xmath51 analysis of the theoretical and experimental uncertainties basically follows the approach developed in @xcite , with the necessary updates to take into account the bp98 ssm predictions and the energy spectrum information .technical details about error estimates are given in the appendix .we conclude this section by comparing the standard ( no oscillation ) predictions with the experimental data for the cl , ga , and sk total rates .figure [ f1 ] shows the 99% c.l .error ellipses for data and expectations in the planes charted by the ( cl , ga ) , ( sk , ga ) , and ( sk , cl ) total rates .the distance between observations and standard predictions makes the solar neutrino problem(s ) evident . at present, such information is the main evidence for solar neutrino physics beyond the standard electroweak model ; however , since the theoretical errors are dominant as far as total rates are concerned no substantial improvements can be expected by a reduction of the experimental errors .conversely , decisive information is expected from the sk spectrum and day - night asymmetry , but no convincing deviation has emerged from such data yet .therefore , it is not surprising that , in oscillation fits , the total rates mainly determine _ allowed _ regions , while the sk spectrum and day - night asymmetry determine _ excluded _ regions .figure [ f2 ] shows the results of our @xmath40 msw analysis of the data discussed in the previous section , shown as confidence level contours in the @xmath52 plane .the choice of the variable @xmath53 , rather than the usual @xmath29 , allows an expanded view of the large mixing region . in each of the six panels , we determine the absolute minimum of the @xmath51 and then plot the iso-@xmath51 contours at @xmath54 , 5.99 , and 9.21 , corresponding to 90% , 95% , and 99% c.l . for two degrees of freedom ( the oscillation parameters ) . in fits including the total rates ,there is a global @xmath51 minimum and two local mimima ; such minima , and the surrounding favored regions , are usually indicated as msw solutions at small mixing angle ( sma ) , large mixing angle ( lma ) , and low @xmath55 ( low ) .additional information on such solutions is reported in tab .[ chisquare ] .concerning the statistical interpretation of the @xmath51 values , a remark is in order .one can attach confidence levels to @xmath51 values in two different ways , depending on the choice between _ hypotheses tests _ and _ parameter estimation _ @xcite .if one is interested in testing the goodness of the msw hypothesis _ a priori _ , then one should compare the absolute @xmath48 with a number of degrees of freedom @xmath56 calculated as number of data minus number of free parameters .the corresponding probability @xmath57 is given in the last column of tab .[ chisquare ] where , for completeness , @xmath57 is reported also for the other two local minima and for the no oscillation case .if the msw hypothesis is accepted , then the msw parameter estimation involves only @xmath51 differences with respect to the global minimum , and the appropriate value of @xmath58 to use is the number of free parameters in the model ( @xmath59 ) , as anticipated for fig .[ f2 ] .the first panel of fig .[ f2 ] refers to the fit to the total rates only .the three @xmath51 minima are indicated by dots .the absolute minimum is reached within the sma solution @xmath60 , which represents a very good fit to the data .the lma solution is also acceptable , while the low solution gives a marginal fit ( see also the upper three rows of tab .[ chisquare ] ) .the sk data on the day - night asymmetry ( second panel ) and energy spectrum ( third panel ) exclude large regions in the mass - mixing parameter space ; but are unable to ( dis)prove any of the three solutions , which in fact are present also in the global fit to all data ( fifth panel ; see also the middle three rows of tab .[ chisquare ] ) .the spectrum information is somewhat sensitive to the ( uncertain ) value of the _ hep _ neutrino flux ; for instance , an enhancement by a factor @xmath61 helps to fit the high - energy part of the sk spectrum @xcite , and thus it produces a reduction of the excluded regions in the mass - mixing plane ( fourth panel in fig .[ f2 ] ) , and a corresponding slight enlargement of the globally allowed regions ( sixth panel ; see also the lower three rows of tab .[ chisquare ] ) . to the _ hep_/ssm flux ratio ( in the absence of oscillations ) .this limit is weakened to @xmath62 in the presence of an oscillation effect of @xmath63 .therefore , we can assume @xmath64 as a relevant , representative case of large _ hep _ flux . ] from the results of fig .[ f2 ] and of tab .[ chisquare ] , it appears that the inclusion of the day - night and spectral information can significantly change the c.l.s associated to each of three solutions sma , lma , and low . in order to understand better the role of difference pieces of data, we show in fig .[ f3 ] the comparison between data and predictions for the @xmath48 point of the total rate fit ( sma solution at best fit , first row of tab . [ chisquare ] ) , in the same coordinates as in fig .analogously , figs .[ f4 ] and [ f5 ] show the analogous comparison for the lma and low solutions ( second and third row of tab .[ chisquare ] ) .figure [ f6 ] shows the spectral information for the three local minima in the global fit to all data with standard _hep _ flux ( upper panel ) and with enhanced ( @xmath65 ) _ hep _ flux ( lower panel ) . from figs .[ f3][f5 ] it appears ( as far as total rates are concerned ) that the sma solution represents a very good fit to the sk , ga , and cl data , that the lma solution underestimates slightly the ga and sk data , and that the low solution underestimates the ga rate and overestimates the cl rate .this explains the ordering in the likelihood of such solutions in tab .[ chisquare ] : @xmath66 .on the other hand , figure [ f6 ] shows that the bulk of the sk observed spectrum is basically consistent with a flat shape ( except for the two highest - energy , low - statistics bins ) , and therefore tends to favor the lma and low solutions , rather than the steadily increasing spectrum predicted by the sma solution .this tendency is slightly enhanced in the case of large _ hep _ neutrino flux ( lower panel of fig .[ f6 ] ) since in this case the lma and low solutions can fit somewhat better the spectrum endpoint .in other words , the inclusion of the spectrum in the global analysis tends to compensate the different likelihoods of the lma and low solutions with respect to the sma solution .moreover , the slight excess of observed nighttime events [ eq . ( [ e4 ] ) ] adds extra likelihood to the lma solution , so that in the global fit one has @xmath67 ( see tab .[ chisquare ] ) , with the lma solution even more likely than the sma solution in the case of large _ hep _ flux .from the previous discussion , the following situation emerges for the three msw solutions sma , lma , and low .none of them can be excluded at 99% c.l . by the present experimental data .different pieces of data give indications that are not as consistent as it would be desirable : the total rate information favors the sma solution , the spectral data favor the lma and low solutions , and the day - night data favor the lma solution . in a global fits ,the three solutions have comparable likelihoods .although such solutions are subject to change shape and likelihood as more accurate experimental data become available , no dramatic improvement can be really expected in their selection , unless : 1 ) the theoretical uncertainties on the total rates are reduced to the size of the corresponding experimental uncertainties ; 2 ) the total errors associated to the sk spectrum and day - night measurement are significantly reduced ( by , say , a factor @xmath68 ) ; or 3 ) decisive results are found in new solar neutrino experiments such as the sudbury neutrino observatory ( sno ) @xcite , the gallium neutrino observatory @xcite , borexino @xcite , and kamland @xcite .any of these conditions require a time scale of a few years at least ; the same time scale should then be expected in order to ( patiently ) single out one of the three msw solutions ( sma , lma , or low ) . another aspect of the lma and low solutions emerging from fig .[ f2 ] is their extension to large values of the mixing angle ( @xmath69 ) , which are often assumed to be realized only through the vacuum oscillation solutions .since the possibility of nearly maximal @xmath3 mixing for solar neutrinos has gained momentum after the sk evidence for maximal @xmath70 mixing @xmath71 , it is interesting to study it in detail by dropping the usual `` @xmath72 '' variable and by exploring the full range @xmath73 $ ] , as it was done earlier in @xcite .the subcase @xmath74 will receive special attention in the next section .as stated in sec . [ s1 ] , for large values of @xmath34 ( @xmath75 ev@xmath6 ) the parameter space relevant for @xmath39 solar neutrino oscillations is spanned by the variables @xmath18 . as far as @xmath76 is taken in its full range @xmath77 $ ] , one can assume @xmath78 , since the msw physics is invariant under the substitution @xmath79 at any @xmath32 .( at any @xmath55 and @xmath80 $ ] ) allows to take @xmath78 with @xmath76 in the restricted range @xmath81 $ ] . ] for graphical representations , we prefer to use the mixing variables @xmath82 introduced in @xcite , which properly chart both cases of small and large mixing .the case @xmath83 corresponds to the familiar two - family oscillation scenario , except that now we consider also the usually neglected case @xmath31 ( @xmath84 ) . for each set of observables ( rates , spectrum , day - night difference , and combined data )we compute the corresponding msw predictions and their uncertainties , identify the absolute minimum of the @xmath51 function , and determine the surfaces at @xmath85 , 7.82 , and 11.36 , which define the volumes constraining the @xmath86 parameter space at 90% , 95% , and 99% c.l .such volumes are graphically presented in @xmath87 slices for representative values of @xmath88 .figure [ f7 ] shows the results for the fit to the total rates only .the absolute minimum @xmath89 is reached within the sma solution at nonzero @xmath88 , as reported in the first row of tab .[ chiminima ] .the preference for @xmath90 is , however , not statistically significant , since @xmath91 only .the qualitative behavior of the @xmath39 msw solutions in fig .[ f7 ] is unchanged with respect to the earlier analysis in @xcite : for increasing @xmath88 , the low solution moves to slightly larger values of @xmath92 and then disappears , while the sma and lma regions tend to merge in a single , broad solution which slowly disappears at large values of @xmath32 .in general , the msw allowed region(s ) become less structured for increasing @xmath32 , due to the the decreasing energy dependence of the @xmath23 survival probability @xcite .in the first panel of fig .[ f7 ] it is interesting to note that both the lma and the low solutions are compatible at 99% c.l . with maximal mixing @xmath93 .this compatibility persists for the low solution at small values of @xmath88 , while it is lost for the lma solution , which migrates towards smaller mixing angles . as for the two - family case discussed in the previous section, the addition of the spectrum information ( and , to a lesser extent , of the day - night asymmetry ) is expected to produce significant changes in the c.l.s associated to the msw solutions in fig .[ f7 ] .figure [ f8 ] shows the region excluded by the sk day - night asymmetry , in the same representation as in fig .such region is generally far from the allowed regions in fig .[ f7 ] , and thus the day - night asymmetry contributes only marginally to shape the msw solutions , except at low values of @xmath88 , where it cuts the lower part of the lma solution and disfavors the rightmost part of the sma solution .the region excluded by the sk day - night asymmetry can extend beyond maximal mixing @xmath94 , as it was observed earlier in ref .@xcite for the corresponding kamiokande variable .figure [ f9 ] shows instead the regions excluded by the sk spectrum , which plays a relevant role in disfavoring the zone where the sma and lma solutions tend to merge ( see the six middle panels of fig .in fact , for @xmath95 ev@xmath6 the normalized sk spectrum is predicted to have a large _ negative _ slope ( see , e.g. , @xcite ) , contrary to the experimental results .relatively large positive values for the spectrum slope are excluded in the left diagonal band shown in the upper panels of fig .both positive and negative values for the spectrum slope can instead occur in the excluded region where the earth regeneration effects are relevant @xcite . for increasing @xmath32 , the spectral information becomes less effective in disfavoring zones of the parameter space , since the energy - dependence of the electron survival probability becomes weaker .therefore , only the strongest distortion effect ( at @xmath96 ev@xmath6 ) survives at large @xmath88 .finally , as in the @xmath40 case , the excluded regions are slightly shrinked if large _ hep _flux values are assumed ( not shown ) .figure [ f10 ] shows the combined fit to all data , which is one one of the main results of this work . as for the total rates , the minimum @xmath51 ( second row of tab .[ chiminima ] ) is reached within the sma solution and shows a very weak preference for nonzero values of @xmath32 ( @xmath97 ) . by comparing fig .[ f10 ] with fig .[ f7 ] , it can be seen that the sk spectrum excludes a significant fraction of the solutions at @xmath98 ev@xmath6 , including the upper part of the lma solution at small @xmath32 , and the merging with the sma solution at large @xmath32 .in particular , at @xmath99 the 95% c.l . upper limit on @xmath55 drops from @xmath100 ev@xmath6 ( rates only ) to @xmath101 ev@xmath6 ( all data ) .this indication tends to disfavor neutrino searches of _ cp _ violation effects , since such effects decrease with @xmath102 at nonzero values of @xmath32 .figure [ f10 ] also shows that the inclusion of the present spectrum and day - night information , rather than helping in constraining @xmath88 , tends to slightly weaken its upper bound : for instance , the region allowed at 95% c.l . in the last panel of fig .[ f10 ] is larger than the corresponding one in fig .in fact , the information coming from total rates prefers some energy dependence in the survival probability @xmath36 , and thus disfavors large values of @xmath32 [ see eq .( [ e2 ] ) ] .however , since no clear indication for @xmath103 emerges from the bulk of the sk spectrum , its addition in the fit makes large values of @xmath32 slightly more `` acceptable . ''in addition , at @xmath104 , the lower part of the broad region allowed by total rates ( last three panels of fig . [ f7 ] ) becomes more consistent with the ( weak ) indication for a nonzero night - day difference [ eq .( [ e2 ] ) ] , thus giving some extra allowance for large values of @xmath32 , as shown in the last three panels of fig .[ f10 ] .figure [ f11 ] is analogous to fig .[ f10 ] , the only difference being the flux of _ hep _ neutrinos ( increased by a factor of 20 with respect to bp98 ) .since a large _ hep _ flux allows a better fit to the sk spectrum endpoint at any value of @xmath32 , the global fit is somewhat improved ( third row of tab .[ chiminima ] ) , and the msw solutions are slightly more extended than in fig .as for the @xmath40 case , the best - fit point migrates from the sma to the lma solution . in general , the comparison of figs .[ f10 ] and [ f11 ] shows that the ( uncertain ) value of the _ hep _ neutrino flux is not crucial in shaping the current @xmath39 msw solutions ( although it might become decisive with more accurate spectral data ) .figure [ f12 ] shows the values of @xmath105 as a function of @xmath88 for unconstrained @xmath55 and @xmath76 , coming from the fit to total rates and to all data .the values of @xmath106 can be read from tab .[ chiminima ] .this figure displays more clearly some anticipated features of the current @xmath39 msw solutions ( for standard _hep _ flux ) : 1 ) a slight preference for nonzero values of @xmath88 ; 2 ) the existence of an upper bound on @xmath32 , slightly stronger in the fit to total rates ( @xmath107 at 95% c.l . ) than in the global fit @xmath108 . in the case ofenhanced ( @xmath65 ) _ hep _ flux we find that : 1 ) the absolute minimum of @xmath51 is reached at @xmath83 within the lma solution ; 2 ) for fixed @xmath90 , the local minimum migrates to the sma solution when @xmath109 ; and 3 ) the 95% c.l .upper limit on @xmath88 is slightly weakened to @xmath110 .the 95% c.l .upper bound on @xmath32 coming from solar neutrino data alone @xmath111@xmath112 is consistent with the one coming from atmospheric neutrino data alone @xmath113 @xcite , as well as with the upper limit coming from the combination of chooz and atmospheric data @xmath114 @xcite .this indication supports the possibility that solar , atmospheric , and chooz data can be interpreted in a single three - flavor oscillation framework @xcite . in this case , the chooz constraints on @xmath32 exclude a large part of the @xmath39 msw parameter space ( basically all but the first two panels in figs . [ f7][f10 ] ) .however , even small values of @xmath32 can be interesting for solar @xmath8 phenomenology , as anticipated in the comment to eq .( [ e3 ] ) . figure [ f13 ] shows , for instance ,how the partial and global fits to neutrino event rates ( cl , ga , and sk experiments ) are modified when passing from @xmath83 to @xmath115 .each of the three experiments , taken separately , allows @xmath116 . in the combination ,the low and lma solutions `` touch '' the @xmath117 line at @xmath118 , while for nonzero @xmath32 only the low solution remains consistent with @xmath74 ( and , to some extent , with @xmath31 ) .this fraction of the low solution , usually ignored in analyses using the @xmath29 variable , can be relevant for model building . in particular , there is currently great interest in models predicting exact or nearly bimaximal mixing , namely , @xmath119 ( see , e.g. , @xcite and refs .therein ) .such models imply , for the neutrino mixing matrix @xmath120 , that @xmath121 and that @xmath122 .recent theoretical predictions for ( small ) nonzero values of @xmath123 are discussed , e.g. , in @xcite .figure [ f14 ] shows the section of the volume allowed in the @xmath39 msw parameter space , for @xmath74 ( maximal mixing ) , in the mass - mixing plane @xmath124 .all data are included in the fit ( with standard _hep _ flux .it can be seen that both the lma and low solutions are consistent with maximal mixing ( at 99% c.l . ) for @xmath125 .moreever , the consistency of the low solution with maximal mixing improves significantly for @xmath126 , while the opposite happens for the lma solution .this gives the possibility to obtain nearly bimaximal mixing ( @xmath127 with @xmath32 small ) within the low solution to the solar neutrino problem an interesting possibility for models predicting large mixing angles .we have performed a thorough analysis of the @xmath40 and @xmath39 msw solutions of the solar neutrino problem , discussing the information coming from the total rates , as well as from the sk energy spectrum and day - night asymmetry .the global fit to the data puts an upper bound to the @xmath0 mixing angle , consistent with atmospheric and reactor oscillation searches .a fraction of the msw solutions extends at and beyond maximal @xmath3 mixing @xmath128 , especially for the so - called low solution .therefore , a novel realization of ( nearly ) bimaximal mixing models appears possible .g.l.f . and e.l .thank the organizers of the _ taup 99 _ workshop on topics in astroparticle and underground physics ( paris , college de france ) , where preliminary results of this work were presented .thanks e. kh.akhmedov and p. i. krastev for useful correspondence .this work is co - financed by the italian ministero delluniversit e della ricerca scientifica e tecnologica ( murst ) within the `` astroparticle physics '' project .in this appendix we present an updated version of the @xmath51 statistical analysis of neutrino event rates discussed in ref .@xcite , which has been widely used in many other analyses of the solar neutrino problem , including the one performed by the super - kamiokande collaboration in @xcite .in particular , we take into account the most recent estimates for the relevant bp98 standard solar model ingredients and for their uncertainties @xcite . unless otherwise noted , the notation is the same as in ref .@xcite , to which the reader is referred for further details .table [ tablefluxes ] reports the solar neutrino fluxes @xmath129 , as calculated in the bp98 standard solar model @xcite .the corresponding energy - averaged interaction cross sections @xmath130 for the @xmath131-th detector are given in tab .[ tablecross ] , taking into account updated energy spectra @xcite and cross sections @xcite .the associated @xmath132 relative uncertainties @xmath133 are reported in tab .[ tablecrosserr ] .such tables allow the calculation of the theoretical uncertainties related to the detection process .the uncertainties of the ssm @xmath8 fluxes are mainly related on eleven basic ingredients @xmath134 : five astrophysical @xmath135-factors , the sun luminosity , the metallicity @xmath136 , the sun age , the opacity , the element diffusion , and the @xmath137be electron capture rate @xmath138 .the last two sources of uncertainties represent new entries with respect to the list given in @xcite .the `` diffusion error '' is estimated by taking differences between quantities calculated with and without diffusion @xcite .the uncertainty in the @xmath137be(@xmath139)@xmath137li capture rate @xcite affects ( inversely ) the @xmath21b flux by changing the @xmath137be density for the ( much slower ) competing reaction @xmath137be(@xmath140)@xmath21b .the fractional @xmath132 uncertainties of the eleven input ingredients ( @xmath141 ) are reported in tab .[ tableinput ] .table [ tablealpha ] reports the matrix @xmath142 of logarithmic derivatives @xmath143 , which parameterize the ssm `` response '' to small variations in the input ingredients , and which are crucial to estimate correctly the correlations among the neutrino flux or event rate uncertainties @xcite .as remarked in @xcite for the opacity uncertainty , we have split the known flux errors @xmath144 due to opacity , diffusion , and @xmath137be electron capture , into two arbitrary factors , @xmath142 and @xmath141 ( @xmath145 , 10 , 11 ) , so as to treat homogeneously all sources of uncertainties in the @xmath51 statistics . given the previous ingredients , the standard event rates @xmath146 are calculated as : @xmath147 and the associated theoretical error matrix @xmath148 reads : @xmath149 the off - diagonal entries of @xmath148 induce strong correlations among the theoretical uncertainties for different experiments , as evident in fig .[ f1 ] . in the presence of oscillations ( figs .[ f3][f5 ] ) , the standard values of the partial rates @xmath150 are simply replaced by the corresponding oscillated values . we remind that the entries of the @xmath142 matrix are not independent , being constrained by the luminosity sum rule @xcite . at equilibrium , the standard neutrino fluxes must obey the energy conservation relation @xcite : @xmath151 where @xmath152 is the average neutrino energy for the @xmath153-th neutrino flux , @xmath154 mev @xmath155 s@xmath156 is the solar luminosity , and @xmath157 mev is the overall @xmath158-value for the fusion reaction @xmath159 ( independent from the intermediate reaction chain ) .non - equilibrium corrections , which are especially relevant for cno fluxes , modify this relation into @xcite @xmath160 where the @xmath161 values are given in tab .[ tableluminosity ] . by taking partial derivatives of the above relation with respect to the input ingredients@xmath134 , one gets the sum rule @xcite : @xmath162 which is satisfied with good approximation ( @xmath163 level ) .similarly , if the flux variations are approximately parametrized as power laws of the central central temperature @xmath164 of the sun , @xmath165 the temperature exponents @xmath166 become constrained by the sum rule @xcite @xmath167 which , using the exponents in ref .@xcite , is satisfied only at the @xmath168 level . however, such exponents can be easily adjusted within their uncertainties @xmath169 so as to satisfy exactly the above sum rule .sets of temperature exponents are given in tab .[ tablebeta ] .the set of @xmath170 s is particularly relevant for analyses of the solar neutrino problem with `` unconstrained '' central temperature ( not performed in this work ) .since the above tables and equations refer only to the analysis of the _ total _ neutrino event rates , a final remark is in order about the fit to the sk energy - angle spectral information .for msw analyses , the maximum amount of information about the sk solar neutrino signal is contained in the double - differential distribution of events in nadir angle @xmath171 and electron recoil energy @xmath172. however , the @xmath173 distribution is not yet available outside the sk collaboration , which has released only the two _ projected _ distributions , namely , the ( angle - averaged ) energy spectrum and the ( energy - averaged ) nadir distribution .the experimental errors of such projections ( including statistical fluctuations ) are _ a priori _ strongly correlated , but the published information is not sufficient to recover such correlations . for such reasons , while we include the complete information in @xmath174 above 5.5 mev ( given as an 18-bin spectrum in @xcite ) , we prefer to use only a minimal ( but relevant ) information in @xmath175 , namely , the night - day asymmetry @xmath176 .the @xmath51 function for the energy spectrum is then constructed in the same way as in ref .@xcite , namely , by separating uncorrelated and correlated error components in the @xmath177 error matrix , with off - diagonal correlations equal to @xmath178 . a numerical table of errorsis explicitly reported in @xcite for the 825 day sk spectrum .double counting of the total rate information in sk is avoided by minimizing the @xmath51 with respect to the overall spectrum normalization , taken as a free parameter @xcite . in the global fit, we then sum up the @xmath51 of the spectrum , the @xmath51 of the total rates , and the ( trivial ) @xmath51 of the day - night asymmetry measurement .a similar construction of the global @xmath51 function is also adopted in @xcite ..solar neutrino event rates observed in the five experiments homestake , kamiokande , sage , gallex , and super - kamiokande ( 825 days , @xmath179 mev ) , as compared with the theoretical expectations from the bp98 standard solar model @xcite .the quoted uncertainties are at 1@xmath180 . in our analysis ,the gallex and sage experimental rates are combined in quadrature in a single ( gallium ) rate of @xmath45 snu . [ cols="<,^,^,^,^ " , ] y. suzuki for the super - kamiokande collaboration , in _ lepton - photon 99 _ , proceedings of the xix international symposium on photon and lepton interactions at high energies ( stanford , california , u.s.a . , 1999 ) , to appear ; transparencies available at http://lp99.slac.stanford.edu t. a. kirsten , in _neutrino 98 _ , proceedings of the xviii international conference on neutrino physics and astrophysics ( takayama , japan , 1998 ) , edited by y. suzuki and y. totsuka ; nucl .b ( proc . suppl . )* 77 * ( 1999 ) , p. 26 .the solar neutrino deficit , shown as a discrepancy between data and expectations in the gallium ( ga ) , chlorine ( cl ) , and super - kamiokande total event rates . in each plane , the error ellipses represent 99% c.l .contours for two degrees of freedom ( i.e. , @xmath182 ) .the projection of an ellipse onto one of the axis gives approximately the @xmath183 range for the corresponding rate .data and expectations refer to table [ tablerates ] .the correlation of ssm errors is calculated as in the appendix .two - generation msw solutions to the solar neutrino problem .the upper four panels correspond to the following separate fits to data subsets : total rates ( cl+ga+k+sk ) ; super - kamiokande night - day asymmetry @xmath184 ; super - kamiokande electron energy spectrum with standard _ hep _neutrino flux ; super - kamiokande spectrum with enhanced @xmath185 _ hep _ neutrino flux .the two lower panels show the results of global fits to all data .the thin solid , thick solid , and dashed curves correspond to @xmath186 , 5.99 , and 9.21 .the positions of the local @xmath51 minima in fits including the total rates are indicated by dots .see also tab .[ chisquare ] . fig .the sma solution at best fit ( total rates only , first row of tab . [ chisquare ] ) , compared with the experimental data , in the same planes as in fig .note the excellent agreement between theory and observations .[ f4 ] . as in fig .[ f3 ] , but for the lma solution at best fit ( total rates only , second row of tab .[ chisquare ] ) .[ f5 ] . as in fig .[ f3 ] , but for the low solution at best fit ( total rates only , third row of tab . [ chisquare ] ) .comparison of data and predictions for the super - kamiokande spectrum shape .the theoretical spectra normalization is taken free in the fit .the upper ( lower ) panel corresponds to the case of standard ( @xmath65 ) _ hep _ flux .the sma , lma , and low spectra are calculated in the global best fit points reported in tab .[ chisquare ] ( middle rows for the upper panel and lower rows for the lower panel ) . the sk data are reported from @xcite , and the error bars refer to statistical and total experimental errors . fig .three - flavor msw oscillations : global fit to cl+ga+k+sk rates in the @xmath187 parameter space .the favored regions in each panel correspond to sections of the volume allowed at 90% , 95% , and 99% c.l .( @xmath85 , 7.82 , and 11.36 ) for representative values of @xmath88 .[ f8 ] . as in fig .[ f7 ] , but for the fit to the super - kamiokande night - day asymmetry .the region inside the curves is excluded .[ f9 ] . as in fig .[ f7 ] , but for the fit to the super - kamiokande energy spectrum .the regions inside the curves are excluded .results of the global three - flavor msw fit to all data .notice that , in the first two panels , the 99% c.l .contours are compatible with maximal mixing ( @xmath188 ) for both the low and the lma solutions . fig .[ f11 ] . as in fig .[ f10 ] , but for the case of enhanced ( @xmath65 ) _ hep _ flux .the allowed regions are slightly enlarged with respect to fig .values of @xmath189 as a function of @xmath32 , for unconstrained @xmath190 and @xmath92 . at 95% c.l . , the upper limit on @xmath32 is in the range @xmath191@xmath2 , depending on the data used in the fit and on the value of the _ hep _ flux .regions allowed at 99% c.l . by the total rates only , for @xmath83 ( solid curves ) and @xmath115 ( dotted curves ) . for @xmath115 ,the sma and lma solutions are slightly shifted to lower values of @xmath92 , while the low solution is shifted to higher values ( including the value @xmath74 ) .section of the allowed volume in the @xmath39 parameter space in the plane @xmath124 , for the case of maximal @xmath3 mixing @xmath25 . for @xmath192 , both the lma and low solutions are compatible with maximal mixing at 99% c.l .for small values of @xmath35 , the maximal mixing case favors the low solution .

<S> we perform an updated phenomenological analysis of the mikheyev - smirnov - wolfenstein ( msw ) solutions of the solar neutrino problem , assuming oscillations between two and three neutrino families . </S><S> the analysis includes the total rates of the homestake , sage , gallex , kamiokande and super - kamiokande experiments , as well as the day - night asymmetry and the 18-bin energy spectrum of super - kamiokande . </S><S> solutions are found at several values of the @xmath0 mixing angle . among the most interesting features </S><S> , we find that solar neutrino data alone put the constraint @xmath1@xmath2 at 95% c.l . , and that a fraction of the msw solutions extends at and beyond maximal @xmath3 mixing @xmath4 , especially if the neutrino square mass splitting is in its lower range @xmath5 ev@xmath6 ) and if @xmath0 is nonzero . </S><S> in particular , bimaximal ( or nearly bimaximal ) mixing is possible for atmospheric and msw solar neutrino oscillations within the stringent reactor bounds on @xmath0 . </S>

cs0510078

in the multiple description problem , an information source is encoded into @xmath0 packets and these packets are sent through parallel communication channels .there are several receivers , each of which can receive a subset of the packets and needs to reconstruct the information source based on the received packets . in the most general case ,there are @xmath1 receivers and the packets received in each receiver correspond to one of @xmath1 subsets of @xmath2 .a long standing open problem in the literature @xcite is to characterize the information - theoretic rate region subject to the specified distortion constraints .practical multiple description codes have been discussed in @xcite and recent work @xcite has considered the multiple description problem in the context of the distributed source coding scenario .optimal descriptions of even the gaussian source with quadratic distortion measures have not been fully characterized . in the special case of two descriptions of a scalar gaussian source with quadratic distortion measures ,however , the entire rate region has been characterized in @xcite .our focus is on @xmath0 descriptions of a memoryless _ vector _ gaussian source forwhere @xmath0 individual and a single common receiver ( cf .figure [ fig : md ] ) .each receiver needs to reconstruct the original source such that the empirical covariance matrix of the difference is less than , in the sense of a positive semidefinite ordering , a `` distortion '' matrix . in this setting, the symmetric rate multiple description problem of a scalar gaussian source with symmetric distortion constraints has been characterized in @xcite , but a complete understanding of all other rate - distortion settings is open .our main result is an _exact _ characterization of the sum rate for any specified @xmath3 distortion matrix constraints . with @xmath4 , we characterize the entire rate region . our contribution is two fold : * first , we derive a novel information - theoretic inequality that provides a lower bound to the sum of the description rates .the key step is to avoid using the entropy power inequality , which was a central part of the proof of two descriptions of the scalar gaussian source in @xcite : the vector entropy power inequality is tight only with a certain covariance alignment condition , which arbitrary distortion matrix requirements do not necessarily allow . *second , we show that jointly gaussian descriptions actually achieve the lower bound not by resorting to a direct calculation and comparison , which appears to be difficult for @xmath5 , but instead by arguing the equivalence of certain optimization problems .consider another two description problem of a pair of jointly gaussian memoryless sources as depicted in figure [ fig : md_special ] .there are two encoders that describe this source to three receivers : receiver @xmath6 gets the description of encoder @xmath6 , with @xmath7 and the third receiver receives both the descriptions .suppose receiver @xmath6 is interested in reconstructing the @xmath6th marginal of the jointly gaussian source , with @xmath7 .the third receiver is interested in reconstructing the entire vector source .this description problem is closely related to the vector gaussian description problem that is the main focus of this paper .we exploit this connection and characterize the rate region where the reconstructions have a constraint on the covariance of error at each of the receivers ( in the sense of a positive semidefinite order ) .we have organized the results in this paper as follows . in section [ sec2 ]we give a formal description of the problem and summarize our main result .the derivation of a lower bound is in section [ sec3 ] . in section [ sec4 ]we provide an upper bound and provide conditions for the achievable sum rate to meet the lower bound .we see in section [ sec5 ] that the conditions are indeed satisfied in the special case of a scalar gaussian source .the solution in the case of the more complicated vector gaussian source is in section [ sec6 ] .the solution to the multiple description problem depicted in figure [ fig : md_special ] is the topic of section [ sec71 ] . finally ,while the characterization of the rate region of general multiple descriptions of the gaussian source ( with each receiver having access to some subset of the descriptions ) is still open , we can use the insights derived via our sum rate characterization to solve this problem for a nontrivial set of covariance distortion constraints ; this is done in section [ sec72 ] . a note about the notation in this paper : we use lower case letters for scalars , lower case and bold face for vectors , upper case and bold face for matrices .the superscript @xmath8 denotes matrix transpose .we use @xmath9 and @xmath10 to denote the identity matrix and the all zero matrix respectively , and @xmath11 to denote a diagonal matrix with the diagonal entries equal to @xmath12 .the partial order @xmath13 ( @xmath14 ) denotes positive definite ( semidefinite ) ordering : @xmath15 ( @xmath16 ) means that @xmath17 is a positive definite ( semidefinite ) matrix .we write @xmath18 to denote a gaussian random vector with mean @xmath19 and covariance @xmath20 .all logarithms in this paper are to the natural base .the information source @xmath21\}$ ] is an i.i.d . random process with the marginal distribution @xmath22 , i.e. , a collection of i.i.d .gaussian random vectors . denoting the dimension of @xmath21\}$ ] by @xmath23, we suppose that @xmath24 is an @xmath25 positive definite matrix .there are @xmath0 encoding functions at the source , encoder @xmath26 encodes a source sequence , of length @xmath27 , @xmath28,\;\dots , \ ; { { { \mathbf{x}}}}[n])^t$ ] to a source code @xmath29 , for @xmath30 .this code @xmath31 is sent through @xmath26th communication channel at the rate @xmath32 .there are @xmath0 individual receivers and one central receiver . for @xmath33, the @xmath26th individual receiver uses its information ( the output of the @xmath26th channel ) to generate an estimate @xmath34 @xmath35 of the source sequence @xmath36 .the central receiver uses the output of all the @xmath0 channels to generate an estimate @xmath37 of the source sequence @xmath36 .since we are interested in covariance constraints , the decoder maps can be restricted to be the minimal mean square error ( mmse ) estimate of the source sequence based on the received codewords .so , @xmath38 , \quad l=1,\;\dots,\;l \\ \hat{{{{\mathbf{x}}}}}_0^{n } & = { { \mathbb{e}}}\left[{{{\mathbf{x}}}}^n|f_1^{(n)}({{{\mathbf{x}}}}^n ) , \ ; \dots , \ ; f_l^{(n)}({{{\mathbf{x}}}}^n)\right ] .\end{split}\ ] ] suppose the reconstructed sequences satisfy the covariance constraints @xmath39-\hat{{{{\mathbf{x}}}}}_l[m])^t({{{\mathbf{x}}}}[m]-\hat{{{{\mathbf{x}}}}}_l[m])\big ] & \preccurlyeq \mathbf{d}_l , \quad l=1,\;\dots,\;l , \\\frac{1}{n}\sum\limits_{m=1}^n { { \mathbb{e}}}\big[({{{\mathbf{x}}}}[m]-\hat{{{{\mathbf{x}}}}}_0[m])^t({{{\mathbf{x}}}}[m]-\hat{{{{\mathbf{x}}}}}_0[m])\big ] & \preccurlyeq \mathbf{d}_0 , \end{split}\ ] ] then we say that multiple descriptions with distortion constraints @xmath40 are achievable at the rate tuple @xmath41 .the closure of the set of all achievable rate tuples is called the rate region and is denoted by @xmath42 . throughout this paper , we suppose that @xmath43 . , is without loss of generality is seen by applying the data processing inequality for mmse estimation errors ; having more access to information can only reduce the covariance of the error in a positive semidefinite sense .similarly , @xmath44 is also not interesting ; here we simplify this condition and take @xmath45 . ]our main result is the precise characterization of the sum rate of multiple descriptions for individual and central receivers . for distortion constraints @xmath40 ,the sum rate is @xmath46 this sum rate is achieved by a _ jointly gaussian random multiple description scheme _ :let @xmath47 be zero mean jointly gaussian random vectors independent of @xmath48 , with the positive definite covariance matricex @xmath49 denoted by @xmath50 .defining @xmath51 we consider @xmath52 such that @xmath53 \overset{\text{def}}{= } & { { \mathbb{e}}}\big[({{{\mathbf{x}}}}-{{\mathbb{e}}}[{{{\mathbf{x}}}}|{{{\mathbf{u}}}}_l])^t({{{\mathbf{x}}}}-{{\mathbb{e}}}[{{{\mathbf{x}}}}|{{{\mathbf{u}}}}_l])\big ] \preccurlyeq \mathbf{d}_l , \quad l=1,\;\dots,\;l , \\ { { \text{cov}}}[{{{\mathbf{x}}}}|{{{\mathbf{u}}}}_1,\;\dots,\;{{{\mathbf{u}}}}_l ] \overset{\text{def}}{= } & { { \mathbb{e}}}\big[({{{\mathbf{x}}}}-{{\mathbb{e}}}[{{{\mathbf{x}}}}|{{{\mathbf{u}}}}_1,\;\dots,\;{{{\mathbf{u}}}}_l])^t({{{\mathbf{x}}}}-{{\mathbb{e}}}[{{{\mathbf{x}}}}|{{{\mathbf{u}}}}_1,\;\dots,\;{{{\mathbf{u}}}}_l])\big ] \preccurlyeq \mathbf{d}_0 . \end{split}\ ] ] to construct the code book for the @xmath26th description , draw @xmath54 @xmath55 vectors randomly according to the marginal of @xmath56 .the encoders observe the source sequence @xmath36 , look for codewords @xmath57 that are jointly typical with @xmath36 and send the index of the resulting @xmath55 through the @xmath26th channel , respectively .the @xmath26th individual receiver uses this index and generates a reproduction sequence @xmath58 $ ] for @xmath30 , the central receiver uses all the @xmath0 indices to generate a reproduction sequence @xmath59 $ ] . for every @xmath50 satisfying ,the rate tuple @xmath60 satisfying @xmath61 is achievable by using this coding scheme , where @xmath62 is the covariance matrix for all @xmath63 , and @xmath64 $ ] . in particular ,the achievable sum rate is @xmath65 we denote this ensemble of descriptions , throughout this paper , as the jointly gaussian description scheme and the time sharing between them as the jointly gaussian description strategy .we show that jointly gaussian description schemes are optimal in achieving the sum rate . for two descriptions, we can characterize the entire rate region .given distortion constraints @xmath66 , the rate region for the two description problem for an i.i.d.@xmath67 vector gaussian source is @xmath68 we show that if the distortion constraints @xmath66 satisfy @xmath69 and @xmath70 , we can get the optimizing @xmath71 by solving a matrix riccati equation .an illustration of the rate region is shown in figure [ fig : region ] . in this case , if we let @xmath72^{-1}$ ] for @xmath73 , then the optimizing @xmath71 is @xmath74 where @xmath75^{\frac{1}{2}}({{\mathbf{k}}}_{w_1}-{{\mathbf{k}}}_{w_0})^{\frac{1}{2 } } -{{\mathbf{k}}}_{w_0}.\ ] ] letting @xmath76 denote the optimal sum rate , the two corner points in figure [ fig : region ] are @xmath77 @xmath78[ sec3 ] by fairly procedural steps , we have the following lower bound to the sum rate of the multiple descriptions : @xmath79 where we have defined @xmath80 and called it the symmetric mutual information between @xmath81 .note that @xmath82 and is also well defined even when @xmath81 are continuous random variables .our main result is the following information theoretic inequality which gives a lower bound to the sum of symmetric mutual information between @xmath83 and mutual information between @xmath84 and @xmath36 for given covariance constraints .[ lemma : inform ] let @xmath28,\;\dots,\;{{{\mathbf{x}}}}[n])$ ] , where @xmath85 $ ] s are i.i.d .@xmath86 gaussian random vectors for @xmath87 .let @xmath81 be random variables jointly distributed with @xmath36 .let @xmath88 $ ] and @xmath89 $ ] for @xmath90 .given positive definite matrices @xmath91 , if @xmath92 - \hat{{{{\mathbf{x}}}}}_l[m])^t({{{\mathbf{x}}}}[m ] - \hat{{{{\mathbf{x}}}}}_l[m ] ) ] & \preccurlyeq \mathbf{d}_l , \quad l=1,\;\dots,\;l ,\\ \frac{1}{n}\sum\limits_{m=1}^n { { \mathbb{e}}}[({{{\mathbf{x}}}}[m ] - \hat{{{{\mathbf{x}}}}}_0[m])^t({{{\mathbf{x}}}}[m ] - \hat{{{{\mathbf{x}}}}}_0[m ] ) ] & \preccurlyeq \mathbf{d}_0 , \end{split}\ ] ] then @xmath93 furthermore , there exists a jointly gaussian distribution of @xmath94 such that the inequality in is tight .this is a fundamental information - theoretic inequality which involves only the joint distribution are not simply functions of @xmath36 and can also be continuous random variables . ] between @xmath84 and @xmath36 and bounds on mean square error estimation of @xmath36 from @xmath84 ; we delegate the proof of this result to appendix [ app : inform ] .we can now use lemma [ lemma : inform ] to derive a lower bound to the sum rate @xmath95 by letting @xmath96 in the lemma above , we can derive a simple lower bound to the rate of the individual descriptions as well : @xmath97 this bound is actually the point - to - point rate - distortion function for individual receivers , since each individual receiver only faces a point - to - point compression problem .note that for any positive definite @xmath71 , @xmath98 is a lower bound to the sum rate of the multiple descriptions .two special choices of @xmath71 are of particular interest : * letting @xmath99 and 0 @xmath100 , we have the following lower bound : @xmath101 this bound is actually the summation of the bounds on the individual rates . *letting some eigenvalues of @xmath102 goes to infinity , we have the following lower bound : @xmath103 this bound is the point - to - point rate - distortion function when we only have the central distortion constraint .we will see later that for some distortion constraints @xmath40 , and can be tight .[ sec4 ] in the previous section we gave a lower bound to the sum rate .now we give a upper bound to the sum rate by using the jointly gaussian description scheme described in section [ sec : sumrate ] .first we give a sketch of the achievable rate region by using jointly gaussian description scheme . given the source sequence @xmath36 , as long as we can find a combination of codewords @xmath57 that are jointly typical with @xmath36 , all the receivers can generate reproduction sequences that satisfy their given distortion constraints .an intuitive way to understand is the following : since @xmath57 are jointly typical with @xmath36 , then for any @xmath104 , we have that @xmath105 are jointly typical with @xmath36 . now the probability that a randomly generated combination of codewords @xmath105 are jointly typical with @xmath36 is roughly @xmath106 and the number of possible combination of codewords @xmath107 are @xmath108 .thus , as long as @xmath109 we can find a combination of codewords @xmath105 that are jointly typical with @xmath36 . rigorously speaking , we need to show that as long as is satisfied , then for any given source sequence @xmath36 we can find a combination of codewords @xmath57 such that @xmath105 are jointly typical with @xmath36 for all @xmath110 .second moment method_@xcite is commonly used to address this aspect , and a proof can be found in @xcite .evaluating based on the jointly gaussian distribution of @xmath48 and @xmath111 , we get that all the rate tuples @xmath60 satisfying @xmath112 are achievable by the jointly gaussian description scheme . in particular , we have that the achievable sum rate is @xmath113 the resulting distortions @xmath114 by using jointly gaussian description scheme can be calculated as @xmath115 = [ { { \mathbf{k}}}_x^{-1 } + \mathbf{k}_{w_l}^{-1}]^{-1 } , \quad l=1,\;\dots,\ ; l , \\ { { \mathbf{d}}}^*_0 = & { { \text{cov}}}[{{{\mathbf{x}}}}|{{{\mathbf{u}}}}_1,\;\dots,\;{{{\mathbf{u}}}}_l ] = [ { { \mathbf{k}}}_x^{-1 } + ( { { \mathbf{i}}},\ ; \dots , \ ; { { \mathbf{i}}})\mathbf{k}_w^{-1}({{\mathbf{i}}},\ ; \dots , \ ; { { \mathbf{i}}})^t]^{-1}. \end{split}\ ] ] the achievable region given in has useful combinatorial properties ; in particular it belongs to the class of _ contra - polymatroids_@xcite . certain rate regions of the multiple access channel @xcite and distributed source coding problems @xcite are also known to have this specific combinatorial property . to see this , let @xmath116 we can readily verify that @xmath117 by definition, we conclude that the achievable rate region of a jointly gaussian multiple description scheme is a contra - polymatroid .the key advantage of this combinatorial propety is that we can exactly characterize the vertices of the achievable rate region .letting @xmath118 to be a permutation on @xmath2 , define @xmath119 and @xmath120 .then the @xmath121 points @xmath122 are the vertices of the contra - polymatroid .our goal is to show that the jointly gaussian description scheme achieves the lower bound to the sum rate . in generalit does not seem facile to do a direct calculation and comparison .we forgo this strategy and , instead , provide an alternative characterization of the achievable sum rate which is much easier to compare with the lower bound .similar to the derivation of the lower bound ( in appendix [ app : inform ] ) , we consider an @xmath123 gaussian random vector @xmath124 , independent of @xmath48 and all @xmath125 s .defining @xmath126 , we have the following achievable sum rate : @xmath127+\mathbf{k}_z\big| } { \big|{{\text{cov}}}[{{{\mathbf{x}}}}|{{{\mathbf{u}}}}_1,\;\dots,\;{{{\mathbf{u}}}}_l]\big|\prod\limits_{l=1}^l\big|{{\text{cov}}}[{{{\mathbf{x}}}}|{{{\mathbf{u}}}}_l]+\mathbf{k}_z\big|},\end{aligned}\ ] ] where the last step is from a procedural gaussian mmse calculation .note that if we have @xmath128 then ( a ) in is actually an equality .thus , if our choice of @xmath52 and @xmath102 satisfy the following two conditions : * is true .* distortion constraints are met with equality , i.e. , @xmath129 = \mathbf{d}_l , \quad l=1,\;\dots,\ ; l , \\ & { { \text{cov}}}[{{{\mathbf{x}}}}|{{{\mathbf{u}}}}_1,\;\dots,\;{{{\mathbf{u}}}}_l ] = \mathbf{d}_0 , \end{split}\ ] ] then the upper bound matches the lower bound and we have characterized the sum rate . in the following we examine underwhat circumstances the above two conditions are true .first , we give a necessary and sufficient condition for to be true , delegating the proof to appendix [ app : nec_suff_condition ] . [prop : nec_suff_condition ] there exists some choice of positive definite @xmath71 such that is true if and only if @xmath50 , the covariance matrix of @xmath130 , takes the following form @xmath131 where @xmath132 .next , we look at the conditions for to be true . from , we have @xmath133^{-1 } = { { \mathbf{k}}}_x^{-1 } + \mathbf{k}_{w_l}^{-1 } , \quad l=1,\;\dots,\ ; l \\\mathbf{d}_0^{-1 } & = { { \text{cov}}}[{{{\mathbf{x}}}}|{{{\mathbf{u}}}}_1,\;\dots,\;{{{\mathbf{u}}}}_l]^{-1 } = { { \mathbf{k}}}_x^{-1 } + ( { { \mathbf{i}}},\ ; \dots , \ ; { { \mathbf{i}}})\mathbf{k}_w^{-1}({{\mathbf{i}}},\ ; \dots , \ ; { { \mathbf{i}}})^t . \end{split}\ ] ] @xmath134 , is calculated in the following lemma ; the proof is available in appendix [ app : inverse ] .[ lemma : inverse ] let @xmath135 if @xmath136 and @xmath137 , then @xmath138^{-1}.\ ] ] using this lemma , from we arrive at @xmath139^{-1 } = \sum\limits_{l=1}^l\left[({{\mathbf{d}}}_l^{-1}-{{\mathbf{k}}}_x^{-1})^{-1}+{{\mathbf{a}}}\right]^{-1}.\ ] ] defining @xmath140 is equivalent to @xmath141^{-1 } = \sum\limits_{l=1}^l\left[{{\mathbf{k}}}_{w_l}+{{\mathbf{a}}}\right]^{-1}.\ ] ] thus , if there exists a positive definite solution @xmath142 to , and the corresponding @xmath50 is positive definite , then the distortion constraints are met with equality , i.e. , holds .it turns out that as long as @xmath142 is a solution to , the resulting @xmath50 is always positive definite ; we state this formally below , delegating the proof to appendix [ app : positivedefinite ] .[ lemma : positivedefinite ] if for some @xmath143 and @xmath144 is true , then the covariance matrix @xmath50 defined in is positive definite .we summarize the state of affairs in the following theorem .[ th : main ] given distortion constraints @xmath145 , let @xmath146 if there exists an solution @xmath147 to and @xmath148 , then the jointly gaussian description scheme with @xmath50 defined in with @xmath149 achieves the optimal sum rate , and the optimal @xmath71 for lower bound is @xmath150 . thus we show that if the given distortion constraints @xmath151 satisfy the condition for theorem [ th : main ] , then the jointly gaussian description scheme achieves the optimal sum rate and we can calculate the optimal @xmath50 by solving a matrix equation .however , for arbitrarily given distortion constraints , may not have a solution @xmath147 such that @xmath148 . in this case, we can show that there exists a jointly gaussian description scheme that achieves the sum rate lower bound , and resulting in distortions @xmath152 such that @xmath153 for @xmath154 . in the following we first study the relatively simpler case of scalar gaussian source , andthen move to discuss the vector gaussian source .[ sec5 ] here we suppose that the information source is an i.i.d . sequence of @xmath155 scalar gaussian random variables .let individual distortion constraints be @xmath156 and the central distortion constraints be @xmath157 , where @xmath158 for @xmath159 .we consider the jointly gaussian description scheme with the following covariance matrix for @xmath160 .@xmath161 consider the condition for theorem [ th : main ] to hold : to meet the individual distortion constraint with equality , we need @xmath162 let @xmath163 we need @xmath164^{-1 } = \sum\limits_{l=1}^l\left[\sigma_l^2+a\right]^{-1}\ ] ] to have a solution @xmath165 , to meet the central distortion constraint with equality . towards this , define @xmath166 and we have @xmath167 using induction , we can show that @xmath168 thus we have @xmath169 then given distortions @xmath170, @xmath171 and @xmath172 falls into the following three cases .* case 1 : * @xmath173 and @xmath174 . in this case , since @xmath175 is a continuous function , there exists an @xmath176 such that @xmath177 . in this casethe condition for theorem [ th : main ] holds and from theorem [ th : main ] we know that jointly gaussian description scheme with covariance matrix for @xmath160 being with @xmath178 achieves the optimal sum rate . * case 2 : * @xmath179 .alternatively , @xmath180 . in this case, the condition for theorem [ th : main ] does not hold .but the jointly gaussian description scheme can still achieve the sum rate . to see this , choosing @xmath181 in @xmath52 we can meet individual distortions with equality and get a central distortion @xmath182 .from we have @xmath183 hence we have achieved distortion @xmath184 where @xmath185 , and from the achievable sum rate is @xmath186 which equals the sum of our bounds on individual rates .* case 3 : * @xmath187 , alternatively , @xmath188 . in this case, the conditions for theorem [ th : main ] do not hold as well .but the jointly gaussian description strategy still achieves the sum rate . to see this , note that we can find a @xmath189 such that @xmath190 and @xmath191 and we choose @xmath192 , @xmath193 for @xmath194 , and @xmath195 in @xmath196 . defining @xmath197 , is equivalent to the following equation : @xmath198^{-1 } = \sum\limits_{l=1}^l\left[\sigma_l^2+\sigma_x^2\right]^{-1}.\ ] ] from lemma [ lemma : positivedefinite ] , our choice of @xmath196 is positive definite .thus the resulting distortions are @xmath199 , where @xmath190 . using the determinant equation @xmath200 and, we have an achievable sum rate @xmath201 we conclude that in this case the point - to - point rate - distortion bound for the central receiver is achievable . in summary , we have shown that the jointly gaussian description scheme achieves the lower bound on the sum rate .further , the sum rate can be calculated either trivially ( by choosing @xmath202 in case ii or @xmath203 in case iii ) or by solving a polynomial equation in a single variable ( case i ) .the essence of our proof of the optimality of jointly gaussian description scheme for scalar gaussian sources is the use of the _ intermediate value theorem _ for scalar continuous functions .however , there is no natural extension of this theorem for vector valued functions . to avoid this problem, we first explicitly solve the two description problem and characterize the optimality of jointly gaussian description scheme .next , we show that the jointly gaussian description scheme is optimal for @xmath204 by showing an equivalence of certain optimization problems . in the last part of this section ,we show that the jointly gaussian description strategy can achieve the optimal rate region for the two description problem . with only two descriptions , we can explicitly solve , thus generalizing the corresponding solution for the scalar gaussian source , derived in @xcite .suppose the distortion constraints are denoted by @xmath205 and let @xmath206 we now solve , which is equivalent to , for @xmath207 , @xmath208 and @xmath147 .from we get @xmath209 and @xmath210 expanding out @xmath211 using lemma [ lemma : blockinverse ] in appendix [ app : matrix ] , we get @xmath212 taking inverse on both sides , we have @xmath213 defining @xmath214 as @xmath215^{-1 } , \ ] ] is equivalent to @xmath216 defining @xmath217 is equivalent to @xmath218 which is further equivalent to @xmath219 this is a version of the so - called _ algebraic riccati equation _ ; the corresponding hamiltonian is readily seen to be positive semidefinite and we can even write down the following explicit solution : @xmath220^{\frac{1}{2}}({{\mathbf{k}}}_{w_1}-{{\mathbf{k}}}_{w_0})^{\frac{1}{2}}. \end{split}\ ] ] thus @xmath75^{\frac{1}{2}}({{\mathbf{k}}}_{w_1}-{{\mathbf{k}}}_{w_0})^{\frac{1}{2 } } -{{\mathbf{k}}}_{w_0}.\ ] ] now , if @xmath148 then we can appeal to theorem [ th : main ] and arrive at the explicit jointly gaussian description scheme parameterized by @xmath50 that achieves the sum rate .analogous to the scalar case ( cf .@xcite ) , we have the following sufficient condition for when this is true ; the proof is available in appendix [ app : twod ] .[ prop : twod ] if the distortion constraints @xmath66 satisfy @xmath221 then @xmath148 .we now complete the proof by considering the cases that are not covered by the conditions in proposition [ prop : twod ] .* when @xmath222 we can choose @xmath223 to achieve the sum of point - to - point individual rate - distortion functions .thus in this case , the sum rate is equal to this natural lower bound .* when @xmath224 we can choose @xmath225 to achieve the point - to - point rate distortion - function for central receiver , also a natural lower bound .* when neither @xmath226 nor @xmath227 is positive or negative semidefinite ( this case can not happen in the scalar case ) , we can not use theorem [ th : main ] , and the trivial choice of @xmath228 or @xmath225 does not meet the lower bound . in the next subsectionwe will address this case and prove that the jointly gaussian description scheme indeed achieves the lower bound on the sum rate for @xmath229 .if we let the source to be scalar gaussian , our result reduces to ozarow s solution of the two description problem for a scalar gaussian source@xcite : this is because the last case described above does not happen in the scalar case . while we exactly characterized the optimal jointly gaussian description scheme and used this characterization in arguing that it achieves the fundamental lower bound to the sum rate, such exact calculations do not appear to be as immediate when @xmath230 .so , we eschew this somewhat brute - force approach and resort to a more subtle proof that involves exploring the structure of the solution to an optimization problem .first , note that by a linear transformation at the encoders and the decoders , we have the following result on rate region for multiple description with individual and central receivers .@xmath231 thus , throughout this subsection we will suppose , for notation simplicity , that @xmath232 .given distortion constraints @xmath233 , let @xmath234 and define @xmath235^{-1 } - \sum\limits_{l=1}^l\left[{{\mathbf{k}}}_{w_l}+{{\mathbf{a}}}\right]^{-1 } , \\f({{\mathbf{a } } } ) \overset{\text{def}}{= } & \log |{{\mathbf{k}}}_{w_0}+{{\mathbf{a}}}| - \sum\limits_{l=1}^l \log |{{\mathbf{k}}}_{w_l}+{{\mathbf{a}}}|.\end{aligned}\ ] ] note that @xmath236 consider the following optimization problem : @xmath237 now , since @xmath238 is a continuous map and @xmath239 is a compact set , there exists an optimal solution @xmath147 to where @xmath147 satisfies the karush - kuhn - tucker ( kkt ) conditions : there exist @xmath240 and @xmath241 such that @xmath242 now @xmath147 falls into the following four cases .* case 1 : * @xmath243 .alternatively , 0 and 1 are not eigenvalues of @xmath147 . in this case ,@xmath244 and @xmath245 ; thus the kkt conditions in reduce to @xmath246 equivalently , @xmath247^{-1 } = \sum\limits_{l=1}^l\left[{{\mathbf{k}}}_{w_l}+{{\mathbf{a}}}^*\right]^{-1}.\ ] ] from theorem [ th : main ] , the jointly gaussian description scheme with covariance matrix for @xmath248 being @xmath249 achieves the lower bound to the sum rate .thus in this case , we have characterized the optimality of the jointly gaussian description scheme parameterized by in terms of achieving the sum rate .* case 2 : * @xmath250 .alternatively , some eigenvalues of @xmath147 are 0 , but no eigenvalues of @xmath147 are 1 . thus @xmath251 and the kkt conditions in reduce to @xmath252 for some @xmath253 satisfying @xmath254 .the _ key idea _ now is to see that the distortion constraint on the central receiver is too loose and we can in fact achieve a _ lesser _ distortion ( in the sense of positive semidefinite ordering ) for the same sum rate .we first identify this lower distortion : defining @xmath255 consider the smaller distortion matrix on the central receiver @xmath256 this new distortion matrix on the central receiver satisfies two key properties , that we state as a lemma ( whose proof is available in appendix [ app : enhanced0 ] ) .[ lemma : enhanced0 ] @xmath257 comparing with , we have @xmath258^{-1 } = \sum\limits_{l=1}^{l}\left[{{\mathbf{k}}}_{w_l}+{{\mathbf{a}}}^*\right]^{-1}.\ ] ] now , the corresponding @xmath259 is singular .if it had nt been , then by theorem [ th : main ] we could have concluded that jointly gaussian description scheme achieves the lower bound to the sum rate .we now address this technical difficulty .our first observation is that there exists @xmath260 such that for all @xmath261 we have @xmath262 , and @xmath263 , @xmath264 , and we can rewrite as @xmath265^{-1 } = \sum\limits_{l=1}^l\big[({{\mathbf{k}}}_{w_l}-\epsilon { { \mathbf{i}}})+({{\mathbf{a}}}^*+\epsilon { { \mathbf{i}}})\big]^{-1}.\ ] ] thus if the distortion constraints were @xmath266 with @xmath267^{-1 } , \quad { { l=1,\;\dots,\;l } } , \\ { { \mathbf{d}}}_0(\epsilon ) & = \left[({{\mathbf{k}}}_{w_0}^*-\epsilon { { \mathbf{i}}})^{-1}+{{\mathbf{i}}}\right]^{-1 } , \end{split}\ ] ] then @xmath268 is a solution to .this situation corresponds to that discussed in case i ; we can conclude that sum rate for this modified distortion multiple description problem is @xmath269 where @xmath270^{-1}-{{\mathbf{i}}}$ ] .we would like to let @xmath271 approach zero and consider the limiting multiple description problem .in particular , we show that @xmath272 as @xmath273 in appendix [ app : depsilon ] .further , we show that @xmath274 as @xmath273 in appendix [ app : kzepsilon ] .thus we can conclude that the sum rate approaches , using , @xmath275 as @xmath276 ; here @xmath259 .we observe that this sum rate is achievable using the jointly gaussian multiple scheme .further , this sum rate is identical to the lower bound to sum rate for the original distortions @xmath277 .thus we conclude the optimality of the jointly gaussian description scheme in this case as well .* case 3 : * @xmath278 .alternatively , some eigenvalues of @xmath147 are 1 , but no eigenvalues of @xmath147 are 0 . in this case , the @xmath244 and the kkt conditions in reduce to @xmath279 for some @xmath280 satisfying @xmath281 . defining @xmath282^{-1}-{{\mathbf{i}}},\ ] ] we have , as in , that @xmath283 the observation @xmath284^{-1}+{{\mathbf{\lambda}}}_2,\ ] ] combined with the proof of suffices to justify .now , from , @xmath285 as in the previous case , the key step is to identify smaller distortion matrices at each of the individual receivers ( ordered in the positive semidefinite sense ) that is achievable at the same sum rate : @xmath286^{-1 } , \quad l = 1 , \ ; \ldots , \ ; l.\ ] ] to see that this is indeed a smaller distortion matrix , observe that since @xmath50 is positive definite , it follows that @xmath287 and @xmath288^{-1 } \\ & = \left[\left(\left ( ( { { \mathbf{k}}}_{w_l}+{{\mathbf{i}}})^{-1}+{{\mathbf{\lambda}}}_2\right)^{-1}-{{\mathbf{i}}}\right)^{-1}+{{\mathbf{i}}}\right]^{-1 } \\ & = \left [ { { \mathbf{i}}}- ( { { \mathbf{k}}}_{w_l}+{{\mathbf{i}}})^{-1}-{{\mathbf{\lambda}}}_2\right ] \\ & = \left [ { { \mathbf{i}}}+ { { \mathbf{k}}}_{w_l } \right]^{-1 } -{{\mathbf{\lambda}}}_2 \\ & = { { \mathbf{d}}}_l - { { \mathbf{\lambda}}}_2 , \quad l = 1 , \ ; \ldots , \ ; l. \end{split}\ ] ] since @xmath289 , it follows that @xmath290 .define @xmath291^{-1 } , \quad l=0,\;1,\;\dots,\;l-1 , \\{ { \mathbf{d}}}_l(\epsilon ) & = \left[({{\mathbf{k}}}_{w_l}^*+\epsilon { { \mathbf{i}}})^{-1}+{{\mathbf{i}}}\right]^{-1 } , \end{split}\ ] ] then there exists @xmath260 such that for all @xmath261 we have @xmath292 , and @xmath293 .we can rewrite as @xmath294^{-1 } = \sum\limits_{l=1}^{l-1}\left[({{\mathbf{k}}}_{w_l}+\epsilon { { \mathbf{i}}})+({{\mathbf{a}}}^*-\epsilon { { \mathbf{i}}})\right]^{-1}+\left[({{\mathbf{k}}}_{w_l}^*+\epsilon { { \mathbf{i}}})+({{\mathbf{a}}}^*-\epsilon { { \mathbf{i}}})\right]^{-1}.\ ] ] thus if the distortion constraints were @xmath295 , then @xmath296 is a solution to .this situation corresponds to that discussed in case i ; we conclude that the sum rate for this modified distortion multiple description problem is @xmath269 where @xmath297^{-1}-{{\mathbf{i}}}$ ] .we would like to let @xmath271 approach zero and consider the limiting multiple description problem .similar to equations and , we have @xmath298 further , we show that @xmath299 .we can now conclude that the sum rate approaches @xmath300 as @xmath271 approaches 0 .in other words , the point - to - point rate - distortion function for central receiver with distortion @xmath301 can be achieved by using the jointly gaussian description scheme , and the resulting distortion is @xmath302 where @xmath303 . in conclusion , the jointly gaussian description scheme is also optimal in this case .* case 4 : * @xmath304 .i.e. , both 0 and 1 are eigenvalues of @xmath147 . in this case , the kkt conditions are : there exist @xmath305 and @xmath306 such that equations , and hold .we can combine equations and to get @xmath307 where @xmath308^{-1}-{{\mathbf{i}}}. \end{split}\ ] ] as in cases 2 and 3 , we want to show the optimality of the jointly gaussian multiple description scheme through a limiting procedure .we do this by first perturbing @xmath147 so that it has no eigenvalue equal to 0 or 1 as follows . without loss of generality ,suppose that @xmath147 has @xmath309 eigenvalues equal to 0 and @xmath310 eigenvalues equal 1 , where @xmath311 and @xmath312 , and there exists @xmath25 orthogonal matrix @xmath20 such that @xmath313 with @xmath314 .we need to perturb the eigenvalues of @xmath147 away from both 0 and 1 . towards this , we define two @xmath25 diagonal matrices : @xmath315 also define @xmath316 further , defining @xmath317 there exists @xmath260 such that for all @xmath318 and @xmath319 we have @xmath320 , and @xmath321 .now , we can rewrite as @xmath322^{-1 } = \sum\limits_{l=1}^{l}\big[{{\mathbf{k}}}_{w_l}(\epsilon_1,\epsilon_2)+{{\mathbf{a}}}^*(\epsilon_1,\epsilon_2)\big]^{-1}.\ ] ] thus if the distortion constraints were @xmath323 , then @xmath324 is a solution to .this situation corresponds to that discussed in case i ; we conclude that the sum rate for this modified distortion multiple description problem is @xmath325^{-1}-{{\mathbf{i}}}$ ] .we would like to let @xmath326 and @xmath327 approach zero and consider the limiting multiple description problem .similar to equations and , when @xmath326 and @xmath327 approach 0 , we get @xmath328 where @xmath329 as in case 3 and @xmath330^{-1}$ ] as in case 2 .further , we show that @xmath331 in appendix [ app : case4e1e2 ] .we conclude that the sum rate approaches @xmath300 as @xmath326 and @xmath327 approach 0 . thus the point - to - point rate - distortion function for central receiver with distortion @xmath301 can be achieved by using the jointly gaussian description scheme , andthe resulting distortions are @xmath332 where @xmath333 and @xmath334 .in other words , the jointly gaussian multiple description scheme is also optimal in this case . to summarize, we see that the jointly gaussian description scheme achieves the limiting sum rate .the limiting sum rate is the solution to an optimization problem .for some specific distortion constraints , the sum rate can be characterized as the solution to a matrix polynomial equation ( case i ) . applying the result in section [ sec : vecl ] to the case of @xmath4 , i.e. , the two description problem, we can see that jointly gaussian description scheme achieves the optimal sum rate .this resolves the case left out in section [ sec : vector2 ] .it also turns out that in the two description problem , we can show that jointly gaussian description strategy achieves the entire rate region .this is the main result of this subsection . from section [ sec : outerbound ]we have a outer bound to the rate region for the two description problem @xmath335 following the discussion in section [ sec : vecl ] , we show in the following that the jointly gaussian description strategy ( jointly gaussian multiple description schemes and the time sharing between them ) achieves the outer bound to the rate region .let @xmath336 and @xmath337 now consider the optimization problem : @xmath338 as in section [ sec : vecl ] , the optimal solution @xmath147 falls into four cases .* case 1 : * @xmath339 . in this case , we know from section [ sec : innerbound ] that the rate pair @xmath340 satisfying @xmath341 is achievable using the jointly gaussian multiple description scheme with the covariance matrix of @xmath342 being @xmath206 denoting the resulting distortions as @xmath343 , we readily calculate @xmath344 for @xmath345 . from the discussion in section [ sec : vecl ] , we know that the lower bound to sum rate is achieved using this jointly gaussian description scheme .thus , in this case , the jointly gaussian description scheme achieves the rate region . as an aside , we note in this case that , @xmath147 satisfies @xmath346^{-1 } = \left[{{\mathbf{k}}}_{w_1}+{{\mathbf{a}}}^*\right]^{-1}+\left[{{\mathbf{k}}}_{w_2}+{{\mathbf{a}}}^*\right],\ ] ] and , from the discussion in section [ sec : vector2 ] , that a sufficient condition for this case to happen is .* case 2 : * @xmath347 .this case is similar to case 1 : the jointly gaussian description scheme with covariance matrix for @xmath342 being @xmath206 achieves the lower bound on the rate region .we note that in this case the resulting distortions are @xmath348 , with @xmath349 .further , we know from the discussion in [ sec : vector2 ] , that a sufficient condition for this case to happen is @xmath350 * case 3 : * @xmath351 . in this case , we know from the discussion in section [ sec : vecl ] that for another two description problem with distortions @xmath352 such that @xmath353 , the jointly gaussian description scheme with covariance matrix for @xmath342 being @xmath354 achieves the lower bound to sum rate @xmath355 to the original distortions @xmath66 .we can see , from the contra - polymatroid structure of the achievable region of jointly gaussian description scheme , that the corner point @xmath356 in figure [ fig : region ] is achievable by this jointly gaussian description scheme. now observe that the discussion in case 3 of section [ sec : vecl ] is symmetric with respect to the individual receivers .thus , by exchanging the role of receiver 1 and receiver 2 , we can achieve the other corner point @xmath357 in figure [ fig : region ] by another appropriate jointly gaussian description scheme .finally , time sharing between these two jointly gaussian multiple description schemes allows us to achieve the lower bound on the rate region . as an aside, we note , as a consequence of the discussion in section [ sec : vector2 ] , that a sufficient condition for this case to happen is @xmath358 * case 4 : * @xmath359 . in this case , we know , from the discussion in section [ sec : vecl ] , that for another two description problem with distortions @xmath360 such that @xmath361 and @xmath349 , the jointly gaussian description scheme with covariance matrix for @xmath362 being @xmath354 achieves the lower bound to sum rate @xmath355 to the original distortions @xmath66 . using an argument entirely analogous to that applied that the jointly gaussian description strategy achieves the rate region .to summarize : the jointly gaussian description strategy achieves the rate region for the two description problem . for a class of distortion constraints ,the corner points of the rate region can be characterized by solving a matrix polynomial equation , as already seen in section [ sec : vector2 ] .although multiple description for individual and central receivers is a special case of the most general multiple description problem , the solution to this problem sheds substantial insight to the issue - at - large . in this section , we discuss two instances of other multiple description problems that can be resolved using the insights developed so far .in particular , we discuss the problem of two descriptions with separate distortion constraints and the general multiple description problem for some special sets of distortion constraints .the problem of two descriptions with separate distortion constraints is ilustrated in figure [ fig : md_special ] .suppose the vector gaussian source @xmath85=({{{\mathbf{x}}}}_1[m],{{{\mathbf{x}}}}_2[m])$ ] , the dimension of @xmath363 $ ] is @xmath364 and the dimension of @xmath365 $ ] is @xmath366 .this implies that the dimension of @xmath85 $ ] is @xmath367 .let @xmath368^t{{{\mathbf{x}}}}[m]]$ ] , @xmath369^t{{{\mathbf{x}}}}_1[m]]$ ] , and @xmath370^t{{{\mathbf{x}}}}_2[m]]$ ] .there are two encoders at the source providing two descriptions of @xmath85 $ ] .there are three receivers : the individual receivers 1 and 2 are only interested in generating reproduction of @xmath363 $ ] with mean square distortion constraint @xmath371 ( an @xmath372 positive definite matrix ) from description 1 and @xmath365 $ ] with mean square distortion constraint @xmath373 ( an @xmath374 positive definite matrix ) from description 2 , respectively .the central receiver uses both descriptions to generate a reproduction of @xmath85 $ ] with the error covariance meeting a distortion constraint @xmath301 ( an @xmath25 positive definite matrix ) from both descriptions .this situation is closely related to the two description problem and we can harness our results thus far to completely characterize the rate region of the problem at hand .the rate region of two description with separate distortion constraints is @xmath375 where @xmath376 is defined as @xmath377 it is clear that any rate pair @xmath378 for some @xmath379 is in the rate region for the two description with separate distortion constraints , and so @xmath380 on the other hand , although receiver 1 ( 2 ) is only interested in reconstructing @xmath381 ( @xmath382 ) , they can actually reconstruct the entire source @xmath48 based on their received descriptions .hence , any coding scheme for the two description with separate distortion constraints will result in some achievable distortions @xmath383 with @xmath384 and @xmath385 . thus any rate pair @xmath386 achieved by this coding scheme is in the rate region @xmath387 for the two description problem .thus @xmath388 from equivalence of the two regions in , the proof is now complete .consider the general gaussian multiple description problem with source covariance @xmath24 and @xmath389 distortion constraints @xmath390 for each @xmath104 .following arguments similar to that used in arriving at the lower bound for sum rate , we have an outer bound on the rate region : @xmath391 following arguments similar to those used in arriving at the upper bound for the sum rate , we can use a jointly gaussian description scheme with covariance matrix of @xmath125 s ( @xmath50 ) taking the form , any tuple @xmath60 satisfying @xmath392+\mathbf{k}_z\big| } { \big|{{\text{cov}}}[{{{\mathbf{x}}}}|{{{\mathbf{u}}}}_l,\;l \in s]\big|\prod\limits_{l \in s}\big|{{\text{cov}}}[{{{\mathbf{x}}}}|{{{\mathbf{u}}}}_l]+\mathbf{k}_z\big|},\quad \forall s \subseteq \{1 , \ ; \dots , \ ; l\ } \end{array } \right\}\ ] ] is achievable .thus if we can find a @xmath50 of the form in such that all of the @xmath389 distortion constraints are met with equality , i.e. , @xmath393 = [ { { \mathbf{k}}}_x^{-1 } + ( { { \mathbf{i}}},\ ; \dots , \ ; { { \mathbf{i}}})\mathbf{k}_{w_s}^{-1}({{\mathbf{i}}},\ ; \dots , \ ; { { \mathbf{i}}})^t]^{-1 } , \quad \forall s \subseteq \{1 , \ ; \dots , \ ; l\},\ ] ] where @xmath62 is the covariance matrix for all @xmath394 , then the achievable region matches the outer bound and we would have characterized the rate region of the multiple description problem . from the above discussion , we see that for some choice of distortion constraints of the multiple description problem , we can indeed do this : first choose @xmath3 distortions @xmath395 such that they satisfy the condition for theorem [ th : main ] for the multiple description problem with individual and central receivers .next we can solve for the @xmath50 which is the covariance matrix of @xmath396 for the sum - rate - achieving jointly gaussian description scheme .for any other @xmath397 , this scheme results in distortion @xmath398^{-1}$ ] .finally we choose these @xmath399 s as the other distortion constraints .now we have a general multiple description problem with @xmath389 distortion constraints @xmath390 for each @xmath104 , and hence we can find a @xmath50 of form such that all of the @xmath389 distortion constraints are met with equality . thus is actually the rate region and it can be achieved by a jointly gaussian description scheme .in this appendix we provide some useful results in matrix analysis that are extensively used in this paper .* theorem 2.5 ) let @xmath142 be an @xmath400 nonsingular matrix and @xmath401 be an @xmath402 nonsingular matrix and let @xmath403 and @xmath404 be @xmath405 and @xmath406 matrices , respectively . if the matrix @xmath407 is nonsingular , then @xmath408 [ lemma : blockinverse ] ( * ? ? ?* theorem 2.3 ) suppose that the partitioned matrix @xmath409 is invertible and that the inverse is conformally partitioned as @xmath410 if @xmath142 is a nonsingular principal sub - matrix of @xmath411 , then @xmath412 ( * ? ? ?* theorem 6.13 ) let @xmath413 be a positive definite matrix and let @xmath414 be an @xmath406 matrix .then for any @xmath415 positive definite matrix @xmath416 , @xmath417 ( * ? ? ?* theorem 6.8 and 6.9 ) let @xmath142 and @xmath401 be positive definite matrices such that @xmath418 .then , @xmath419define an i.i.d . random process @xmath420\}$ ] , @xmath87 of @xmath123 gaussian random vectors , where @xmath421 $ ] , @xmath87 are independent of @xmath36 and @xmath422 , @xmath423 .form a random process @xmath424,\;\dots,\;{{{\mathbf{y}}}}[n])^t$ ] by @xmath425 = { { { \mathbf{x}}}}[m]+{{{\mathbf{z}}}}[m ] , \quad m=1,\ ; \dots,\ ; n.\ ] ] it follows that @xmath426\}$ ] is an i.i.d . random process of @xmath427 gaussian random vectors , where @xmath428 .then @xmath429 since @xmath36 and @xmath430 are gaussian vectors , for the first two terms in , we have @xmath431 we also have the following bound on @xmath432 for @xmath90 : @xmath433|c_l ) \\ & \le \sum\limits_{m=1}^n \frac{1}{2}\log(2\pi e)^n\big|{{\text{cov}}}[{{{\mathbf{y}}}}[m]|c_l]\big| \\ & \le \frac{1}{2}\log(2\pi e)^{nn}+\frac{n}{2}\log\left|\frac{1}{n}\sum\limits_{m=1}^n { { \text{cov}}}[{{{\mathbf{y}}}}[m]|c_l]\right| \\ & = \frac{1}{2}\log(2\pi e)^{nn}+\frac{n}{2}\log\left|\frac{1}{n}\sum\limits_{m=1}^n { { \text{cov}}}[({{{\mathbf{x}}}}[m]+{{{\mathbf{z}}}}[m])|c_l]\right| \\ & = \frac{1}{2}\log(2\pi e)^{nn}+\frac{n}{2}\log\left|\frac{1}{n}\sum\limits_{m=1}^n { { \text{cov}}}[{{{\mathbf{x}}}}[m]|c_l]+{{\mathbf{k}}}_z\right| \\ & \le \frac{1}{2}\log(2\pi e)^{nn}+\frac{n}{2}\log\left|{{\mathbf{d}}}_l+{{\mathbf{k}}}_z\right| \\ & = \frac{1}{2}\log(2\pi e)^{nn}\left|{{\mathbf{d}}}_l+{{\mathbf{k}}}_z\right|^n . \end{split}\ ] ] next we bound the last two terms of as follows .@xmath434 letting @xmath435 \stackrel{\rm def}{= } { { \text{cov}}}[{{{\mathbf{x}}}}[m]-\hat{{{{\mathbf{x}}}}}_0[m]],\label{eq : covdefn}\ ] ] we have @xmath436)-h({{{\mathbf{z}}}}[m]|{{{\mathbf{z}}}}[1],\;\dots,\;{{{\mathbf{z}}}}[m-1],{{{\mathbf{y}}}}^n-\hat{{{{\mathbf{x}}}}}_0^n)\big ) \\ & \ge \sum\limits_{m=1}^n \big(h({{{\mathbf{z}}}}[m])-h({{{\mathbf{z}}}}[m]|{{{\mathbf{y}}}}[m]-\hat{{{{\mathbf{x}}}}}_0[m])\big ) \\ & = \sum\limits_{m=1}^ni({{{\mathbf{z}}}}[m];{{{\mathbf{x}}}}[m]-\hat{{{{\mathbf{x}}}}}_0[m]+{{{\mathbf{z}}}}[m ] ) \\ & \overset{(a)}{\ge } \sum\limits_{m=1}^n\frac{1}{2}\log\frac{|{{\mathbf{k}}}_c[m]+{{\mathbf{k}}}_z[m]|}{|{{\mathbf{k}}}_c[m]| } \\ & \overset{(b)}{\ge}\frac{n}{2}\log\frac{|{{\mathbf{d}}}_0+{{\mathbf{k}}}_z|}{|{{\mathbf{d}}}_0| } , \end{split}\ ] ] where ( a ) is from and ( * ? ? ?* lemma ii.2 ) .the justfication for ( b ) is from the convexity of @xmath437 in @xmath438 and . fromand we have @xmath439 combining , and , we have @xmath440 taking the supremum over all positive definite @xmath102 , we can sharpen the lower bound in : @xmath441 first we assume @xmath456 , and hence @xmath457^{-1 } \\= & { { \mathbf{a}}}- { { \mathbf{a}}}\left({{\mathbf{i}}}\ ; { { \mathbf{i}}}\ ; \dots { { \mathbf{i}}}\right)\left[{{\mathbf{k}}}_w+\left({{\mathbf{i}}}\ ; { { \mathbf{i}}}\ ; \dots \ ; { { \mathbf{i}}}\right)^t { { \mathbf{a}}}\left({{\mathbf{i}}}\ ; { { \mathbf{i}}}\ ; \dots \ ; { { \mathbf{i}}}\right)\right]^{-1}\left({{\mathbf{i}}}\ ; { { \mathbf{i}}}\ ; \dots \ ; { { \mathbf{i}}}\right)^t { { \mathbf{a}}}\\ = & { { \mathbf{a}}}- { { \mathbf{a}}}\left({{\mathbf{i}}}\ ; { { \mathbf{i}}}\ ; \dots { { \mathbf{i}}}\right ) \big[{{\text{diag}}}\{{{\mathbf{k}}}_{w_1}+{{\mathbf{a}}},\ ; { { \mathbf{k}}}_{w_2}+{{\mathbf{a } } } , \ ; \dots \ ; { { \mathbf{k}}}_{w_l}+{{\mathbf{a}}}\ } \big]^{-1}\left({{\mathbf{i}}}\ ; { { \mathbf{i}}}\ ; \dots { { \mathbf{i}}}\right)^t { { \mathbf{a}}}\\ = & { { \mathbf{a}}}- { { \mathbf{a}}}\sum\limits_{l=1}^l [ { { \mathbf{k}}}_{w_l}+{{\mathbf{a}}}]^{-1 } { { \mathbf{a}}}. \\ \end{split}\ ] ] thus , @xmath458^{-1}-{{\mathbf{a}}}^{-1 }\\ = & { { \mathbf{a}}}^{-1 } - { { \mathbf{a}}}^{-1 } { { \mathbf{a}}}\left [ -\left(\sum\limits_{l=1}^l ( { { \mathbf{k}}}_{w_l}+{{\mathbf{a}}})^{-1}\right)^{-1 } + { { \mathbf{a}}}{{\mathbf{a}}}^{-1}{{\mathbf{a}}}\right]^{-1}{{\mathbf{a}}}{{\mathbf{a}}}^{-1 } - { { \mathbf{a}}}^{-1 } \\ = & \left[\left(\sum\limits_{l=1}^l ( { { \mathbf{k}}}_{w_l}+{{\mathbf{a}}})^{-1}\right)^{-1}-{{\mathbf{a}}}\right]^{-1}. \end{split}\ ] ] when @xmath142 is singular , we can choose @xmath260 such that @xmath459 for @xmath460 , and thus we can apply the previous argument and let @xmath100 in the end .we use induction .first consider the matrix @xmath461 we have @xmath462 where ( a ) is from .next we define @xmath463 and suppose @xmath464 for @xmath465 . then @xmath466^{-1}{{\mathbf{a}}}\\ \ ; \longleftrightarrow \ ; & { { \mathbf{k}}}_{w_l}+{{\mathbf{a}}}\succ { { \mathbf{a}}}\left[\left(\sum\limits_{k=1}^{l-1}({{\mathbf{k}}}_{w_k}+{{\mathbf{a}}})^{-1}\right)^{-1}-{{\mathbf{a}}}\right]^{-1}{{\mathbf{a}}}+{{\mathbf{a}}}\\ \ ; \longleftrightarrow \ ; & ( { { \mathbf{k}}}_{w_l}+{{\mathbf{a}}})^{-1 } \prec \left[{{\mathbf{a}}}\left[\left(\sum\limits_{k=1}^{l-1}({{\mathbf{k}}}_{w_k}+{{\mathbf{a}}})^{-1}\right)^{-1}-{{\mathbf{a}}}\right]^{-1}{{\mathbf{a}}}+{{\mathbf{a}}}\right]^{-1 } \\ \ ; \longleftrightarrow \ ; & ( { { \mathbf{k}}}_{w_l}+{{\mathbf{a}}})^{-1 } \prec { { \mathbf{a}}}^{-1 } - \left[\left(\sum\limits_{k=1}^{l-1}({{\mathbf{k}}}_{w_k}+{{\mathbf{a}}})^{-1}\right)^{-1}-{{\mathbf{a}}}+{{\mathbf{a}}}\right]^{-1 } \\ \ ; \longleftrightarrow \ ; & ( { { \mathbf{k}}}_{w_l}+{{\mathbf{a}}})^{-1 } \prec { { \mathbf{a}}}^{-1 } - \sum\limits_{k=1}^{l-1}({{\mathbf{k}}}_{w_k}+{{\mathbf{a}}})^{-1 } \\ \ ; { \longleftarrow } \ ; & \sum\limits_{k=1}^{l}({{\mathbf{k}}}_{w_k}+{{\mathbf{a}}})^{-1 } \prec { { \mathbf{a}}}^{-1 } \\ \ ; \overset{(b)}{\longleftrightarrow } \ ; & ( { { \mathbf{k}}}_{w_0}+{{\mathbf{a}}})^{-1 } \prec { { \mathbf{a}}}^{-1 } \\ \ ; \longleftrightarrow \ ; & { { \mathbf{k}}}_{w_0}+{{\mathbf{a}}}\succ { { \mathbf{a}}}\\ \ ; \longleftrightarrow \ ; & { { \mathbf{k}}}_{w_0 } \succ 0 , \end{split}\ ] ] where ( b ) is from .@xmath470^{-1 } & = \left[({{\mathbf{k}}}_{w_0}+{{\mathbf{a}}}^*)^{-1}({{\mathbf{i}}}+ ( { { \mathbf{k}}}_{w_0}+{{\mathbf{a}}}^*){{\mathbf{\lambda}}}_1)\right]^{-1 } \\ & \overset{(a)}{= } ( { { \mathbf{i}}}+{{\mathbf{k}}}_{w_0}{{\mathbf{\lambda}}}_1)^{-1}({{\mathbf{k}}}_{w_0}+{{\mathbf{a}}}^ * ) \\ & = ( { { \mathbf{i}}}+{{\mathbf{k}}}_{w_0}{{\mathbf{\lambda}}}_1)^{-1}({{\mathbf{k}}}_{w_0}+{{\mathbf{a}}}^ * - ( { { \mathbf{i}}}+{{\mathbf{k}}}_{w_0}{{\mathbf{\lambda}}}_1){{\mathbf{a}}}^*)+{{\mathbf{a}}}^ * \\ & \overset{(b)}{= } ( { { \mathbf{i}}}+{{\mathbf{k}}}_{w_0}{{\mathbf{\lambda}}}_1)^{-1}{{\mathbf{k}}}_{w_0}+{{\mathbf{a}}}^ * \\ & = \left({{\mathbf{k}}}_{w_0}^{-1}({{\mathbf{i}}}+{{\mathbf{k}}}_{w_0}{{\mathbf{\lambda}}}_1)\right)^{-1 } + { { \mathbf{a}}}^ * \\ & = \left({{\mathbf{k}}}_{w_0}^{-1}+{{\mathbf{\lambda}}}_1\right)^{-1 } + { { \mathbf{a}}}^ * , \end{split}\ ] ] where ( a ) and ( b ) are from @xmath471 .@xmath472 where ( c ) is from @xmath471 .conditioned on @xmath442 , the collection of random variables @xmath443 are jointly gaussian and thus we have latexmath:[\[\sum\limits_{l=1}^lh({{{\mathbf{u}}}}_l|{{{\mathbf{y}}}})-h({{{\mathbf{u}}}}_1,\;\dots,\;{{{\mathbf{u}}}}_l|{{{\mathbf{y } } } } ) = \frac{1}{2}\log\frac{\prod\limits_{l=1}^l from mmse of @xmath56 from @xmath442 we have @xmath445 = { { \mathbf{k}}}_x+\mathbf{k}_{w_l}-{{\mathbf{k}}}_x({{\mathbf{k}}}_x+\mathbf{k}_z)^{-1}{{\mathbf{k}}}_x , \quad l=1,\;\dots,\ ; l\ ] ] and @xmath446 where @xmath447 is an @xmath448 matrix of all ones and @xmath449 is the kronecker product . by fischer inequality (the block matrix version of hadamard inequality , see ( * ? ? ? * theorem 6.10 ) ) we know that @xmath450| } = { \big|{{\text{cov}}}[{{{\mathbf{u}}}}_1,\;\dots,\;{{{\mathbf{u}}}}_l|{{{\mathbf{y}}}}]\big|}$ ] if and only if the off - diagonal block matrices of @xmath451 $ ] are all zero matrices .thus we have @xmath452 if and only if @xmath453 or equivalently , if and only if @xmath454 to get a valid @xmath455 , we need the additional condition @xmath132 .first we prove that @xmath467 we have @xmath75^{\frac{1}{2}}({{\mathbf{k}}}_{w_1}-{{\mathbf{k}}}_{w_0})^{\frac{1}{2 } } -{{\mathbf{k}}}_{w_0}.\ ] ] thus @xmath468^{\frac{1}{2}}({{\mathbf{k}}}_{w_1}-{{\mathbf{k}}}_{w_0})^{\frac{1}{2 } } \succ { { \mathbf{k}}}_{w_0 } \\ \ ; \longleftrightarrow \ ; & \left[({{\mathbf{k}}}_{w_1}-{{\mathbf{k}}}_{w_0})^{-\frac{1}{2 } } ( { { \mathbf{k}}}_{w_2}-{{\mathbf{k}}}_{w_0})({{\mathbf{k}}}_{w_1}-{{\mathbf{k}}}_{w_0})^{-\frac{1}{2}}\right]^{\frac{1}{2 } } \succ ( { { \mathbf{k}}}_{w_1}-{{\mathbf{k}}}_{w_0})^{-\frac{1}{2}}{{\mathbf{k}}}_{w_0}({{\mathbf{k}}}_{w_1}-{{\mathbf{k}}}_{w_0})^{-\frac{1}{2 } } \\ \ ; { \longleftarrow } \ ; & ( { { \mathbf{k}}}_{w_1}-{{\mathbf{k}}}_{w_0})^{-\frac{1}{2 } } ( { { \mathbf{k}}}_{w_2}-{{\mathbf{k}}}_{w_0})({{\mathbf{k}}}_{w_1}-{{\mathbf{k}}}_{w_0})^{-\frac{1}{2 } } \\ & \quad \succ ( { { \mathbf{k}}}_{w_1}-{{\mathbf{k}}}_{w_0})^{-\frac{1}{2}}{{\mathbf{k}}}_{w_0}({{\mathbf{k}}}_{w_1}-{{\mathbf{k}}}_{w_0})^{-1}{{\mathbf{k}}}_{w_0}({{\mathbf{k}}}_{w_1}-{{\mathbf{k}}}_{w_0})^{-\frac{1}{2 } } \\ \ ; \longleftrightarrow \ ; & { { \mathbf{k}}}_{w_2}-{{\mathbf{k}}}_{w_0 } \succ { { \mathbf{k}}}_{w_0}\left({{\mathbf{k}}}_{w_1}-{{\mathbf{k}}}_{w_0}\right)^{-1}{{\mathbf{k}}}_{w_0 } \\ \ ; \longleftrightarrow \ ; & { { \mathbf{k}}}_{w_2}-{{\mathbf{k}}}_{w_0 } \succ { { \mathbf{k}}}_{w_0}\left(-{{\mathbf{i}}}+ ( { { \mathbf{k}}}_{w_1}-{{\mathbf{k}}}_{w_0})^{-1}\right){{\mathbf{k}}}_{w_1 } \\ \ ; \longleftrightarrow \ ; & { { \mathbf{k}}}_{w_2}-{{\mathbf{k}}}_{w_0 } \succ -{{\mathbf{k}}}_{w_0 } + { { \mathbf{k}}}_{w_0}({{\mathbf{k}}}_{w_1}-{{\mathbf{k}}}_{w_0})^{-1}{{\mathbf{k}}}_{w_1 } \\ \ ; \longleftrightarrow \ ; & { { \mathbf{k}}}_{w_2 } \succ { { \mathbf{k}}}_{w_0}({{\mathbf{k}}}_{w_1}-{{\mathbf{k}}}_{w_0})^{-1}{{\mathbf{k}}}_{w_1 } \\ \ ; \longleftrightarrow \ ; & { { \mathbf{k}}}_{w_2 } \succ { { \mathbf{k}}}_{w_0}{{\mathbf{k}}}_{w_0}^{-1}({{\mathbf{k}}}_{w_0}^{-1}-{{\mathbf{k}}}_{w_1}^{-1})^{-1}{{\mathbf{k}}}_{w_1}^{-1}{{\mathbf{k}}}_{w_1 } \\ \ ; \longleftrightarrow \ ; & { { \mathbf{k}}}_{w_2 } \succ ( { { \mathbf{k}}}_{w_0}^{-1}-{{\mathbf{k}}}_{w_1}^{-1})^{-1 } \\ \ ; \longleftrightarrow \ ; & { { \mathbf{k}}}_{w_1}^{-1 } + { { \mathbf{k}}}_{w_2}^{-1 } \prec { { \mathbf{k}}}_{w_0}^{-1 } \\ \ ; \longleftrightarrow \ ; & { { \mathbf{d}}}_0^{-1 } + { { \mathbf{k}}}_x^{-1 } - { { \mathbf{d}}}_1^{-1}-{{\mathbf{d}}}_2^{-1 } \succ \mathbf{0}. \end{split}\ ] ] the proof of @xmath469 is similar and hence is omitted .we first prove the following lemma .[ lemma : depsilon ] let @xmath473 be an @xmath25 matrix such that @xmath474 .let @xmath475 .choose @xmath476 such that @xmath477 .define @xmath478^{-1}.\ ] ] then , there exist constants @xmath479 , such that @xmath480 there exists an @xmath25 orthogonal matrix @xmath20 such that @xmath481 where @xmath482 are eigenvalues of @xmath483 .we have @xmath484 and @xmath485^{-1 } { { \mathbf{q}}}^t \\ & = \big[({{\text{diag}}}\{k_1,\;\dots,\;k_n\}-\epsilon { { \mathbf{i}}})^{-1}+{{\mathbf{i}}}\big]^{-1 } \\ &= { { \text{diag}}}\left\{\frac{k_1-\epsilon}{1+k_1-\epsilon},\;\dots,\;\frac{k_n-\epsilon}{1+k_n-\epsilon}\right\}\\ & = { { \text{diag}}}\left\{\frac{k_1}{1+k_1}-\frac{\epsilon}{(1+k_1)^2}+o(\epsilon ) , \;\dots,\;\frac{k_n}{1+k_n}-\frac{\epsilon}{(1+k_n)^2}+o(\epsilon ) \right\}. \end{split}\ ] ] we now have @xmath486 where @xmath487 are some constants . hence @xmath488 equations and are a direct consequence of this lemma .we first prove the following lemma .[ lemma : kzepsilon ] let @xmath142 be an @xmath25 matrix such that @xmath489 .let @xmath490 .choose @xmath476 such that @xmath491 .define @xmath492^{-1}-{{\mathbf{i}}}.\ ] ] then , there exist constants @xmath493 such that @xmath494 there exists an @xmath25 orthogonal matrix @xmath20 such that @xmath495 where @xmath496 are the eigenvalues of @xmath142 .we have @xmath497 and @xmath498 we now have @xmath499 where @xmath500 are some constants .hence @xmath501 equation is a direct result of this lemma .we first prove the following lemma .[ lemma : kzinf ] let @xmath142 be an @xmath25 matrix such that @xmath502 . choose @xmath476 such that @xmath503 . define @xmath504^{-1}-{{\mathbf{i}}}.\] ] then , for any @xmath505 and @xmath414 such that @xmath506 and @xmath507 , we have @xmath508 using this lemma , we can show a property similar to as @xmath326 approaches zero .first note that similar to case 2 , we have @xmath519 where @xmath520 and @xmath521 are constants .hence we have @xmath522 similarly , we have @xmath523 thus @xmath524 where the last step is similar to .r. puri , s. s. pradhan , and k. ramchandran , `` n - channel symmetric multiple descroption - part ii : an achievable rate - distortion region , '' _ ieee trans .inform . theory _51 , no . 4 , pp .13771392 , apr .2005 .v. a. vaishampayan , n. sloane , and s. servetto , `` multiple description vector quantizers with lattice codebooks : design and analysis , '' _ ieee trans .inform . theory _47 , no . 4 , pp .17181734 , july 2001 .d. tse and s. hanly , `` multi - access fading channels : part i : polymatroid structure , optimal resource allocation and throughput capacities , '' _ ieee trans .inform . theory _44 , no . 7 , pp . 27962815 , nov .p. viswanath , `` sum rate of a class of gaussian multiterminal source coding problems , '' in _ advances in network information theory _ , p. gupta , g. kramer and a. wijngaarden editors , rutgers , nj , 2004 , pp . 4360 .

<S> @xmath0 multiple descriptions of a vector gaussian source for individual and central receivers are investigated . </S><S> the sum rate of the descriptions with covariance distortion measure constraints , in a positive semidefinite ordering , is exactly characterized . for two descriptions , </S><S> the entire rate region is characterized . </S><S> jointly gaussian descriptions are optimal in achieving the limiting rates . </S><S> the key component of the solution is a novel information - theoretic inequality that is used to lower bound the achievable multiple description rates . _ </S><S> submitted to the ieee transactions on information theory , oct . 2005 _ </S><S> 1.3ex * vector gaussian multiple description with individual and central receivers * + hua wang and pramod viswanath </S>

1206.6353

fomalhaut ( @xmath6 psa , hd 216956 , hip 113368 ) is a famous nearby a3v star with a large resolved dusty debris disk ( e.g. * ? ? ?* ) and an imaged candidate extrasolar planet @xcite .the age of fomalhaut is mainly of interest for predicting the infrared brightnesses of substellar companions @xcite , calculations of the total mass of the parent bodies generating the dust @xcite , and placing the dusty debris disk in evolutionary context with other stars ( e.g. * ? ? ?* ) . in general , accurate ages for host stars of substellar objects are useful for constraining not only the masses of the companions , but accurate ages forthe youngest stars may help constrain the initial conditions for the substellar objects @xcite .@xcite recently announced the discovery of a faint optical companion ( likely @xmath73 @xmath8 ) at separation 12.7 ( 96 au ) from fomalhaut . while the companion has been imaged multiple times at optical wavelengths , ( ** kalas et al . , in prep ) , it has eluded detection in the infrared @xcite . given the importance of fomalhaut as a benchmark resolved debris disk system and possible planetary system , a detailed reassessment of its age is long overdue .this paper is split into the following sections : 1 ) a review of published age estimates for fomalhaut , 2 ) estimation of a modern isochronal age for fomalhaut , 3 ) demonstration of the physicality of the fomalhaut - tw psa binary system , 4 ) age estimates for tw psa based on multiple calibrations , and 5 ) estimation of a consensus age for the fomalhaut - tw psa system .these results supersede the age analysis for the fomalhaut - tw psa system presented at the 2010 spirit of lyot meeting in paris @xcite .ccccc 200@xmath0100 & 1 & isochrones ( fom ) , li , x - ray , rotation ( tw ) + 224@xmath9 & 2 & isochrones ( fom ) + 156@xmath10 & 3 & isochrones ( fom ) + 290 & 4 & isochrones ( fom ) + 480 & 5 & isochrones ( fom ) + 220 & 6 & `` fomalhaut '' + 419@xmath031 & 7 & isochrones ( fom )previously published ages for fomalhaut and tw psa are listed in table [ tab : ages ] , and span a factor of 3 , from 156 myr @xcite to 480 myr @xcite .the most often cited age for fomalhaut is 200 @xmath0 100 myr from @xcite and @xcite .the @xcite estimate comes from multiple age indicators ( including li abundance , rotation , hr diagram position , and x - ray emission ) for its purported common proper motion companion tw psa .later , @xcite assigned the same age to fomalhaut based on its purported membership to the castor moving group ( cmg ; which included castor , vega , and roughly a dozen other systems ) .these analyses relied heavily on a few assumptions , worth reexamining namely that fomalhaut and tw psa are physically related , that the castor group is physical ( i.e. useful for age - dating ) , and that fomalhaut and tw psa belong to the castor group . the question of whether the cmg is actually useful for age - dating will await a future investigation . for this study ,i focus solely on the ages of fomalhaut and tw psa , and assess the physicality of that binary .an isochronal age for fomalhaut can be estimated through comparing its @xmath11 and luminosity to modern evolutionary tracks .the stellar parameters for fomalhaut are fairly well determined due to its brightness and proximity , which has enabled the star to have its diameter measured interferometrically . herei estimate refined hr diagram parameters for fomalhaut and estimate an isochronal age .basic stellar parameters for fomalhaut are listed in table [ tab : data ] .@xcite estimated the bolometric flux of fomalhaut to be [email protected] @xmath1210@xmath13 wm@xmath14 , which i adopt . combining this with the revised hipparcos parallax from @xcite of @xmath15 = [email protected] mas ( @xmath16 = [email protected] pc ) , this results in a bolometric luminosity for fomalhaut of [email protected] l@xmath5 or log(@xmath17 ) = [email protected] dex = 3.8270(@xmath00.0014 ) @xmath12 10@xmath18 erg / s based on the total solar irradiance ( tsi ) of @xmath19 = 1360.8(@xmath00.5 ) wm@xmath14 @xcite calibrated to the nist radiant power scale , and the iau 2009 value for the astronomical unit ( 149597870700@xmath03 m ) . using the bolometric magnitude zero - point proposed by iau commissions 25 and 36 of l = 3.055 @xmath1210@xmath20 w ,this translates to a solar absolute bolometric magnitude of m@xmath21 = 4.7554(@xmath00.0004 ) mag on that scale . to force the recent tsi measurement to a scale where m@xmath21 = 4.75 ( a commonly adopted value ; ** ) , the zero - point luminosity could be adjusted to 3.040@xmath1210@xmath20 w. one can calculate a modern @xmath11 for the sun by combining the new luminosity with the solar radius ( 695660 km ) from ( * ? ? ?* where i adopt @xmath0100 km error based on their discussion ) .the resultant solar @xmath11 is [email protected] k. ] .@xcite resolved a small amount of k - band excess due to circumstellar dust , and reported a revised limb - darkened diameter taking into account all the available vlti data : @xmath22 = [email protected] mas . using the relation from @xcite ( @xmath11 = ( 2341 k ) @xmath12 @xmath23 , where @xmath24 is in units of 10@xmath25erg cm@xmath14 s@xmath4 and @xmath22 is in mas ) with the bolometric flux from @xcite and the limb - darkened diameter from @xcite , i derive a new @xmath11 of 8590@xmath073 k and radius [email protected] r@xmath5 .the @xmath11 is only slightly lower than recent estimates ( e.g. * ? ? ?lccccc @xmath27 & 344.411773 & 344.099277 & deg & 1 + @xmath28 & -29.621837 & -31.565179 & deg & 1 + parallax & [email protected] & [email protected] & mas & 1 + distance & [email protected] & [email protected] & pc & 1 + @xmath29 & [email protected] & [email protected] & masyr@xmath4 & 1 + @xmath30 & [email protected] & [email protected] & masyr@xmath4 & 1 + @xmath31 & [email protected] & [email protected] & kms@xmath4 & 2,3 + m@xmath32 & [email protected] & [email protected] & mag & 4 + period & ... & 10.3 & day & 5 + spec .type & a3 va & k4 ve & ... & 6,7 + @xmath11 & 8590@xmath073 k & 4594@xmath080 & k & 8,9 + @xmath33 & [email protected] & 0.10075 & nw m@xmath14 & 10,9 + u & [email protected] & [email protected] & kms@xmath4 & 8 + v & [email protected] & [email protected] & kms@xmath4 & 8 + w & [email protected] & [email protected] & kms@xmath4 & 8 + s@xmath34 & [email protected] & [email protected] & kms@xmath4 & 8 + x@xmath35 & 3.06 & 3.14 & pc & 8 + y@xmath35 & 1.14 & 0.90 & pc & 8 + z@xmath35 & -6.98 & -6.88 & pc & 8 + mass & [email protected] & 0.73@xmath36 & @xmath37 & 8,9 + m@xmath32 & [email protected] & [email protected] & mag & 8 + luminosity & [email protected] & [email protected] & @xmath38 & 8 + log(@xmath17 ) & [email protected] & [email protected] & dex & 8 + radius & [email protected] & [email protected] & @xmath39 & 8 + to calculate an isochronal age , i overlay the new hr diagram point for fomalhaut on the evolutionary tracks of @xcite ( fig .[ hrd ] , top ) .i assume that fomalhaut has a chemical composition similar to the proto - sun , with an asteroseismically - motivated ( and diffusion corrected ) composition of y = 0.27 and z = 0.017 ( see * ? ? ?* and references therein ) .i generate hr diagram positions by monte carlo sampling the bolometric flux , parallax , and limb - darkened radius values from their quoted values and uncertainties ( assuming a normal distribution ) , and interpolating within the @xcite tracks .the @xmath11 and luminosity of fomalhaut are consistent with a mass of [email protected] @xmath37 and age of 433@xmath036 myr .the uncertainties only take into account observational errors , and not systematic uncertainties in chemical composition and input physics . to estimate systematic uncertainties due to different assumed solar composition and input physics, i estimate calculate ages and masses for 3 more sets of tracks ( see table [ tab : ages_iso ] ) .the expectation ages for the 4 sets of tracks are very similar , and the mean of the ages from the 4 tracks is 450 myr with a 22 myr ( 5% ) rms scatter ( which is a reasonable estimate of the systematic error considering slightly different assumed protosolar abundances and input physics ) .considering the typical observational error in age ( @xmath033 myr ; 7% ) , this suggests a total isochronal age uncertainty of @xmath040 myr ( 9% ) .for the four sets of tracks , the average mass is 1.923 @xmath37 , with @xmath00.014 @xmath37 rms ( systematic error component ) and @xmath00.016 @xmath37 scatter due to the observational errors .note that this new mass ( [email protected] @xmath37 ) is similar to previous estimates ( e.g. * ? ? ?* ) , but 16% lower than the 2.3 @xmath37 quoted by @xcite .the new estimated mass is in line with observed trends in @xmath11 and log(@xmath17 ) vs. mass for main sequence stars in eclipsing binaries @xcite , which empirically predict @xmath401.9 - 2.0 @xmath37 .lllll @xcite & 0.270 & 0.017 & 444@xmath41 & [email protected] + @xcite & 0.273 & 0.019 & 427@xmath42 & 1.911@xmath43 + @xcite & 0.274 & 0.0189 & 478 & 1.92 + @xcite & 0.264 & 0.017 & 453@xmath44 & 1.943@xmath45 although overlooked in most recent literature on fomalhaut , the star has a likely stellar companion : tw psa ( gj 879 , hip 113283 ) . that tw psa and fomalhaut appear to share proper motion and parallax appears to have been first noticed by @xcite .tw psa is an active k4ve star @xcite , at a projected separation of [email protected] ( @xmath407100 ) , and has been listed as among the widest ( @xmath4050000 au ) candidate binaries known @xcite .the proximity of tw psa and its approximate co - motion with fomalhaut led to @xcite using tw psa to age - date fomalhaut .@xcite included fomalhaut and tw psa as a wide binary in their bayesian search for multiple systems in the hipparcos catalog , and the system was one of only two binaries with separations @xmath460.25 pc identified within 10 pc .given the utility of tw psa to age - dating fomalhaut , we should test the physicality of the purported binary system using the best available astrometry .the best available astrometric and radial velocity data for fomalhaut and tw psa are listed in table [ tab : data ] , and their degree of similarity is striking .i adopt the literature mean @xmath31 for fomalhaut from @xcite ( [email protected] kms@xmath4 ) .this should reflect center - of - mass motion , as the revised hipparcos astrometric analysis was able to statistically fit an unperturbed single - star solution for fomalhaut s astrometry @xcite .the astrometric acceleration from the original hipparcos reduction reported by @xcite is statistically insignificant ( 1.7@xmath47 ) , and likely spurious . for tw psa ,i adopt the @xmath31 from @xcite , who reported a mean @xmath31 for tw psa of 6.6 kms@xmath4 over 7 epochs over 3794 days .the star apparently showed remarkable stability , with a quoted rms of @xmath00.1 kms@xmath4 .i calculate the galactic velocity vectors u , v , w ( in the direction of galactic center , rotation , and north galactic pole , respectively ) from the astrometry and radial velocities of fomalhaut and tw psa .the 3d velocities are listed in table [ tab : data ] .the degree of similarity is embarrassingly good , as their velocities agree within @xmath00.1 kms@xmath4 in all three directions .the barycentric speeds of the star differ by only [email protected] kms@xmath4 .just how remarkable is the agreement in velocities ?i generated a catalog of 3d velocities for 34,817 stars with updated hipparcos astrometry from @xcite ( those with positive parallaxes ) and literature mean radial velocities from @xcite .the only stars with 3d velocities within 1 kms@xmath4 of fomalhaut are tw psa and hip 3800 ( 0.99 kms@xmath4 different ) .this suggests that @xmath4810@xmath49 of field stars have velocities within 1kms@xmath4 of fomalhaut s , and that the similarity in velocities between fomalhaut and tw psa ( separated by only @xmath400.3 pc ) is more than just a coincidence .what is the 3d separation of fomalhaut and tw psa ? taking the parallax values and uncertainties from @xcite ,i generate 10@xmath50 monte carlo realizations of the distances to fomalhaut and tw psa ( assuming @xmath16 = 1/parallax ) .fomalhaut has a parallax distance of [email protected] pc , while tw psa is at [email protected] pc .the 3d separations in the simulations have a median separation of 0.280@xmath2 pc ( 57.4@xmath3 kau ; 68%cl range quoted ) .many plausible binary systems are known with larger separations ( e.g. * ? ? ?assuming that the census of the nearest 100 star systems is complete , the local density of star systems is 0.085 pc@xmath51 .the chances of having an unrelated star ( system ) within 0.28 pc of a nearby star is approximately 1 in 130 .so the proximity in space and velocity between fomalhaut and tw psa appear to be more than coincidental , but are they bound ?the escape velocity from fomalhaut ( 1.92@xmath37 ) at the separation of tw psa is 0.21 kms@xmath4 , suggesting that the observed difference in velocities ( [email protected] kms@xmath4 ) is statistically consistent with the hypothesis of tw psa and fomalhaut constituting a bound pair . if one sets the semi - major axis of tw psa s orbit to be equal to the observed 3d separation , and adopt a tw psa mass of 0.73 @xmath37 @xcite , then one estimates an orbital period of @xmath408 myr .the predicted amplitude of the orbital velocities would be 0.06 kms@xmath4 for fomalhaut and 0.15 kms@xmath4 for tw psa .rotation rates among late - f through early - m dwarfs appear to spin down as they age through magnetically braking , approximately as rotation period @xmath52 age@xmath53 @xcite . using the rotation period from (* days ) and the gyrochronology curves from @xcite , and assuming @xmath01.1 day rms fit to the gyro relation , i estimate a gyrochronology age of 410@xmath080 myr .attempts to derive an isochronal age for tw psa were discussed by @xcite . herei use the @xmath11 and luminosity from table 2 to derive new estimates of isochronal ages from evolutionary tracks . using the pre - ms evolutionary tracks of @xcite , tw psa is consistent with having a mass of 0.71 @xmath37 and an age of 52 myr . using the @xcite pre - ms tracks , tw psa is consistent with having a mass of 0.72 @xmath37 and 66 myr . as discussed by @xcite in their review of tw psa s other youth diagnostics ( li , activity ), it is unlikely that the star is @xmath48100 myr .the spectroscopic surface gravity appears to be log(@xmath54 ) @xmath55 4.5 - 4.7 ( e.g. * ? ? ?* ) , again consistent with a main sequence star . in light of those findings ,i consider the star to be a young main sequence star , rather than pre - ms .hence , i consider the pre - ms isochronal age estimates to represent strict lower limits to tw psa s age ( i.e. @xmath4650 myr ) , rather than useful age estimates themselves .tw psa is also a coronal x - ray source , with l@xmath56 = 10@xmath57 ergs@xmath4 , and fractional x - ray luminosity of log(@xmath58/@xmath59 ) = -4.57 @xcite . although the star s color is slightly redder ( b - v = 1.1 ) than the range probed by the calibrations in @xcite , using the x - ray age relation from equation a3 of that paper , this log(@xmath58/@xmath59 ) value would be consistent with an age of @xmath40380 myr . given the scatter in x - ray luminosities among stars in clusters of similar mass ( @xmath400.4 dex ; * ?* ) , the age uncertainty is approximately @xmath60 myr . as pointed out by @xcite , tw psa shows detectable li ( ew(li i @xmath616707 ) = 33@xmath02 m ) consistent with an abundance of log @xmath62(li ) = 0.6 .the recent photometric teff from @xcite that i adopt ( @xmath11 = 4594@xmath080 k ) is not far from the @xmath11 originally adopted by @xcite ( 4500 k ) . @xcite plotted the li i abundances for members of 4 clusters of different ages ( their fig .2 ; pleiades , m34 , uma , hyades ) , and given its completeness , there is little reason to repeat the plot here . what _ has _ changed in the past 15 yearsis the age scale for these benchmark clusters .@xcite adopted the following age scale : pleiades - 85 myr , m34 - 200 myr , uma - 300 myr , hyades - 700 myr .more recent evolutionary tracks are leading to slightly older ages among the younger clusters .i adopt the following age scale : pleiades - 130 myr @xcite , 220 myr @xcite , uma - 500 myr @xcite , & hyades - 625 myr @xcite .the li abundance for tw psa appears to be intermediate between the hyades and pleiades , and m34 and uma .it is more li - poor than the pleiades and m34 stars ( hence @xmath46220 myr ) , but more li - rich than the uma stars and hyades ( hence @xmath48500 myr ) .based on the li abundances alone , i adopt an estimate of 360@xmath0140 myr .the three independent age estimates for tw psa listed in table [ tab : ages ] are consistent with a weighted mean age of 400@xmath070 myr .this age estimate for tw psa is independent of any genetic association with fomalhaut .ll 450@xmath040 & isochrones ( fomalhaut ) + 410@xmath080 & rotation ( tw psa ) + @xmath4650 & isochrones ( tw psa ) + @xmath40380@xmath60 & x - ray ( tw psa ) + 360@xmath0140 & lithium ( tw psa ) + * 440@xmath040 * & * final ( both ) *the kinematic data are consistent with fomalhaut and tw psa comoving within [email protected] kms@xmath4 , and separated by only 0.28 pc . given their coincidence in position , velocity , and statistical agreement in velocities expected for a wide bound binary , and remarkable agreement among independent age estimates ( @xmath4010% agreement ) , i conclude that fomalhaut and tw psa constitute a physical binary .therefore a cross - comparison of their ages is useful .the new age estimates for fomalhaut and tw psa are listed in table [ tab : ages_new ] .the new isochronal age for fomalhaut ( 450@xmath040 myr ) is in good agreement with two recent isochronal estimates : 480 myr @xcite and 419@xmath031 myr @xcite .it is clear that more modern evolutionary tracks and constraints on the hr diagram position of fomalhaut are leading to an age twice as old as the classic age ( 200 myr ; * ? ? ?[ hrd ] ( bottom ) shows a pleasing overlap between the inferred age probability distribution for fomalhaut ( using the @xcite tracks ) and the gyrochronology and li ages for tw psa ( the two estimates with the smallest uncertainties ) .based on the 4 independent ages in table [ tab : ages_new ] , the rounded weighted mean age for the fomalhaut - tw psa system is 440@xmath040 myr .this new estimate has relative uncertainties @xmath405@xmath12 smaller than the age quoted by @xcite and @xcite ( 200@xmath0100 myr ) , and is tied to the contemporary open cluster age scale and modern evolutionary tracks .a factor of 2@xmath12 older age for fomalhaut has consequences for the predicted brightnesses of substellar companions . using the @xcite evolutionary tracks, it appears that a factor of 2@xmath12 older age indicates that a given brightness limit at 4.5 @xmath63 m ( or m band ) corresponds to thermal emission from a planet roughly 2@xmath12 as massive if it were 200 myr .a 1 m@xmath64 planet of age 200 myr has absolute magnitude @xmath65 = 20.4 , but at 440 myr is approximately 1.2 magnitudes fainter ( @xmath65 = 21.6 ) .future searches for thermal emission from exoplanets orbiting fomalhaut should take into account this older age .eem acknowledges support from nsf award ast-1008908 , and thanks paul kalas , mark pecaut , erin scott , tiffany meshkat , and matt kenworthy for comments on the manuscript , and the referee david soderblom for a helpful review. , g. t. , lane , b. f. , thompson , r. r. , boden , a. f. , colavita , m. m. , dumont , p. j. , mobley , d. w. , palmer , d. , shao , m. , vasisht , g. x. , wallace , j. k. , creech - eakman , m. j. , koresko , c. d. , kulkarni , s. r. , pan , x. p. , & gubler , j. 1999 , , 117 , 521

<S> the nearby ( d = 7.7 pc ) a3v star fomalhaut is orbited by a resolved dusty debris disk and a controversial candidate extrasolar planet . </S><S> the commonly cited age for the system ( 200@xmath0100 myr ) from @xcite relied on a combination of isochronal age plus youth indicators for the k4v common proper motion system tw psa . </S><S> tw psa is [email protected] away from fomalhaut , and was first proposed as a companion by luyten ( 1938 ) , but the physicality of the binarity is worth testing with modern data . i demonstrate that tw psa is unequivocally a physical stellar companion to fomalhaut , with true separation 0.280@xmath2 pc ( 57.4@xmath3 kau ) and sharing velocities within [email protected] km s@xmath4 consistent with being a bound system . </S><S> hence , tw psa should be considered `` fomalhaut b '' . </S><S> combining modern hr diagram constraints with four sets of evolutionary tracks , and assuming the star was born with protosolar composition , i estimate a new isochronal age for fomalhaut of 450@xmath040 myr and mass of [email protected] m@xmath5 . </S><S> various stellar youth diagnostics are re - examined for tw psa . the star s rotation , x - ray emission , and li abundances are consistent with approximate ages of 410 , 380 , and 360 myr , respectively , yielding a weighted mean age of 400@xmath070 myr . combining the independent ages , </S><S> i estimate a mean age for the fomalhaut - tw psa binary of 440 @xmath0 40 myr . </S><S> the older age implies that substellar companions of a given mass are approximately one magnitude fainter at ir wavelengths than previously assumed . </S>

cond-mat9810067

quantum features of conductance in mesoscopic electronics is currently a major theoretical and experimental research interest in condensed matter physics@xcite .developments are driven by two complementary imperatives .firstly a technological trend to fabricate devices on smaller and smaller scales is rapidly approaching the point where quantum effects will become a problem unless explicit attempts to exploit quantum features are made .quantum tunnelling can lead to undesired coupling between fabricated structures . on the other hand , tunnelling offers the possibility of very fast switching times .secondly , the new devices require improvements to the theoretical description of electronic transport in a low temperature , high mobility regime .small devices with very long coherence times can be dominated by coherent quantum effects .it is becoming increasingly clear that intrinsic quantum fluctuations play an important role at low temperatures@xcite current noise in resonant tunnelling devices ( rtd ) provides a path to understanding noise in the deep quantum domain . in a biased rtd one or more bound quantum statesare coupled incoherently to two electron reservoirs maintained at different chemical potential .there are a number of experimental@xcite and theoretical@xcite results .rtds involve exchange of fermions between the reservoirs , and the bound states .we propose in this paper an approach to such devices based on quantum markov master equations@xcite .such an approach to quantum noise in nonequlibrium systems has been used with great success in quantum optics .this provides an alternative approach to the conventional green s functions methods , and offers additional physical insights .for example it enables one to deal with coherent coupling between adjacent well states which couples off - diagonal elements of the density matrix in the occupation number basis and can not be described by population master equations .such coupling can occur in strongly coupled quantum dots , as in the recent experiments of blick et al@xcite and oosterkamp et al@xcite .if the strength of this coherent coupling dominates the time scales of elastic and inelastic relaxation , a population master equation can not describe the system .coherently coupled nanostructures are likely to become increasingly important and thus there is considerable motivation to develop theoretical schemes that go beyond population rate equations . in the first part of this paper we derive the operator master equation to describe a bound electronic system coupled incoherently to two reservoirs .we then apply this equation to calculate the current two - time correlation function for a single well , with a single bound state .this model of course can equally well be treated by a population master equation approach as in the approach of carlos egues et al.@xcite , but we rederive the known results simply to display the method in a familiar context . in section iiiwe apply our methods to treat the case of coherent coupling between the bound states of adjacent wells . in this caseour approach yields results that go beyond the traditional population master equation approach .we derive the current spectrum in the device and demonstrate new features that arise precisely because of the coherent coupling between the two wells . to make contact with previous work we show that in the limit when elastic scattering dominates the coherent coupling , a population rate equation may be derived that is equivalent , in the appropriate limit , to that obtained by carlos egues et al.@xcite .we begin with the derivation of the master equation for a single quantum tunnelling channel connecting two reservoirs under external bias .this system is quite adequately described by other methods , including population master equations .however we treat it here simply to demonstrate our approach in a familiar setting .therefore our results are not new and could equally well be obtained by other methods .this is not the case for the coherently coupled double well system we discuss next .the hamiltonian describing this process is given by@xcite @xmath0 where @xmath1 and @xmath2 are the annihilation ( creation ) operators of electrons in the emitter ( @xmath3 ) reservoir , in the central quantum well and in the collector ( @xmath4 ) reservoir respectively .the energy of the bound state without bias is @xmath5 which under bias becomes @xmath6 where @xmath7 is a structure dependent coefficient .the single particle energies in the emitter and collector are respectively , @xmath8 and @xmath9 .the energy reference is at the bottom of the conduction band of the emitter reservoir .the fourth and fifth terms in the hamiltonian describe the coupling between the quasibound electrons in the well and the electrons in the reservoir .the tunnelling coefficients @xmath10 depend on the barrier profile and the bias voltage .we will assume that at all times the two reservoirs remain in thermal equilibrium , with chemical potentials @xmath11 , with @xmath12 , despite the tunnelling of electrons .this is one of the key defining characteristics of a reservoir .it assumes in effect that two very different time scales describe the dynamics of the reservoirs and the quaisbound quantum state in the well . in the interaction picturethe hamiltonian may be written as @xmath13 where the bound state frequency is @xmath14 and the reservoir operators are given by @xmath15 where @xmath16 we now obtain an equation of motion for the density operator of the bound state , @xmath17 , in the well following the standard method based on second order perturbation theory and tracing over reservoir states@xcite .thus we need @xmath18\ ] ] where @xmath19 is the thermal equilibrium state of the two reservoirs , and @xmath20 denotes a trace over the reservoir variables .note that the factorisation of the well state and reservoir states has been assumed .this is reasonable if the well state and the reservoir states are initially uncorrelated and provided there is a wide separation in the relaxation time scales of the well state and the reservoirs .the only non zero correlation functions we need to compute are @xmath21 in order to illustrate the important physical approximations required in deriving the master equation we will now explicitly evaluate the first of these correlation functions . using the definition of the reservoir operators and the assumed thermal fermi distribution of the electrons in the emitter we find @xmath22 as the reservoir is a large system by definition we can replace the sum over @xmath23 by an integral to obtain @xmath24 where we have changed the variable of time integration .the dominant term in the frequency integration will come from frequencies near @xmath25 as the time integration is significant at that point .we assume that the bias is such that the quasibound state is well below the fermi level in the emitter .this implies that near @xmath26 , the average occupation of the reservoir state is very close to unity .this is an effective low temperature approximation .now we make the first markov approximation .we assume that the function @xmath27 is slowly varying around @xmath26 , and thus the frequency integration will lead to a function which is a rapidly decaying function of time compared to dynamical time scales for the quasibound state .this implies that on time scales of interest in an experiment we can extend the upper limit of the time integration to infinity as a good approximation . in that case@xmath28 becomes time independent and may be approximated by @xmath29 which defines the effective rate @xmath30 of injection of electrons from the left reservoir ( the emitter ) into the quasibound state of the well .this rate will have a complicated dependence on the bias voltage through both @xmath25 and the coupling coefficients @xmath31 . in this paperwe do not address this issue .we simply seek the noise properties as a function of the rate constants .evaluating all the other correlation functions under similar assumptions we find that the quantum master equation for the density operator representing the well - state in interaction picture is given by @xmath32 @xmath30 and @xmath33 are constants determining the rate of injection of electrons from emitter into the well and from the well into the collector respectively .the rate constants can be determined by a self - consistent band calculation involving the bias voltage .two poisson processes shown in the master equations : the injection of electrons into the well described by the first term in the right hand and the emission of electrons out of the well by the second term , are conditioned by the rates @xmath34 and @xmath35 : @xmath36 where the average is taken with respect to the well state at any time t. the master equation [ master1 ] is diagonal in the occupation number representation .the mean occupation number @xmath37can therefore be determined easily from the rate equation @xmath38 however the occupation number of the well states is not directly measured in current experiments .the current noise is a fluctuation in classical stochastic processes .it is measured in the relatively high temperature reservoirs of the leads , well away from the well state , and the strong , fast electron - electron interactions in the reservoir establish the classical level of the observed variable .it is however conditioned on the underlying quantum stochastic process in the well , which is described by the master equation .we thus have the familiar problem of connecting the observed classical stochastic process to the quantum source of information in an open quantum system . in this problemwe proceed as follows .the current pulse in the emitter and collector may be determined from the ramo - shockley theorem @xcite . for a symmetric geometrythis takes the form , @xmath39 the connection to the quantum source is then made by equations ( [ poisson ] ) . using eq(16 ) , the average current is given by @xmath40 . in the steady state this is @xmath41 where @xmath42 and the subscript @xmath43 indicates the steady state .the fluctuations in the observed current , @xmath44 are quantified by the two - time correlation function : @xmath45 to relate these classical averages to the fundamental quantum processes occurring in the well we apply the theory of open quantum system @xcite to the present system and calculate the following correlation components with @xmath46@xcite : @xmath47 @xmath48 @xmath49 @xmath50 calculating the above correlation components using master equation with corresponding initial conditions , and substituting them together with the shot - noise component into equation [ eq - corr ] yield : @xmath51 thus the spectral density of current fluctuation in frequency domain is given by @xmath52 the current fano factor @xmath53 is defined as the ratio of current noise density over the full shot noise density , and for low frequencies ( @xmath54 ) : @xmath55 the shot noise is suppressed and reaches the minimum of 50 % in a symmetric structure with @xmath56 the result is same as those derived by chen and ting @xcite using non - equilibrium green s function method .the result can also be obtained by a classical master equation calculation @xcite .however the classical master equation can not be used to treat the case of coherent coupling in a double well system discussed below .the suppression of fluctuations at low frequency is due to the exclusion principle in the well state , reflected in the master equation by the appearance of the anti - commuting field operators .no electron can tunnel onto the well if an electron is already there .we need to wait a time of the order of @xmath57 for the electron to tunnel back out into the collector .thus strong anti - correlations are established in the two fundamental poisson processes , @xmath58 .if the tunnelling particles were bosons , the well could accumulate a large number of particles , enhancing the probability for emission into the collector .this would lead to a rapid bunching of emission events into the collector and a super - shot noise current would result . at high frequencies , we are looking at fast processes in which an electron tunnels into the well and immediately tunnels out .the fano factor at high frequencies is 0.5 due to the assumed form of ramo - shockley theorem .we now apply our approach to a triple barrier and double quantum well involving elastic scattering within the wells and coherent coupling between the wells . the main procedures are parallel to those in single well case but now involve off - diagonal elements of the density matrix .the master equation is , @xmath59 \right ] \nonumber \\ & & -\eta _ {2}\left [ c_{2}^{\dagger } c_{2},\left [ c_{2}^{\dagger } c_{2},\rho \right ] \right ] \nonumber \\ & & -i\omega \left [ ( c_{1}^{\dagger } c_{2}+c_{2}^{\dagger } c_{1}),\rho \right]\end{aligned}\ ] ] where @xmath60 and @xmath61 are annihilation ( creation ) operator of electron in the left and right quantum well respectively , @xmath62 is the rate of elastic scattering ( electron - phonon for example ) in the @xmath63 well , and @xmath64 is the coherent coupling rate between the two well states . the irreversible term describing the elastic scattering is derived in much the same way as the inelastic tunnelling terms that describe electrons entering and leaving the device , with one additional assumption . to get a markov master equation for number conserving scattering eventswe must assume that the temperature of the bath describing such processes is high enough that the bath states are well away from the ground states .this is not a very restrictive assumption for realistic devices at millikelvin temperatures .the deviations that can result for very low temperatures are described in gardiner@xcite the derivation of the scattering term in the master equation ( 27 ) is detailed in the appendix .the last term in this equation represents a coherent coupling between the two wells and causes a single electron to periodically tunnel backward and forward between the two wells , until it is eventually lost through the final barrier .recently blick et al@xcite , have made measurements on a structure that can be roughly approximated by our model .as they point out this device exhibits a new feature , in that a single electron can be in a superposition state between the two wells and is thus like an artificial molecule .we first derive the noise features in the presence of this coherent coupling .we will then show that , in the limit of strong elastic scattering , @xmath65 , the system can be described in terms of population rate equations that have been extensively used in the past .the steady state current is easily found to be given by @xmath66 for a symmetric system , @xmath67 ; @xmath68 .the appropriate correlation functions may be evaluated to give , @xmath69 \right\}\end{aligned}\ ] ] where @xmath70 and @xmath71 ; @xmath72 .the noise spectra are derived in two cases . in case1 : @xmath73 , when the elastic scattering rate , @xmath74 , is higher than the coherent coupling rate between the well states , the current noise spectrum is @xmath75 \right\}\end{aligned}\ ] ] the current fano factor against normalised frequency is plotted in fig . 1 where the spectrum shows lorentzian feature .in case2 : @xmath76 , the opposite situation , when coherent coupling is much stronger than elastic scattering , the noise spectrum @xmath77\right \}\ ] ] where @xmath78 , is symmetrically double peaked about the free particle frequency ( @xmath79 ) as shown in the fig .a comparison of these two quantum processes is shown in fig .when elastic scattering increases , the fano factor increases . increasing the coherent coupling results in noise suppression and the double peak feature are more significant as the two well coupling and the quantum correlations are stronger .further when elastic scattering is extremely weak : @xmath80 , and coherent coupling is strong : @xmath81 , the steady state current @xmath82 approaches the single well case as expected in this limit .a significant outcome is that the best noise reduction at low frequency when @xmath83 reaches 0.22 .the coherent tunnelling between the two wells has a strong effect on the noise characteristics .electrons are periodically transferred between the two wells at the tunnelling frequency .if an electron from the emitter is injected into the first well , no further electrons can enter this well until this electron is removed , which takes place on a time scale determined by @xmath84 .thus at frequencies smaller than @xmath64 , noise is suppressed by the exclusion principle , just as for the single well case . at the tunnel frequencyhowever we expect the noise to increase , as electrons injected into the first well are quickly cycled to the second well , where they can incoherently escape to the collector .this explains the two peaked structure of the noise power spectrum . in the case of large @xmath64however , coherent coupling dominates .in that case if an electron tunnels into the first well it periodically returns to that well at a frequency of @xmath85 . to see thisit is sufficient to note that the two levels which are degenerate in the absence of tunnelling become split into symmetric and anti symmetric combinations , separated in energy by @xmath86 .a state initially localised in one well can then be written as a linear combination of the two new eigenstates .the phase difference in the superposition rotates through @xmath87 at the frequency @xmath88 which leads to a state localised in the other well .this is just the standard description of tunnelling in a two state system .the periodic of return of the electron to the first well suppresses another electron from entering the well .thus at large values of @xmath64 we expect noise suppression to occur at @xmath89 .this behaviour is indeed seen in figure4 .we now show that in the limit of strong elastic scattering @xmath90 ( case 1 above ) , a population master equation can be derived that describes a classical sequential tunnelling structure .the sequential model is traditionally formulated in terms of a classical master equation for the occupation probabilities of each well . in our case , we have restricted the discussion to a single bound state in each well and thus the maximum population in each well is unity .however we can derive an equivalent classical master equation to describe sequential tunnelling even in this case .our method is an extension of adiabatic methods used in quantum optics to obtain rate equations .we assume that the off - diagonal elements of the double well density operator are rapidly damped due to the elastic scattering rates @xmath62 .the off - diagonal elements are then assumed to relax almost instantaneously to their steady state values and adiabatically follow the more slowly changing diagonal matrix elements . from eq(27 )we find the following equations of motion for the matrix elements in the occupation number basis for each well , @xmath91 @xmath92-% % tcimacro{\dfrac{\gamma _ { r}}{2 } } % beginexpansion { \displaystyle { \gamma _ { r } \over 2}}% % endexpansion ( n_{2}\delta _ { n_{2},1}+m_{2}\delta _ { m_{2},1 } ) \\-\eta _ { 1}(n_{1}^{2}\delta _ { n_{1},1}-2n_{1}m_{1}\delta _ { n_{1},1}\delta _ { m_{1},1}+m_{1}^{2}\delta _ { m_{1},1 } ) \\-\eta _ { 2}(n_{2}^{2}\delta _ { n_{2},1}-2n_{2}m_{2}\delta _ { n_{2},1}\delta _ { m_{2},1}+m_{2}^{2}\delta _ { m_{2},1 } ) \end{array } \right\ } < n_{1}n_{2}|\rho |m_{1}m_{2}>$ ] @xmath93 @xmath94 @xmath95 @xmath96 @xmath97 @xmath98 where @xmath99 refer to the occupation number of the first and second wells respectively .note that the diagonal matrix elements represent the occupation probabilities of each well , @xmath100 the diagonal matrix elements then obey the equation , @xmath101 @xmath102<n_{1}n_{2}|\rho |n_{1}n_{2}>$ ] @xmath103 @xmath104 @xmath105 @xmath106 @xmath107 therefore , the population equation we are interested is , @xmath108 @xmath109p(n_{1},n_{2},t)$ ] @xmath110 @xmath111 $ ] note that the elastic scattering rates , @xmath112 , do not directly enter this equation .this is because elastic scattering does not change the occupation of the well states but does disrupt the phase coherence between the wave functions in the wells .this will lead to a decay of the relevant off - diagonal matrix elements , which obey the equations , @xmath113 @xmath114-\eta _ { 2}n_{2}^{2}\delta _ { n_{2},1}\}y_{1}(t)$ ] @xmath115 @xmath116 @xmath117 @xmath118 @xmath119-% % tcimacro{\dfrac{\gamma _ { r}}{2 } } % beginexpansion { \displaystyle { \gamma _ { r } \over 2}}% % endexpansion [ n_{2}\delta _ { n_{2},1}+(n_{2}+1)\delta _ { n_{2}+1,1}]$ ] @xmath120\}y_{2}(t)$ ] @xmath121 @xmath122 @xmath123 @xmath124 @xmath125y_{3}(t)$ ] @xmath126 . to proceed we solve the equations for @xmath127 in the steady state , assuming that the diagonal matrix elements are time constant in time over the lifetime of the of - diagonal matrix elements .this is the adiabatic approximation .these steady state values are then substituted back into the equation for the diagonal matrix elements to obtain a classical jump process master equation to describe sequential tunnelling .the algebra is tedious , so we will not give details .the result is @xmath108 @xmath128p(n_{1},n_{2},t)$ ] @xmath110 @xmath129 $ ] @xmath130\}$ ]. where @xmath131 is given by @xmath132 with@xmath133 @xmath134 @xmath135-% % tcimacro{\dfrac{\gamma _ { r}}{2 } } % beginexpansion { \displaystyle { \gamma _ { r } \over 2}}% % endexpansion [ n_{2}\delta _ { n_{2},1}+(n_{2}+1)\delta _ { n_{2},0}]$ ] @xmath136 $ ] @xmath137@xmath138 in addition to the incoherent tunnelling of electrons between the wells and the external reservoirs , we now have incoherent ( sequential ) tunnelling between the two wells at rates determined by @xmath139 .the form of this equation corresponds to the sequential tunnelling master equation obtained by carlos egues et al@xcite .we have thus shown that , in the limit of strong decoherence induced by elastic scattering of the bound states , a population master equation may describe sequential tunnelling in the device .this will be the appropriate limit in the case that @xmath65 .however future quantum nanostructures devices are likely to operate in the opposite limit . in that caseour method is ideally suited for determining the device characteristics .we have shown how the quantum theory of open systems , formulated as a quantum stochastic process , enables the current noise spectrum to be calculated for mesoscopic tunnelling devices .our approach explicitly treats quantum noise properties of the charge carriers , and gives a simple intuitive picture to understand the results .as fabrication technology develops , quantum noise limited networks of _ coherent _ tunnelling devices , such as quantum dots and quantum point contacts , will become increasingly important .such coherently coupled devices are essential for the implementation of a quantum computer , which must operate reversibly@xcite .the full operator master equation methods we have demonstrated here provide a powerful description , including both diagonal and off - diagonal matrix elements in the same equation .our model does not treat the transverse unbound modes in the well of a realistic resonant tunnelling device .these can easily be incorporated by additional states in the well and additional jump process channels in the master equation .we have not done that here as we sought to derive the irreducible level of current noise in tunnelling devices .our model may in fact apply to very tightly confined quantum dot structures which could conceivably be fabricated with a single bound well states at donor impurities .further extensions of the model are also needed to treat the case where the well state is just below the fermi level in the collector in which case the current noise acquire an additional temperature dependent classical component .these more general cases will be treated in a larger publication . in this appendixwe derive the master equation describing elastic scattering of the quasi - bound states of the well which cause a dephasing of the electron quasi - bound states but do not change their populations .the hamiltonian for double - well system in schroedinger picture is where @xmath144 where @xmath145 are bose destruction operators describing independent reservoir oscillators .note that each bound state in the well is coupled to an independent reservoir .this assumes that there are no correlations between well states due to the dephasing that takes place through elastic collisions .we will only consider here the derivation of the master equation arising from the elastic scattering of bound states and the harmonic oscillator reservoirs .the relevant part of the master equation is @xcite where @xmath147 is the equilibrium state of the bath , and where @xmath148 means to trace over the bath variables .this equation is based on a second order expansion in the interaction energy between the reservoir states and the bound states of the well .we have also assumed that the system and bath states are decorrelated very rapidly on the time scale of interest in the system , so that the bath remains close to its equilibrium state .the bath hamiltonian is @xmath155 & = & tr_{b}[\alpha _{ q}^{\dagger } \alpha _ { q^{^{\prime } } } ^{\dagger } \rho _ { b}]=0 \eqnum{a13 } \\ tr_{b}[\alpha_ { q}^{\dagger } \alpha _ { q^{^{\prime } } } \rho _ { b } ] & = & \delta _ { qq^{^{\prime } } } % % tcimacro{\dfrac{1}{e^{e_{q}/k_{b}t}-1 } } % beginexpansion { \displaystyle { 1 \over e^{e_{q}/k_{b}t}-1}}% % endexpansion \eqnum{a14}\end{aligned}\ ] ] @xmath156 } \nonumber \\ & & + ( 1+% % tcimacro{\dfrac{1}{e^{e_{q}/k_{b}t}-1 } } % beginexpansion { \displaystyle { 1 \over e^{e_{q}/k_{b}t}-1}}% % endexpansion e^{i[(\omega _ { q}-\omega _ { q^{^{\prime } } } ) t-\omega _ { q}\tau ] } ] ( c_{n}^{\dagger } c_{n})^{2}\rho \nonumber \\ & = & \int_{0}^{t}d\tau \sum\limits_{q}|m_{q}|^{2}[% % tcimacro{\dfrac{1}{e^{e_{q}/k_{b}t}-1 } } % beginexpansion { \displaystyle { 1 \over e^{e_{q}/k_{b}t}-1}}% % endexpansion e^{i\omega _ { q}\tau } + ( 1+% % tcimacro{\dfrac{1}{e^{e_{q}/k_{b}t}-1 } } % beginexpansion { \displaystyle { 1 \over e^{e_{q}/k_{b}t}-1}}% % endexpansion ) e^{-i\omega _ { q}\tau } ] ( c_{n}^{\dagger } c_{n})^{2}\rho \nonumber \\ & = & \sum\limits_{q}|m_{q}|^{2}[% % tcimacro{\dfrac{\sin ( \omega _ { q}t)}{\omega _ { q } } } % beginexpansion { \displaystyle { \sin ( \omega _ { q}t ) \over \omega _ { q}}}% % endexpansion ( 1+% % tcimacro{\dfrac{2}{e^{e_{q}/k_{b}t}-1 } } % beginexpansion { \displaystyle { 2 \over e^{e_{q}/k_{b}t}-1}}% % endexpansion ) + i% % tcimacro{\dfrac{\cos ( \omega _ { q}t)-1}{\omega _ { q } } } % beginexpansion { \displaystyle { \cos ( \omega _ { q}t)-1 \over \omega _ { q}}}% % endexpansion ] ( c_{n}^{\dagger } c_{n})^{2}\rho \nonumber \\ & = & ( \eta _ { n}+i\xi _ { n})(c_{n}^{\dagger } c_{n})^{2}\rho \eqnum{a15}\end{aligned}\ ] ] the coefficients @xmath162 appear to be time dependant , but under reasonable physical assumptions are time independant@xcite . these assumptions are , firstly that @xmath163 is assumed to be a time scale over which the system operators vary significantly . on this time scale bath correlation functions decay rapidly . secondly , that the bath is at finite temperature and there is significant excitation above the reservoir ground statefinally that the coupling constants @xmath164 are independent of @xmath165 up to some large cut - off wave number . under these assumptionsthese coefficients can be evaluated in the limit of @xmath166 .we refer the reader to reference @xcite for more details .

<S> a quantum markovian master equation is derived to describe the current noise in resonant tunnelling devices . </S><S> this equation includes both incoherent and coherent quantum tunnelling processes . </S><S> we show how to obtain the population master equation by adiabatic elimination of quantum coherences in the presence of elastic scattering . </S><S> we calculate the noise spectrum for a double well device and predict sub - shot noise statistics for strong tunnelling between the wells . </S><S> the method is an alternative to greeen s functions methods , and population master equations for very small coherently coupled quantum dots . </S>

hep-th9511177

investigation of the quark dynamics inside hadrons remains as before the most important task of theoretical and experimental studies in hadronic physics . in view of asymptotic freedom in qcd, this problem at small distances can be treated in the framework of perturbative calculations .however at distances like hadron size @xmath0 and larger nonperturbative effects become important . therefore in thisregion lattice simulations and string models are used for revealing the interquark interaction ( for a review and further references , see , e.g. @xcite , @xcite ) .usually one assumes that at distances @xmath1 the most important are such configurations of gluonic field when this field is concentrated along the line connecting quarks ( the flux tube between quarks ) .such infinitely thin flux tube is simulated by relativistic string . in the most simple case ,the dynamics of this system is determined by the nambu goto string action @xcite .however other string models were also proposed to describe quark interaction , for example polyakov kleinert rigid string @xcite , @xcite .search for new string models is stimulated by comparing the calculations in qcd with those in the framework of known string models @xcite .a large body of literature is devoted to the calculation of the interquark potential in string models ( see , for example , @xcite@xcite and references therein ) .the central point in these calculations is a determination of the string energy as a function of the distance between quarks .when deconfinement temperature ( or critical temperature ) is investigated then the dependence of the effective string tension on temperature should be calculated .practically in all these investigations only the static interquark potential has been considered .it means that the ends of the string connecting quarks are fixed . on the one hand, this assumption essentially simplifies the problem because the string frequencies in this case are integer . on the other hand ,results obtained in this way are applied , strictly speaking , only to the infinitely heavy quarks which can not move .however in this case the notion of the interquark potential practically loses its usefulness . as a matter of factstatic string potential is treated , after its derivation , as a usual potential describing the interaction between quarks with finite masses .here it is implicitly assumed that the interquark potential is a smooth function of quark mass @xmath2 when @xmath3 .probably it is right .calculation of the interquark potential generated by string with massive ends carried out in our recent paper @xcite testifies to this conclusion .nevertheless we would like to draw attention to one peculiarity in energy spectrum of the relativistic string with heavy quarks at its ends which is , to our view , very important .the point is that the position of the first excited level in this system depends on the quark mass @xmath2 and when @xmath3 this level may be arbitrary close to the ground state .obviously this testifies to the infrared instability of the string connecting quarks with sufficiently large masses .it is this problem that will be discussed in our short comunication .the layout of the paper is the following . in section 2 the basic facts from the classical and quantum theories of the nambu goto string with massive ends are given .then linearization of equations of motion and boundary conditions is considered .it should be noted here that it is these linearized equations that are used as a starting point in all the calculations of the interquark potential in string models both in perturbation theory and by making use of the functional integration ( variational estimations ) .therefore the spectrum of eigenfrequencies in this system is of great importance . in section 3the quark mass dependence of the string spectrum is investigated .it is shown that the energy of the first excited level , @xmath4 , may be arbitrary close to the ground state with energy @xmath5 when quark mass tends to infinity .this is a direct indication about the infrared instability in the system under consideration . in conclusion ( section 4 )the physical implications of this result and possible modifications of the string model for removing this drawback are shortly discussed .the nambu goto string with point like masses attached to its ends is described by action @xcite @xmath6 where @xmath7 is infinitesimal area of the string world surface , @xmath8 , @xmath9 , are the world trajectories of the string massive ends , @xmath10 are the quark masses ( @xmath11 ) .@xmath12 is the string tension . in hadronic physics oneusually sets @xmath13 gev .the gauss parametrization of the string world surface embedded into @xmath14dimensional space time with signature @xmath15 @xmath16 @xmath17 permits us to write the induced metric on this surface , @xmath18 , in the following way @xmath19 here @xmath20 and @xmath21 . from eq .( 2.3 ) it follows that in the quadratic approximation the determinant @xmath22 of the induced metric is given by @xmath23 in this approximation , the infinitesimal area @xmath7 can be written as @xmath24 \,dt\,dr$ ] , while the line elements @xmath25 take the form @xmath26\,dt$ ] .finally the action ( 2.1 ) becomes @xmath27\,+\,\sum_{a=1}^{2}\ , \frac{m_a}{2}\,\int\limits_{t_1}^{t_2}\!dt\ , \dot{\bf { u}}^2(t , r_a)\,{,}\ ] ] @xmath28 equations of motion and boundary conditions for the system described by ( 2.5 ) can be immediately deduced @xmath29 @xmath30 @xmath31 where @xmath32 and we put for simplicity @xmath33 .the general solution to eq .( 2.6)(2.8 ) is given by @xmath34}\,\frac{\alpha^j_n}{\omega_n}\ , u_n(r),\quad j=1,\dots , d-2\,{.}\ ] ] amplitudes @xmath35 satisfy the usual rule of complex conjugation @xmath36 .the eigenfunctions @xmath37 in ( 2.9 ) are defined by @xmath38\,{.}\ ] ] @xmath39 s are normalization constants and @xmath40 is the dimensionless parameter @xmath41 the eigenfrequencies @xmath42 are the roots of the transcendental equation @xmath43 on the @xmath44-axis these roots are placed symmetrically around zero . hence they can be numbered in the following way @xmath45 therefore it will be sufficient to consider only the positive roots .the hamiltonian operator of the system under consideration in terms of creation and annihilation operator , @xmath46 and @xmath47 respectively , reads @xmath48 where @xmath49\,=\,\delta^{ij}\,\delta_{nm},\quad n , m\,=1,2,\dots \,{.}\ ] ] the last term in eq .( 2.13 ) is the casimir energy @xcite , @xcite .in this section we investigate the behaviour of the energy levels in string spectrum at large quark masses . to this end , let us separate the integer part in each root of eq .( 2.12 ) @xmath50 where the correction @xmath51 to the frequencies @xmath42 depends on quark masses . by substituting eq . ( 3.1 ) into eq . ( 2.12 ) , one finds , for arbitrary @xmath52 and in the limit @xmath53 , that @xmath51 satisfies the following equation @xmath54 it means that @xmath55 when @xmath56 .solution to eq .( 3.2 ) for @xmath57 is @xmath58 for @xmath59 we have @xmath60 energy levels of the system are calculated by making use of eq .the energy of the first excited state , @xmath4 , is @xmath61 for fixed string tension @xmath62 and string length @xmath63 , the energy @xmath4 decreases for raising values of the dimensionless parameter @xmath40 . from physical point of view, this behaviour of @xmath4 implies the instability of the string with heavy quarks at its ends .figure 1 shows the dependence of the dimensionless energy of the first excited state @xmath64 on the dimensionless quark mass @xmath65 ( bold face curve ) .this curve is obtained for the string length of order ofthe hadron size . from frequency equation ( 2.12 )one can easily deduce the following scaling property of its roots .let the functions @xmath66 determine the roots of this equation for given value of the mass parameter @xmath67 and for @xmath68 i.e. @xmath69 then for arbitrary string length @xmath70 the following scaling rule holds @xmath71 ] @xmath72 . when @xmath73 ( region of the top quark mass ) we have @xmath74 .there are no any restrictions on the number of excitations with energy @xmath4 .therefore , we really have infinite sequence of equidistant energy levels @xmath75 which is superimposed on the basic string spectrum with scale @xmath62 ( see eq .( 2.13 ) ) . as a consequence , for sufficiently large quark massesthe ground state of the string becomes fuzzy , i.e. weak , compared with @xmath62 , external perturbations will result in excitation of the string ( see fig . 1 ) .obviously this implies the infrared instability in this system .it should be noted that all these considerations are applicable to the rigid string too .in fact , in ref . @xcite the same frequency equation as ( 2.11 ) was derived in the model of the rigid string with massive ends . where may this infrared instability be manifested ? at first , when going beyond the linear approximation in string model . to this endwe have to take into account the `` interaction '' in the string action ( 2.4 ) which is described by terms @xmath76 with @xmath77 .besides , this infrared instability may be revealed in calculations at finite temperature . for example , effective string potential generated by linearized string action ( 2.4 ) is obtained by functional intagration over `` field '' @xmath78 . as a result ,one arrives at the following well known expression @xcite @xmath79\,{,}\ ] ] where @xmath80 are the matsubara frequencies and @xmath81 are the roots of eq .when @xmath3 , @xmath82 and we obtain infinity in ( 3.7 ) .on the other hand , calculations in string models with fixed ends are well defined at finite temperature @xcite .the instability of the relativistic string , connecting heavy quarks is quiet clear because fixed string tension @xmath12 proves to be insufficient to keep in bound state very heavy quarks . in the framework of the string approach to hadron physicsthis instability apparently implies that string like collective excitations of gluon field do not dominate in heavy quark dynamics . though other points of view can be proposed .for example , one way suppose that string tension for heavy quarks should be greater than @xmath12 .however the following problem arises here : in what way this assumption can be put in agreement with basic concepts of qcd . in the same way one may doubt in applicability of the linearized string action ( 2.5 ) which has been used to derive frequency equation ( 2.12 ) . but in this case it would be difficult to understand why this action is good for description of the string with light quarks at its ends or for fixed string ends but is not applicable to heavy quarks . proceeding from simple physical considerationsone may expect that situation should be opposite because linearization of the string action is , in some sense , equivalent to the nonrelativistic approximation ( see , for example , @xcite ) .thus the problem of infrared instability in the dynamics of the relativistic string connecting sufficiently heavy quarks remains , in our opinion , open .+ * acknowledgements * + one of the authors , v.v .n. , would like to thank n.r .shvetz ( yale university ) for valuable discussion of some topics considered in this article .acknowledge the financial support from bogoliubov laboratory of theoretical physics ( jinr ) and the salerno university .this work has been partly supported by the russian foundation for fundamental research through project no .93 - 02 - 3972 .99 m. flensburg and c. peterson , nucl .b238 ( 1987 ) 141 .born et al .b329 ( 1994 ) 325 .barbashov and v.v .nesterenko , introduction to the relativistic string theory ( world scientific , singapore , 1990 ) .a.m. polyakov , nucl .b286 ( 1986 ) 406 .h. kleinert , phys .b174 ( 1986 ) 335 ; phys .58 ( 1987 ) 1915 .j. polchinski and z. yang , phys .d46 ( 1992 ) 3667 .o. alvarez , phys.rev .d24 ( 1981 ) 440 .h. kleinert , phys .d40 ( 1989 ) 473 .g. german , mod .lett . a6 ( 1991 ) 1815 .e. dhoker , p. sikivie and y. kanev , phys .b347 ( 1995 ) 56 .e. elizalde , s. leseduarte and s.d .odintsov , phys .d48 ( 1993 ) 1757 .h. kleinert , quark mass dependence of @xmath83 term in string potential .preprint of fub , berlin , 1986 ( unpublished ) .g. lambiase and v.v .nesterenko , quark mass correction to the string potential .jinr preprint e2 - 95 - 443 , dubna , 1995 ; hep - th/9510221 .submitted to phys .timosenko and d.h .young , vibration problems in engineering , 3rd ed .( van nostrand , new york , 1955 ) .koshlyakov , e.b .gliner and m.m .smirnov , the basic differential equations of mathematical physics , moscow : fizmatgiz 1962 .mostepanenko and n.n .trunov , the casimir effect and its applications , ( moscow energoatomizdat , 1990 ) ( in russian ) .g. pluniem , b. mller and w. greiner , phys .* 134 * ( 1986 ) 87 .kapusta , finite temperature field theory , cambridge university press , cambridge 1989 .a. antillon and g. german , phys .d47 ( 1993 ) 4567 .nesterenko , z. phys .c - particle and fields 51 ( 1991 ) 643 . fig . 1 .quark mass dependence of the energy levels in string spectrum .the bold faced curve is the dimensionless energy of the first excited state , @xmath84 ; @xmath65 is dimensionless quark mass .other curves present the next 5 energy levels , @xmath85

<S> the quark mass dependence of the energy spectrum in the nambu </S><S> goto string with point like masses ( quarks ) at its ends is analyzed . to this end , </S><S> linearized equations of motion and boundary conditions in this model are considered . </S><S> it is shown that for sufficiently large quark masses , the first excited state in string spectrum may be arbitrary close to the ground state . </S><S> obviously this points to infrared instability in the system under consideration . </S><S> possible modifications of string model which could remove this drawback are discussed . </S><S> -12pt 4000 4000 1.2 cm pacs numbers : 11.17.+y , 12.40.qq , 14.80.d + key words : </S>

1511.07079

electrical impedance tomography ( eit ) is a non - invasive imaging technique which aims at reconstructing the inner structure of a reference subject from the knowledge of the current and voltage measurements on the boundary of the subject .typically , an array of electrodes are attached to the boundary of the reference subject , then small currents are applied to some or all of the electrodes and the resulting electric voltages are measured at the electrodes .these current and voltage measurements on the boundary of the reference subject are used to estimate the value of the conductivity inside the subject .the result is an image of the inner structure of the subject due to the fact that different materials have different conductivities .compared with computerized x - ray tomography , eit is less costly and requires no ionizing radiation ; hence , it qualifies for many clinical applications including lung ventilation ( e.g. @xcite ) , brain imaging ( e.g. @xcite ) , breast cancer diagnosis ( e.g. @xcite ) , etc . on the other hand , eit can also be used for nonclinical purposes such as determining the location of mineral deposits ( e.g. @xcite ) , describing soil structure ( e.g. @xcite ) , identifying cracks in non - destructive testing ( e.g. @xcite ) and so on .the inverse problem of reconstructing the conductivity from voltage - current - measurements is known to be highly ill - posed and nonlinear , and reconstructions suffer from an enormous sensitivity to modeling and measurement errors . to reduce modeling errors ,one usually concentrates on reconstructing a conductivity change with respect to a known reference conductivity .then , the most natural approach is to parametrize the support of the conductivity change and determine these parameters by an iterative , nonlinear inverse problems solver .such iterative methods yield good reconstructions for a given good initial guess ; however , they require expensive computation and have no convergence results .non - iterative methods such as the factorization method ( @xcite ) and the monotonicity - based method ( @xcite ) , on the other hand , are rigorously justified and require no initial guess .however , the reconstructions of both factorization method and monotonicity - based method tend to be rather sensitive to measurement errors when phantom data or real data are applied @xcite . in clinical studies , it is common practice to start by linearizing the eit problem around a known reference conductivity and minimizing the residual functional of the linearized equation ( herein called linearized residual functional for the sake of brevity ) .the resulting problem can be regularized in various ways .for example , @xcite considered the minimization problem of a linearized residual functional with a regularization term , which is similar to the standard tikhonov regularization method .the algorithm proposed in @xcite is fast and simple .however , no convergence proofs are available so far . besides , one can also use sparsity reconstruction @xcite .this is an effective method to detect piecewise constant inclusions .however , strictly speaking , convergence to the true conductivity is still an open issue , as it is not clear how to obtain a global minimizer .our new method is also based on minimizing the linearized residual functional .however , instead of adding a regularization term , we employ a linear constraint defined by the monotonicity test ( * ? ? ?* theorem 4.1 ) which plays the role of a special regularizer. global convergence of the shape reconstructions is then proved and numerical results show that this method provides good shape reconstructions even under high levels of noise . to the authors knowledge , this is the first reconstruction method based on minimizing the residual that has a rigorous global convergence property . for the question of globally convergent algorithms for other classes of inverse problems , see for example @xcite . the paper is organized as follows .section [ sec : pre ] presents some preliminaries and notations . in section [ sec : theo ] , we state andprove our theoretical results .section [ sec : num ] shows the numerical experiments and we conclude with a brief discussion in section [ sec : con ] .we consider a bounded domain @xmath0 in @xmath1 ( i.e. @xmath0 is a bounded open connected subset of @xmath1 ) , @xmath2 with smooth boundary @xmath3 and outer normal vector @xmath4 .we assume that @xmath0 is isotropic so that the electric conductivity @xmath5 is a scalar function , and that @xmath6 is bounded and strictly positive . inside @xmath0 ,the electric potential @xmath7 is governed by the so - called _conductivity equation_. on the boundary of @xmath0 , @xmath8 satisfies the _neumann condition_. @xmath9 in this work , we shall follow the _ continuum model _( e.g. ( * ? ? ?* subsection 12.3 ) ) , which assumes that there are no electrodes and that the current density @xmath10 is applied over all @xmath3 .the electric voltage , in this case , can be measured at every point of the boundary @xmath3 and is denoted by @xmath11 .the forward problem of ( [ eqn : conductivity ] ) is to determine the potential @xmath8 for given data @xmath6 and @xmath12 .the existence of a variational solution @xmath13 for this neumann boundary value problem is obtained due to the lax - milgram theorem , while the uniqueness ( up to an additive constant ) is straight - forward .the forward problem of ( [ eqn : conductivity ] ) is well - posed in the sense that the potential @xmath8 depends continuously on the neumann data @xmath12 . to guarantee that @xmath8 is uniquely defined, one would require furthermore that the solution @xmath8 has zero integral mean , i.e. @xmath14 notice that , the current density @xmath12 should satisfy the conservation of charge @xmath15 , since @xmath16 to guarantee that @xmath8 is uniquely defined , one would require furthermore that the solution @xmath8 has zero integral mean , i.e. @xmath14 in fact , by fixing a ground voltage @xmath17 and considering @xmath18 , we see that @xmath19 satisfies ( [ eqn : conductivity ] ) and @xmath20 since @xmath8 is unique up to an additive constant , @xmath19 is unique with respect to the ground voltage @xmath21 .the unique solvability of the forward problem ( [ eqn : conductivity ] ) guarantees the existence of the neumann - to - dirichlet ( ntd ) operator @xmath22 , which maps each current @xmath12 to the voltage measurement @xmath23 on the boundary : @xmath24 here , @xmath25 is the unique variational solution of the forward problem ( [ eqn : conductivity ] ) corresponding to the conductivity @xmath6 and the boundary current @xmath12 , and @xmath23 is understood as the trace of @xmath26 on the boundary @xmath3 .@xmath27 is the subspace of @xmath28 with positive essential infima . @xmath29 and @xmath30 denote the spaces of @xmath31- and @xmath32-functions with vanishing integral mean on @xmath3 .it is well - known that @xmath22 is a linear , bounded , compact , self - adjoint , positive operator from @xmath30 to @xmath30 ( see for example ( * ? ? ?* chapter 5 ) for two - dimensional cases ) . for each @xmath33 the quadratic form associated with @xmath22 is : @xmath34 the existence of the frchet derivative @xmath35 of the ntd operator @xmath36 can be found for example in @xcite .given some direction @xmath37 , the derivative @xmath38 is associated with the quadratic form : @xmath39 the inverse problem of ( [ eqn : conductivity ] ) is to determine the conductivity @xmath6 from the knowledge of the ntd operator @xmath22 .obviously , @xmath22 depends on @xmath6 nonlinearly , and like many other nonlinear inverse problems , this is an ill - posed problem .in fact , it is known that small amounts of noise or model errors may cause poor spatial resolution .the reader is referred to mueller and siltanen s book @xcite for further explanation about the nonlinearity and the ill - posedness of this inverse problem .the uniqueness of solutions of the inverse problem ( [ eqn : conductivity ] ) has been investigated for different classes of conductivities and dimensions immediately following calderon s pioneer paper @xcite in 1980 , for example , kohn and vogelius @xcite for piecewise analytic conductivities , sylvester and uhlmann @xcite for @xmath40-conductivities in dimension @xmath41 , nachman @xcite for @xmath42-conductivities in dimension @xmath43 , astala and pivrinta @xcite for @xmath44-conductivities in dimension @xmath43 . in the next section, we shall propose a regularization scheme to construct an approximate solution and prove stability in the presence of noise .in this work , we need the definiteness assumption , i.e. either @xmath45 a.e . @xmath0 or @xmath46 a.e .@xmath0 . for the sake of simplicity, we shall assume that the background conductivity @xmath47 and that the conductivity of the investigated subject is defined by @xmath48 , where the open set @xmath49 denotes the unknown inclusions .we assume furthermore that @xmath50 and that @xmath51 has a connected complement .the goal of eit is to determine the inclusions shape @xmath49 from the knowledge of the ntd operators @xmath22 and @xmath52 .notice that our method also works for inhomogeneous background conductivity .we start by describing our method for exact data .the idea of this method is inspired by a result of seo and one of the authors in @xcite .it is proved that , if @xmath53 is an exact solution of the linearized equation @xmath54 then the support of @xmath53 coincides with @xmath49 .however , it is not clear in general whether such an exact solution exists .in addition , one can not get a similar result for noisy data .it is natural to ask whether minimizing the residual functional @xmath55 under appropriate regularization can yield a solution @xmath53 with correct support . in this work, we can prove that this is indeed possible .more precisely , denote by @xmath56 the @xmath57 given injected currents which are assumed to form an orthonormal subset of @xmath30 , we can replace @xmath58 by the matrix @xmath59 and minimize this @xmath57-by-@xmath57 matrix under the frobenius norm : @xmath60 where @xmath61 stands for the matrix @xmath59 , @xmath62 denotes the frobenius norm and @xmath63 is an admissible set of conductivity change @xmath53 . since there is no hope to reconstruct the conductivity change at every point inside @xmath0 from the knowledge of a finite number of measurements , it is reasonable to restrict @xmath63 to the class of piecewise constant functions .more precisely , the admissible conductivity change @xmath53 is assumed to be constant on a fixed partition @xmath64 of the bounded domain @xmath0 : @xmath65 where the @xmath66 s are real constants and each @xmath67 is assumed to be open , @xmath68 , @xmath69 is connected and @xmath70 for @xmath71 .notice that , this partition is not unique , and any minimizer @xmath72 of ( [ 29.10.1 ] ) as well as any reconstruction shape ( that is , the support of some minimizer @xmath73 ) depend heavily on the choice of this partition .it is well - known that @xmath58 is a linear , bounded , compact , self - adjoint operator from @xmath30 to @xmath30 .perhaps a reasonable choice of an appropriate norm to minimize @xmath58 is the operator norm . in fact , all of the following theoretical results remain true for the operator norm . the numerical results for the operator norm can be easily obtained by considering the equivalent problem of minimizing the maximum eigenvalue of an approximate matrix of @xmath58 .another commonly used norm is the hilbert - schmidt norm .nevertheless , we could not minimize @xmath58 under the hilbert - schmidt norm , since it is not clear whether or not @xmath58 belongs to the class of hilbert - schmidt operators .the idea of using the frobenius norm comes from the fact that in realistic models , one always applies a finite number of currents @xmath12 on the boundary ; and hence , only a finite number of measurements @xmath74 are known .problem ( [ 29.10.1 ] ) was actually considered decades ago in clinical eit such as @xcite .typically , one usually adds a regularization term into the minimization functional , similar to the standard tikhonov regularization method . by this method, good shape reconstructions with real - time implementation can be obtained .however , no rigorous convergence results have been established so far , and the reconstructions usually contain ringing artifacts . in the present paper , we do not follow the standard tikhonow regularization method .instead , we use a linear constraint defined by the monotonicity test @xcite as a special regularizer . a lower bound for an admissible conductivity change @xmath53 is , in fact , due to the fact that @xmath75 . an upper bound for @xmath53 , on the contrary ,is numerically defined by the idea of the monotonicity test @xcite as follows : ( * ? ? ?* example 4.4 ) has proved that , for @xmath76 and for every ball @xmath77 @xmath78 we will show in the proof of lemma [ le : position ] that for any real constant @xmath79 satisfying @xmath80 , it holds that @xmath81 .\hfill\\ \fl \mbox{(ii ) if } p_k \nsubseteq d \mbox { then } \lambda(1)-\lambda(\sigma)+\alpha \lambda'(1)\chi_{p_k}\not\geq 0\ ; \mbox { in quadratic sense for all } \alpha > 0 .\hfill\end{aligned}\ ] ] we see that , when @xmath82 , we have @xmath83 and in this case , @xmath84 is actually the number @xmath85 in ( * ? ? ?* example 4.4 ) .although the formula of @xmath84 depends heavily on the inclusion contrast @xmath86 , in many applications a bound for @xmath86 is known a - priori . for each pixel @xmath67 , the biggest coefficient @xmath87 such that @xmath88\end{aligned}\ ] ] then satisfies : * @xmath89 if @xmath90 .* @xmath91 if @xmath92 .this motivates the constraint @xmath93 on each pixel @xmath67 .note that @xmath87 is allowed to be @xmath94 .our following theory , therefore , requires us to use a stronger upper bound @xmath95 where @xmath96 plays the role of a special regularizer . since @xmath79 is smaller than the true contrast @xmath86 , this seems over - constrained " , but we will show that the minimizer of this over - constrained problem possesses the correct support .thus , we can define the admissible set @xmath63 as follows : @xmath97 here comes our main result .[ thm : main ] consider the minimization problem @xmath98 the following statements hold true : * problem ( [ 5.11.15.1 ] ) admits a unique minimizer @xmath99 . *@xmath100 and @xmath49 agree up to the pixel partition , i.e. for any pixel @xmath67 @xmath101 moreover , @xmath102 .[ re10.11.15 ] notice that @xmath87 is defined via the infinite - dimensional ntd operator @xmath36 and does not involve the finite - dimensional matrix @xmath103 .theorem [ thm : main ] holds regardless the number @xmath57 of applied boundary currents .we would like to emphasize that the goal of our method is to show an approximation of the size and the location of the inclusion @xmath49 , not the value of the conductivity @xmath6 inside @xmath49 . indeed , as we can see from the above theorem , the support of the unique minimizer @xmath99 agrees with @xmath49 up to pixels , while the value of @xmath99 is always smaller than @xmath104 .* theorem 4.3 ) has proved that @xmath105 is injective if there are enough boundary currents ( that is , if @xmath57 is sufficiently large ) . in that case, it is obvious to see that @xmath61 is also injective .this fact together with the fact that the square frobenius norm is strictly convex imply @xmath106 is strictly convex . before proving theorem [ thm : main ] , we need the following lemmas . [le : position ] for any pixel @xmath67 , @xmath90 if and only if @xmath107 . one special case of lemma [ le : position ] has been proved in ( * ? ? ?* example 4.4 ) .we have slightly modified the proof there to fit with our notations and settings .* step 1 : * we shall check that , @xmath90 implies @xmath107 .indeed , employing the following monotonicity principle ( see e.g. ( * ? ? ?* lemma 3.1 ) ) @xmath108 for @xmath109 and @xmath110 , we get the following inequalities for all pixels @xmath67 , all @xmath111 $ ] and all @xmath33 : @xmath112 here @xmath113 is the unique solution of the forward problem ( [ eqn : conductivity ] ) when the conductivity is chosen to be @xmath114 .the last inequality holds due to the fact that @xmath115 in @xmath49 and that @xmath67 lies inside @xmath49 .* step 2 : * if @xmath107 , we will show that @xmath90 by contradiction : assume that @xmath107 and @xmath92 .applying the following monotonicity principle ( * ? ? ?* lemma 3.1 ) @xmath116 where we choose @xmath117 , and taking into account the definition of @xmath87 in ( [ ineq:1 ] ) , we obtain @xmath118 thus , for all @xmath33 : @xmath119 with some positive constant @xmath120 . on the other hand ,applying localized potential ( see e.g. ( * ? ? ?* theorem 3.6 ) and @xcite for the origin of this idea ) , we can find a sequence @xmath121 such that the solutions @xmath122 of the forward problem ( [ eqn : conductivity ] ) ( when the conductivity is chosen to be @xmath114 and boundary currents @xmath123 ) fulfill @xmath124 this contradicts ( [ ineq:2 ] ) .[ sk.positive ] for all pixels @xmath67 , denote by @xmath125 the matrix @xmath126 .then @xmath125 is a positive definite matrix . for all @xmath127, we have @xmath128 where @xmath129 .this means that @xmath125 is a positive semi - definite symmetric matrix in @xmath130 .we shall prove that @xmath125 is , in fact , a positive definite matrix by showing that @xmath131 assuming , by contradiction , that there exists @xmath132 such that @xmath133 since @xmath67 is open , there exists an open ball @xmath134 .it holds that @xmath135 a.e .@xmath77 , where @xmath136 is some real constant .this fact can be obtained in many different ways , for example , by the poincar s inequality . on the other hand, @xmath137 is a solution of the forward problem ( [ eqn : conductivity ] ) with conductivity @xmath114 and homogeneous neumann boundary data .therefore , @xmath138 is a solution of ( [ eqn : conductivity ] ) when the conductivity is @xmath114 and the neumann boundary is @xmath139 .moreover , we have proved that @xmath140 a.e . in an open ball @xmath141, the unique continuation principle implies that @xmath140 a.e .@xmath0 , and hence @xmath142 a.e .this contradiction implies that ( [ 6.11.15.1 ] ) holds .now we are in the position to prove our main result . *( i ) * existence of minimizer : since the functional @xmath143 is continuous , it admits a minimizer in the compact set @xmath63 .uniqueness obviously follows when we have proven ( ii ) . *( ii ) step 1 : * denote by @xmath144 .we shall check that , for all @xmath145 satisfying @xmath146 , it holds that @xmath147 in quadratic sense .notice that @xmath148 in @xmath49 . in the same manner as the proof of lemma [ le : position ] , we can write @xmath149 for any @xmath150 .if @xmath151 , we have @xmath152 . by lemma [ le :position ] , it holds that @xmath90 . taking into account that @xmath153 and that @xmath154 for @xmath155 , we get that @xmath156 for any @xmath150 . * step 2 : * let @xmath157 be a minimizer of ( [ 5.11.15.1 ] ) .we prove that @xmath158 . indeed , if @xmath159 , it follows from step @xmath114 and the monotonicity of @xmath105 that @xmath160 this implies @xmath107 .thanks to lemma [ le : position ] , we have @xmath90 . * step 3 : * if @xmath157 is a minimizer of ( [ 5.11.15.1 ] ) , then @xmath161 indeed , it holds that @xmath162 . if there exists a pixel @xmath67 such that @xmath163 in @xmath67 , we can choose @xmath164 such that @xmath165 we will show that @xmath166 which contradicts the minimality of @xmath99 .let @xmath167 be @xmath57 eigenvalues of @xmath168 and @xmath169 be @xmath57 eigenvalues of @xmath170 .since @xmath168 and @xmath170 are both symmetric , all of their eigenvalues are real . by the definition of the frobenius norm ,we get @xmath171 thanks to step 1 , @xmath172 and @xmath173 in the quadratic sense . thus , for all @xmath127, we have @xmath174 where @xmath129 .this means that @xmath175 is a positive semi - definite symmetric matrix in @xmath130 .it is well - known that all eigenvalues of a positive semi - definite symmetric matrix should be non - negative .thus , @xmath176 in the same manner , @xmath177 is also a positive semi - definite symmetric matrix .hence , @xmath178 on the other hand , lemma [ sk.positive ] claims that @xmath125 is a positive definite matrix .thus , all @xmath57 eigenvalues @xmath179 of @xmath125 are positive .since @xmath180 and the matrices @xmath181 and @xmath125 are all symmetric , we can apply weyl s inequalities ( * ? ? ?* theorem iii.2.1 ) to get @xmath182 in summary , ( [ 7.11.15.12 ] ) , ( [ cond:1 ] ) , ( [ cond:2 ] ) and ( [ cond:3 ] ) imply ( [ 7.11.15.11 ] ) .this ends the proof of step 3 .* step 4 : * if @xmath90 , then @xmath183 . indeed , since @xmath99 is a minimizer of ( [ 5.11.15.1 ] ), step 3 implies that @xmath161 since @xmath90 , it follows from lemma [ le : position ] that @xmath184 .thus , @xmath183 . in conclusion , problem ([ 5.11.15.1 ] ) admits a unique minimizer @xmath185 .this minimizer fulfills @xmath186 in the presence of noise , we denote by @xmath187 the noise level .similarly as above , we call @xmath188 the @xmath57-by-@xmath57 matrix @xmath189 , where the residual for noisy data now reads @xmath190 and the error is bounded from above by @xmath187 in the operator norm , i.e. @xmath191 when we replace the exact data @xmath22 by the noisy data @xmath192 , we have to change the definition of the biggest coefficient @xmath87 , too . to this end , we need the following lemma : [ le : positive ] for any bounded linear operator @xmath193 on a real hilbert space @xmath194 : * ( square root lemma ) if @xmath193 is positive , i.e. @xmath193 is self - adjoint and @xmath195 for all @xmath196 , then there exists a unique bounded linear positive operator @xmath197 on @xmath194 such that @xmath198 .moreover , @xmath197 commutes with every bounded linear operator which commutes with @xmath193 .we call @xmath197 the positive square root of @xmath193 and denote by @xmath199 . *( absolute value of a bounded linear operator ) the modulus of @xmath193 @xmath200 is a positive operator , where @xmath201 is the adjoint operator of @xmath193 .moreover , @xmath202 commutes with every bounded linear operator which commutes with @xmath203 . *( positive decomposition ) if @xmath193 is self - adjoint , then there exists a unique pair of bounded positive operators @xmath204 and @xmath205 such that @xmath206 , @xmath207 , @xmath204 and @xmath205 commute with each other and with @xmath193 .moreover , @xmath208 . *if @xmath193 is positive , then @xmath209 .* if @xmath193 is self - adjoint , then @xmath210 in quadratic sense .* for any bounded linear operators @xmath211 on a real hilbert space @xmath194 : @xmath212 consequently , the absolute value of operators is continuous w.r.t . the operator norm . *this famous fact can be proved either by means of classical functional analysis ( see , for example , ( * ? ? ?* appendix ) ) or by using @xmath213-algebra techniques ( ( * ? ? ?* proposition 4.33 ) ) .* for all @xmath214 , it holds that @xmath215 hence , @xmath203 is a positive operator .thanks to ( i ) , @xmath203 has a unique positive square root @xmath216 , which commutes with every bounded linear operator commuting with @xmath203 .* we shall prove this fact by using techniques in @xmath213-algebra .indeed , since @xmath193 is self - adjoint , it is normal .hence , the gelfand map @xmath217 establishes a @xmath218-isometric isomorphism between the space of continuous functions on @xmath219 ( here @xmath219 denotes the spectrum of @xmath193 ) and the @xmath213-algebra @xmath220 ( see , e.g. ( * ? ? ?* theorem 2.31 ) ) .define by @xmath221 then both @xmath222 and @xmath223 are continuous positive functions on @xmath219 .thus , @xmath224 and @xmath225 are well - defined positive operators on the space of bounded linear operators @xmath226 .moreover , @xmath204 and @xmath205 commute with each other and with @xmath193 .since @xmath222 and @xmath223 satisfy @xmath227 for all @xmath228 , it follows that the two positive operators @xmath204 and @xmath205 satisfy @xmath229 + on the other hand , we also have @xmath230 by the same way , one can prove that @xmath205 is also a bounded operator .it remains to show that the pair @xmath231 is unique .indeed , if @xmath232 where @xmath211 are positive bounded operators satisfying @xmath233 , then @xmath234 .thus @xmath235 is the unique square root of @xmath236 , i.e. @xmath237 .hence , @xmath238 and @xmath239 .the uniqueness holds . *since @xmath193 is positive , it holds that @xmath240 . thanks to ( i ) , @xmath193 is a unique positive square roof of @xmath203 , i.e. @xmath241 . on the other hand , by ( ii ) , @xmath242 .thus , @xmath243 . *first we prove that , for any linear , bounded , positive operators @xmath244 on a real hilbert space h , it holds that @xmath245 the above inequality had been proved in ( * ? ? ?* theorem 1 ) .we shall recite the proof here for the reader s convenience .+ denote by @xmath246 then @xmath247 is linear bounded self - adjoint operator on @xmath194 .hence @xmath248 thus , we can find a sequence @xmath249 such that @xmath250 denote by @xmath251 , we have @xmath252 since both @xmath197 and @xmath253 are positive operators , we have @xmath254 .thus , @xmath255 on the other hand , @xmath256 as @xmath257 because @xmath258 combining ( [ 30.10.1000 ] ) , ( [ 30.10.1001 ] ) and taking @xmath257 , we get ( [ 30.10.100 ] ) .+ thanks to ( [ 30.10.100 ] ) , for all bounded linear operators @xmath259 and @xmath77 we have @xmath260 back to our issues , since @xmath261 is a linear bounded positive operator , lemma [ le : positive ] yields that @xmath262 . the monotonicity test ( [ ineq:1 ] ) can be rewritten as latexmath:[\ ] ] where the modulus @xmath373 in this case is called the absolute value of matrices .it is easy to see that @xmath374 as @xmath375 .we shall follow the argument in @xcite to calculate @xmath369 .since @xmath376 is hermitian positive - definite , the cholesky decomposition allows us to decompose it into the product of a lower triangular matrix and its conjugate transpose , i.e. @xmath377 where @xmath378 is a lower triangle matrix with real and positive diagonal entries .this decomposition is unique .moreover , since @xmath379 it follows that @xmath378 is invertible . for each @xmath380 , we have that @xmath381 hence , the positive semi - definiteness of @xmath382 is equivalent to the positive semi - definiteness of @xmath383 .since both @xmath384 and @xmath383 are hermitian matrices , we can apply weyl s inequalities ( * ? ? ?* theorem iii.2.1 ) to obtain @xmath385 where @xmath386 denote the @xmath57-eigenvalues of some matrix @xmath387 .it is known from lemma [ sk.positive ] that @xmath125 is a positive definite matrix .thus , the matrix @xmath388 is positive definite , too .this implies @xmath389 for all @xmath390 .it then follows from ( [ eqn : num_eig ] ) that for every @xmath391 , the functional @xmath392 is decreasing in @xmath393 .the biggest @xmath393 that fulfills @xmath394 should be @xmath395 where @xmath396 is the smallest ( most negative ) eigenvalue of @xmath384 .since @xmath376 is hermitian positive - definite , the cholesky decomposition allows us to decompose it into the product of a lower triangular matrix and its conjugate transpose , i.e. @xmath377 where @xmath378 is a lower triangle matrix with real and positive diagonal entries .this decomposition is unique .moreover , since @xmath379 it follows that @xmath378 is invertible . for each @xmath380, we have that @xmath381 hence , the positive semi - definiteness of @xmath382 is equivalent to the positive semi - definiteness of @xmath383 .since both @xmath384 and @xmath383 are hermitian matrices , we can apply weyl s inequalities ( * ? ? ?* theorem iii.2.1 ) to obtain @xmath397 where @xmath386 denote the @xmath57-eigenvalues of some matrix @xmath387 .it is known from lemma [ sk.positive ] that @xmath125 is a positive definite matrix .thus , the matrix @xmath388 is positive definite , too .this implies @xmath389 for all @xmath390 .it then follows from ( [ eqn : num_eig ] ) that for every @xmath391 , the functional @xmath392 is decreasing in @xmath393 .the biggest @xmath393 that fulfills @xmath394 should be @xmath398 where @xmath396 is the smallest ( most negative ) eigenvalue of @xmath384 .the minimization problem ( [ prb:3 ] ) can be rewritten as follows .@xmath399 since the minimization functional @xmath400 is convex , every minimizer should be global minimizer .the existence of the minimizer of ( [ prb:4 ] ) follows the continuity of the minimization functional @xmath401 and the fact that @xmath402 is uniformly bounded .let @xmath403 , where @xmath404 be a minimizer of ( [ prb:4 ] ) .when @xmath57 is large enough , in the same manner as step 2 , proof of theorem [ thm : stability ] , we can show that @xmath405 for @xmath300 .hence , we can follow the proof of theorem [ thm : stability ] to conclude that @xmath406 pointwise converges to the unique minimizer @xmath99 of ( [ 5.11.15.1 ] ) as @xmath187 goes to @xmath271 ( and @xmath57 is large enough ) .problem ( [ prb:4 ] ) can be solved either by using ` cvx ` , a package for specifying and solving convex programs @xcite , or with the ` matlab ` built - in function ` quadprog ` .notice that reconstruction images are highly affected by the choice of the minimization algorithms ( see figure [ fig:5 ] and table 1 ) .while ` cvx ` allows us to directly work with matrices , ` quadprog ` requires us to rearrange the matrix @xmath333 into a long vector and define the @xmath407-by-@xmath408 matrix @xmath409 whose @xmath295th - column stores the matrix @xmath125 : @xmath410 for @xmath411 and @xmath412 the minimization functional in ( [ prb:4 ] ) then becomes @xmath413 where @xmath414 .and we end up with the following quadratic minimization problem under box constraints : @xmath415 in our numerical experiments , we plot the support of the minimizer @xmath416 of the minimization problem ( [ prb:4 ] ) , where @xmath417 is a partition of the unit disk @xmath0 in @xmath418 centered at the origin and is chosen independently of the finite element mesh that is used for solving the forward problems . in the first experiment ,the true inclusion @xmath49 includes a small ball @xmath77 centered at @xmath419 radius @xmath420 , a rectangle @xmath421 whose lower - left corner is located at @xmath422 and upper - right corner is located at @xmath423 and an ellipse @xmath424 centered at @xmath425 , horizontal semi - axis @xmath426 , vertical semi - axis @xmath420 .the reference conductivity @xmath427 is assumed to be @xmath114 on @xmath0 .the true conductivity @xmath6 is assumed to be @xmath428 on @xmath77 , @xmath429 on @xmath424 , @xmath430 on @xmath421 and @xmath114 outside @xmath49 ; and is plotted in the first picture of figure [ fig:1 ] .the next three pictures of figure [ fig:1 ] show the reconstruction images ( aka .the support of the minimizer @xmath416 of the minimization problem ( [ prb:4 ] ) ) of our method with respect to different levels of noise .figure [ fig:1 ] yields that minimizing the residual under constraint @xmath431 is not affected much by the noise .moreover , our method produces no artifacts even under high levels of noise . in figure[ fig:2 ] , we consider the reconstruction images of the minimization residual problem under different constraints , to see that the upper bound @xmath79 is essential not only for preventing infinity upper bounds when @xmath432 , but also for guaranteeing a good shape reconstruction ( figures [ fig:1]d and [ fig:2]b ) . in many applications ,a bound for the conductivity change @xmath86 is known ; hence , in these cases , the value of @xmath433 can be calculated a - priori .all of the reconstruction images showed in figure [ fig:1 ] and figure [ fig:2 ] are obtained by using ` cvx ` to minimize the residual ( [ prb:4 ] ) . in the second experiment, we consider the true inclusion @xmath49 as a ball centered at the origin and with radius @xmath420 .the reference conductivity @xmath427 is also assumed to be @xmath114 on @xmath0 .the true conductivity @xmath6 is @xmath428 on @xmath49 and @xmath114 outside ; and is plotted in the first picture of figure [ fig:3 ] .the next two pictures of figure [ fig:3 ] plot the support of the minimizer @xmath416 of the minimization problem ( [ prb:4 ] ) following the ` cvx ` minimization algorithm when the @xmath368 s do and do not involve in the box constraint .a cut through the origin of three pictures of figure [ fig:3 ] via the @xmath228-axis is presented in figure [ fig:4 ] .these pictures show that the upper bound @xmath268 s play an important role in the proof of the theoretical part and makes the reconstructions slightly better when the noise level is very low ( figures [ fig:3 ] and [ fig:4 ] ) .figure [ fig:5 ] shows the reconstruction images of the true conductivity in the first experiment ( aka . the first picture of figure [ fig:1 ] ) when different minimization algorithms are applied to solve the minimization ( [ prb:4 ] ) .under the constraint @xmath434 and @xmath435 relative noise , the reconstruction image using ` cvx ` looks perfectly well and contains no ringing artifact at all ; while the ` matlab ` built - in function ` quadprog ` with ` trust - region - reflective ` algorithm also yields very good result that reduces a lot of ringing artifacts compared with the standard tikhonov approach ( the third picture from the left of figure [ fig:2 ] ) .however , the ` matlab ` built - in function ` quadprog ` with default option ( in this case ` interior - point - convex ` algorithm ) totally fails to produce an approximation of the true conductivity .the minimum values and the amount of time taken when solving the minimization problem ( [ prb:4 ] ) to produce pictures in figure [ fig:5 ] are shown in table 1 .we believe that the ` trust - region - reflective ` algorithm can be optimized for real - time implementation .[ tab:1 ] @l*15@l algorithm&minimum value&runtime ( second ) + ` cvx`&0.0126 & 4818.4263 + ` trust - region - reflective`&0.0131&235.2102 + ` interior - point - convex`&0.2439&67.2561 + the numerical experiments confirm that minimizing the residual of the linearized eit equation under the constraint @xmath436 yields good approximations to the true conductivity change .the algorithm yields no artifacts and produces good shape reconstructions even under high levels of noise .we also expect the same results for the complete electrode model setting .all of the above arguments hold if we replace the frobenius norm in ( [ 29.10.1 ] ) by the operator norm. however , the numerical results with respect to the frobenius norm are nicer ( the last picture of figures [ fig:2 ] and the first picture of figure [ fig:5 ] ) .the popularly used reconstruction methods based on minimizing the usual linearized eit equation are simple and fast for real - time implementation and produce good reconstruction images . however , these methods have no rigorous convergence results , and the reconstruction images usually contain ringing artifacts . on the other hand ,monotonicity - based methods allow globally convergent implementation but usually produce bad images under high levels of noise or when real data are used .our method is a combination of the usual minimization problem of the linearized eit equation and the monotonicity - based method , which inherits most of the good properties of these two methods , such as stability under high noise and rigorous global convergence property . besides ,if the lower and upper bounds of the conductivity are known , all parameters of our method can be calculated a - priori . moreover , to the best of our knowledge , this is the first reconstruction method based on minimizing the residual which has a rigorous global convergence property .however , we would admit that our method requires the definiteness assumption , that is , the true conductivity should be either always bigger or always smaller than the reference conductivity over all the reference body . in this paper, we establish rigorously theoretical results of our method and provide a few numerical experiments in an idealistic setting , i.e. the continuum model setting . in future works , we will apply this method to more realistic models such as the shunt model or the complete electrode model , which are commonly used models in practice . the authors would like to thank marcel ullrich , dominik garmatter and janosch rieger for many fruitful discussions .mnm is also grateful to li - chang hung and professor giovanni alberti for their constant support .10 a. adler , j. h. arnold , r. bayford , a. borsic , b. brown , p. dixon , t. j. faes , i. frerichs , h. gagnon , y. grber , et al . : a unified approach to 2d linear eit reconstruction of lung images ., 30(6):s35 , 2009 .m. grant and s. boyd .graph implementations for nonsmooth convex programs . in v.blondel , s. boyd , and h. kimura , editors , _ recent advances in learning and control _ , lecture notes in control and information sciences , pages 95110 .springer - verlag limited , 2008 .n. t. thanh , l. beilina , m. v. klibanov , and m. a. fiddy .reconstruction of the refractive index from experimental backscattering data using a globally convergent inverse method ., 36(3):b273b293 , 2014 .

<S> in electrical impedance tomography , algorithms based on minimizing a linearized residual functional have been widely used due to their flexibility and good performance in practice . </S><S> however , no rigorous convergence results are available in the literature yet , and reconstructions tend to contain ringing artifacts . in this work </S><S> , we shall minimize the linearized residual functional under a linear constraint defined by a monotonicity test , which plays the role of special regularizer . </S><S> global convergence is then established to guarantee that this method is stable under the effects of noise . moreover , numerical results show that this method yields good shape reconstructions under high levels of noise without the appearance of artifacts . at ( current page.south ) ; </S>

1106.5685

observations of neutrino oscillations at the mass splittings ( @xmath1 ) of @xmath4 and @xmath5 are consistent with three generations of neutrinos and a unitary mixing matrix @xcite . however , the lsnd @xcite and miniboone @xcite experiments observe an excess of @xmath6 events from a @xmath7 beam , indicating possible new physics with @xmath1 @xmath8 .this could be explained by introducing additional generations of light neutrinos , and electroweak data @xcite require such additional light neutrino species to be sterile , i.e. with negligible couplings to @xmath9 and z bosons .sensitive searches for the disappearance of @xmath10 ( and @xmath7 ) into sterile neutrino species can constrain models used to explain the lsnd and miniboone @xmath6 appearance results . first , neutrino and antineutrino oscillations are expected to be either identical in such models or related via cp - violating phases @xcite or non - standard matter - like effects @xcite .second , tests of @xmath10 and @xmath7 disappearance ( @xmath11 ) probe elements in the neutrino mixing matrix that also govern @xmath12 appearance ( @xmath13 ) . as a result, a number of global analyses have been performed to study whether neutrino and antineutrino data sets relevant to the @xmath14 oscillation region in three different channels ( @xmath12 appearance , @xmath0 disappearance , @xmath12 disappearance ) can all be accommodated within sterile neutrino models @xcite .the miniboone collaboration has previously reported limits on @xmath10 and @xmath7 disappearance in this region of @xmath1 @xcite . substantial neutrino flux and interaction cross section uncertainties limited the sensitivity of the miniboone only analyses .data from the sciboone experiment can be used to reduce these errors because sciboone shares the same neutrino flux and target material .this paper describes the @xmath10 disappearance analysis using data from both experiments .the muon neutrino beam , with mean energy of 0.8 gev , interacts in the sciboone ( miniboone ) detectors at 100 m ( 541 m ) from the neutrino production target .neutrino disappearance at these distances would be observable as a distortion in the neutrino energy spectrum and as a deficit of the total event rate . at @xmath1 of @xmath15, the miniboone neutrino spectrum will have a deficit relative to sciboone , where the neutrinos have not yet oscillated . above @xmath8 , both the sciboone and miniboone neutrino energy spectrawill be altered due to disappearance .the fermilab booster neutrino beamline ( bnb ) provides the @xmath0 flux to both experiments .protons at kinetic energy of 8 gev interact with a 1.7 interaction length beryllium target .the mesons are then focused by a magnetic field and decay in a 50 m long tunnel to produce the 93.6% pure @xmath0 beam ( 5.9% @xmath16 , 0.5% @xmath17 ) .the bnb neutrino flux @xcite is simulated using geant4 @xcite and includes updated p - be particle production data @xcite .the sciboone experimental hall houses three sub - detectors ; in order from upstream to downstream they are a fully active and finely segmented scintillator tracker ( scibar ) , an electromagnetic calorimeter ( ec ) , and a muon range detector ( mrd ) . the scibar detector @xcite consists of 14,336 extruded polystyrene ( @xmath18 ) strips .the scintillators are arranged vertically and horizontally to construct a 3 @xmath19 3 @xmath19 1.7 m@xmath20 volume with a total mass of 15 tons which serves as the primary neutrino interaction target for sciboone .each scintillator strip is read out by a 64-channel multi - anode photomultiplier tube ( pmt ) via wavelength shifting fibers .the ec is a `` spaghetti''-type calorimeter ; 64 modules made of 1 mm scintillating fibers embedded in lead foil are bundled in 64 modules and read out at both ends by pmts . the mrd is built from 12 iron plates , each 5 cm thick , sandwiched between planes of 6 mm thick scintillation counters ; there are 13 alternating horizontal and vertical planes read out via 362 individual 2 inch pmts . a geant4 @xcite monte carlo ( mc ) simulationis used to model particle and light propagation through sciboone and the surrounding material .the miniboone detector @xcite is a mineral oil cherenkov light detector .it is a 12 m diameter spherical tank lined with 1280 inward - facing 8 inch pmts with an optically isolated outer region used to reject cosmic - ray induced events .neutrino interactions in the oil produce charged particles which cause emission of primarily cherenkov light and a smaller amount of scintillation light .a geant3 @xcite based simulation , tuned on miniboone and external data , is used to model light and particle production and propagation within the detector .neutrino interactions are simulated with the v3 nuance event generator @xcite , with cross section parameters set to match miniboone and sciboone data @xcite .prior to selection , approximately 44% ( 39% ) of all events in sciboone ( miniboone ) are due to charged current quasi - elastic ( ccqe ) scattering and 25% ( 29% ) to charged current single pion production ( cc@xmath21 ) ; the small difference in event fractions in miniboone and sciboone is due to the somewhat higher energy flux at miniboone .while all @xmath22 cc and nc interactions are sensitive to oscillation into sterile states , ccqe is the dominant interaction mode after selection cuts .the reconstructed neutrino energy ( @xmath23 ) is calculated assuming the neutrino interaction was ccqe , and assuming the target nucleon is at rest : @xmath24}\ ] ] where @xmath25 , @xmath26 indicates the muon , proton or neutron mass with appropriate subscripts , @xmath27 is the nucleon binding energy which is set to 34 mev , @xmath28(@xmath29 ) is the reconstructed muon energy ( momentum ) , and @xmath30 is the reconstructed muon scattering angle with respect to the neutrino beam direction .charged current ( cc ) neutrino interactions in sciboone are selected by identifying single muon tracks @xcite .the @xmath31 and @xmath32 of each muon track are reconstructed from the hits and energy loss in the sciboone sub - detectors .the highest momentum track in the beam on - time window is required to have @xmath33 to reduce the number of neutral current ( nc ) events .the energy loss of the track in scibar must be consistent with a muon hypothesis , and must originate within the 10.6 ton scibar fiducial volume ( fv ) .muon candidate tracks , where the endpoint of the track is also in the scibar fv , are classified as scibar - stopped tracks ; tracks which stop in the mrd constitute the mrd - stopped sample .tracks which exit the last layer of the mrd form the mrd - penetrated sample .figure [ efficiency_sb ] shows the event selection efficiency of each sciboone sample for cc interactions which occur within the scibar fv , as estimated by mc simulation .we estimate the selection cuts for the mrd - stopped sample are 18% efficient for all cc interactions , with 90% cc neutrino interaction purity ; the main sources of inefficiency stem from the detector volume acceptance and the momentum requirement that the events be contained in scibar or the mrd , or exit through the last plane of the mrd .the mrd - penetrated sample selection cuts are 4% efficient and produce a 97% pure @xmath0 cc sample .the scibar - stopped cuts are 12% efficient and the sample is 85% pure @xmath0 cc interactions .the mean neutrino energies of the selected events are 1.2 , 2.4 and 1.0 gev for mrd - stopped , mrd - penetrated and scibar - stopped samples , respectively .together , the sciboone samples cover the muon kinematic region ( @xmath29-@xmath30 ) relevant for events in miniboone @xcite .the selection criteria yield 13589 data events in the scibar - stopped sample , 20236 data events in the mrd - stopped sample , and 3544 events in the mrd - penetrated sample after subtracting cosmic ray events , for @xmath34 protons on target ( pot ) collected ; with @xmath35 , there are 13592 ( 20166 ) data events in scibar - stopped ( mrd - stopped ) samples .the @xmath23 resolution is 13% for ccqe events in the mrd - stopped sample , and 20% for scibar - stopped events ; the mrd - penetrated sample can only provide the muon angle for each event , not the neutrino energy . according to the simulation ,the scibar - stopped sample is 51% ccqe and 31% cc@xmath21 .the remaining events are from cc or nc multi - pion events in scibar .the mrd - stopped sample is 52% ccqe and 34% cc@xmath21 ; mrd - stopped sample background events also include interactions in the ec or mrd which send a muon backwards into scibar .the mrd - penetrated sample is 57% ccqe and 32% cc@xmath21 .the small background due to cosmic events is estimated from beam - off time windows .[ datamc_sb ] shows the @xmath23 spectrum for scibar - stopped and mrd - stopped data and prediction , assuming no oscillations , with cosmic background subtracted .both samples are divided into 16 bins for @xmath36 gev , 0.1 gev wide .the miniboone event selection is similar to the sciboone selection in that it is based on identifying a single muon in the detector @xcite .clusters of pmt hits in time are categorized as sub - events within a neutrino interaction event .the timing and charge response of the pmts is then used to reconstruct the position , kinetic energy , and direction vector of the primary particle within each sub - event .the muon produced from cc neutrino interactions is reconstructed as the first of two sub - events , with the decay electron as the second .exactly two sub - events are required and both sub - events must have minimal activity in the outer veto region . the first sub - event must be within the beam - on time window and have more than 200 inner tank pmt hits ( @xmath37 gev / c ) to eliminate electrons from stopped cosmic ray muon decays .the mean emission point of cherenkov light of the first sub - event must be within the 442 ton fiducial volume .the second sub - event must have fewer than 200 inner pmt hits and the reconstructed vertex must be within 100 cm of the endpoint of the first sub - event to be consistent with an electron from the ( first sub - event ) muon decay . requiring one decay electron eliminates most nc interactions and allowing only one decay electron removes most cc charged pion interactions .miniboone took neutrino events in two periods .the first period ( run i ) comprises @xmath38 pot , with 190,454 events with @xmath39 gev .the second period ( run ii ) was taken during sciboone s run , comprising @xmath40 pot and 29269 events for the same energy range .the two data sets are consistent with each other within the time dependent uncertainties , such as pot normalization , but are treated separately in the analysis , as these uncertainties cancel between miniboone and sciboone for run ii but not for run i. the miniboone selection is 35% efficient with 74% purity for ccqe events ; the resultant sample has average @xmath23 resolution of 11% for ccqe events @xcite . the rest of the @xmath0 sample is dominated by cc@xmath21 events ( @xmath4175% ) where the outgoing pion is unobserved ( due , e.g. , to absorption in the nucleus ) .[ datamc_mb ] shows the the @xmath23 spectrum for miniboone ( run i ) with the prediction , assuming no oscillations , in 16 bins , 0.1 gev wide except for the first bin ( 0 - 0.4 gev ) .[ cols="^,^,^,^,^,^,^",options="header " , ] systematic uncertainties are included for the underlying neutrino flux prediction , neutrino interaction cross section , and detector response .the method used to estimate the uncertainties due to the underlying neutrino flux prediction and the miniboone detector model is identical to the method used in previous miniboone results @xcite ; the cross section uncertainties and sciboone detector uncertainties assumed are identical to previous sciboone results @xcite .the dominant uncertainties on the neutrino event rate at sciboone and miniboone are the uncertainty of the neutrino flux due to production of @xmath42 mesons from the be target , and the ccqe and cc1@xmath43 neutrino cross sections .the detector uncertainties affect the muon energy and angular resolution . for sciboone ,these include uncertainties on the energy loss of muons through scintillator and iron , light attenuation in the wls fibers and pmt response ; for miniboone , detector uncertainties include particle and light propagation through the mineral oil , scattering , detection and pmt response .table [ tab : ratio ] shows the ratio of predicted to observed number of events for the sciboone and miniboone samples , and the size of the flux , cross section and detector uncertainties .each source of systematic uncertainty produces correlations between @xmath23 bins within a single data sample and between samples , e.g. between sciboone and miniboone .these correlations are represented by developing a covariance matrix , @xmath44 , which is the sum of the individual covariance matrices for each source of systematic uncertainty .each covariance matrix is constructed in the same manner as in previous sciboone cc inclusive analysis @xcite and miniboone oscillation analyses @xcite ; a fractional error matrix is generated using the mc expectations for the central value ( @xmath45 ) and many different sets for systematic variations ( @xmath46 ) for each @xmath47 bin : @xmath48 where @xmath49 denotes the index for each systematic variation and @xmath26 denotes the total number of variations .for the simultaneous fit the covariance matrix ranges over @xmath23 bins from scibar - stopped , mrd - stopped and miniboone samples .the sizes of the diagonal elements of the covariance matrix , @xmath50 are shown as the error bands on the predictions for the individual samples in figs .[ datamc_sb ] and [ datamc_mb ] , and also in figs .[ simfit_ratio ] and [ specfit_ratio ] .the correlation matrix component of the covariance matrix , @xmath51 , is shown in fig .[ ematrix_simfit ] . where there is a shared source of uncertainty ( e.g. , neutrino flux or cross section ) , miniboone bins are strongly correlated with sciboone bins .the correlations between sciboone and miniboone are exploited by the simultaneous fit @xmath52 , described in sec .[ sec : fit ] , to constrain the rate at miniboone .when a bin at miniboone is strongly correlated to one at sciboone , a difference between the data and prediction at miniboone will only increase the @xmath52 if the difference is not also present in the appropriate sciboone bin .the differences in the event rates in the two detectors due to flux or cross section systematics , unlike detector systematics , tend to cancel . as a result ,the largest uncertainty for the simultaneous fit analysis is the miniboone detector uncertainty .distribution for the simultaneous fit analysis .bins 0 - 15 and bins 16 - 31 are respectively for the scibar - stopped and the mrd - stopped samples from sciboone , shown in fig .[ datamc_sb ] . bins 32 - 47 are for the miniboone sample shown in the top panel of fig .[ datamc_mb ] ., height=302 ] in the spectrum fit method , uncertainties of neutrino flux and cross sections are constrained by the sciboone corrections .the systematic variations of the predicted number of events in miniboone after the sciboone corrections , @xmath53 , are calculated as : @xmath54 where @xmath55 and @xmath56 are the central value and the systematic variation ( sys ) , respectively , of the predicted number of events in sciboone , and @xmath57 is the systematic variation of the predicted number of events at miniboone for @xmath23 bin @xmath58 and incident neutrino energy bin @xmath59 .then , @xmath53 is used to construct the covariance matrix for the spectrum fit analysis for miniboone @xmath23 bins .the ratio of the observed number of events predicted for the miniboone sample and the size of the reduced systematic uncertainties after applying the sciboone corrections are shown in tab . [tab : ratio_afterfit ] . as a closure test, the sciboone corrections are re - applied to the sciboone mc samples and compared with the data , as shown in tab . [tab : ratio_afterfit_sciboone ] .the data to mc ratios for the sciboone samples are not exactly zero due to the small data and mc disagreement at the low @xmath60 region @xcite , however the differences are well within the systematic uncertainties .the fractional sizes of the diagonal elements of the covariance matrix are shown in fig .[ errors ] , and the correlation coefficients for the total systematic uncertainty after the sciboone corrections are shown in fig .[ ematrix_specfit ] for miniboone @xmath23 bins . after the sciboone corrections ,the flux and cross section uncertainties are reduced by a factor of approximately two as compared to the previous miniboone - only analysis .the spectrum fit method has uncertainties associated with the extraction of the @xmath61 rate factors , given in table [ tab : sb_spectrum ] , which are included in @xmath44 . as shown in tab .[ tab : ratio_afterfit ] , the dominant uncertainty for the spectrum fit analysis is the miniboone detector uncertainty which does not cancel in the ratio between the two experiments. bins for the spectrum fit analysis .the dashed lines show the uncertainties using miniboone alone , and the solid lines show the uncertainties after the sciboone corrections ., height=302 ] distribution for the spectrum fit analysis , shown in the right panel of fig .[ datamc_mb ] . , height=302 ] the disappearance search uses a frequentist - based @xmath62 approach @xcite to determine allowed regions in the @xmath63 plane .the @xmath52 is calculated from a comparison of the data , @xmath64 , in @xmath23 bin @xmath58 , to a prediction @xmath65 : @xmath66 where @xmath44 is the covariance matrix . as defined in section [ sec : pred ] , the simultaneous fit prediction @xmath67 has 48 bins and the spectrum fit prediction , @xmath68 has 32 bins .the covariance matrix @xmath44 is either 48 @xmath19 48 bins ( simultaneous fit ) or 32 @xmath19 32 bins ( spectrum fit ) . at each oscillation point in a grid across the region of interest , @xmath69 is calculated .the probability distribution for the @xmath62 statistic is determined from an analysis of a set of simulated data samples generated with neutrino flux , cross section , and detector systematics varied .both the simultaneous fit and spectrum fit are consistent with the null hypothesis .[ simfit_ratio ] and fig .[ specfit_ratio ] show the ratio of data to the null hypothesis and three oscillation scenarios for illustration , for the simultaneous fit and the spectrum fit , respectively .the @xmath52 value for the null hypothesis is 45.1 ( 48 degrees of freedom ( dof ) , 59% probability ) , with @xmath70 at @xmath71 , @xmath72 in the simultaneous fit analysis , fit to miniboone run i data .the expected @xmath62 for the 90% cl for the simultaneous fit is 9.34 , so the result is consistent with the null hypothesis .a fit to run ii data is consistent with a @xmath73 = 41.5 ( 48 dof , 73% probability ) ; a fit to both run i and run ii produces a negligible improvement on the simultaneous fit sensitivity as compared to fitting just the larger run i data set . for the spectrum fit to run i and runii , the @xmath52 found comparing the data to the null hypothesis is 41.5 ( 32 dof , 12% probability ) with the @xmath52 minimum value of 35.6 at @xmath74 , @xmath75 .the expected @xmath62 for the 90% cl is 8.41 , so the fit is consistent with the null hypothesis .figure [ limit ] shows the 90% cl sensitivity for the simultaneous fit analysis compared to the previous miniboone - only sensitivity .the 90% cl sensitivity is the average limit of 1000 fake experiments with no underlying oscillation assumed .both sensitivities use the miniboone run i data to show how the addition of sciboone data improves the sensitivity . in the miniboone - only fit , flux andcross section uncertainties are dominant at low @xmath1 ( @xmath76 ) , resulting in a reduced sensitivity in that region . in the simultaneous fit, the presence of flux and cross sections uncertainties does not increase the @xmath52 once sciboone data are included and the sensitivity improves . in the spectrum fit , which has a similar sensitivity as the simultaneous fit ,the uncertainties are directly reduced by the sciboone corrections and so the sensitivity is improved . the 90% cl limit for both analyses is also shown in fig .[ limit ] along with other relevant experimental results .while the limit is an improvement in the @xmath1 region of @xmath77 , the 90% cl limits are worse than the sensitivity at @xmath14 because of the small deficit at miniboone relative to sciboone .the energy dependence of this deficit is within the uncertainties , thus the limit is consistent with the region covered by 68% of the limits of fake null experiments .the spectrum fit limit is not as strong as the simultaneous fit limit because the corrected spectrum creates a larger deficit between miniboone data and prediction than the default prediction , while the simultaneous fit applies no correction and so is consistent with the small deficit . in summary, we search for @xmath0 disappearance using both sciboone and miniboone data sets . the two analyses presented reduce the flux and cross section uncertainties to the level of the detector uncertainties of the two experiments .we set 90% cl limits on @xmath0 disappearance in the @xmath1 region of @xmath2 , with an improvement over previous experimental results between @xmath78 and @xmath3 .we acknowledge the support of fermilab .we acknowledge the support of grants , contracts and fellowships from mext , jsps ( japan ) , the infn ( italy ) , the ministry of science and innovation and csic ( spain ) , the stfc ( uk ) , and the doe and nsf ( usa ) .we also acknowledge the use of condor software for the analysis of the data .34ifxundefined [ 1 ] ifx#1 ifnum [ 1 ] # 1firstoftwo secondoftwo ifx [ 1 ] # 1firstoftwo secondoftwo `` `` # 1''''@noop [ 0]secondoftwosanitize@url [ 0 ] + 12$12 & 12#1212_12%12@startlink[1]@endlink[0]@bib@innerbibempty link:\doibase 10.1088/0954 - 3899/37/7a/075021 [ * * , ( ) ] link:\doibase 10.1103/physrevd.64.112007 [ * * , ( ) ] , link:\doibase 10.1103/physrevlett.105.181801 [ * * , ( ) ] , link:\doibase 10.1016/j.physrep.2005.12.006 [ * * , ( ) ] , link:\doibase 10.1103/physrevd.75.013011 [ * * , ( ) ] , link:\doibase 10.1103/physrevd.77.033001 [ * * , ( ) ] , @noop link:\doibase 10.1103/physrevd.80.073001 [ * * , ( ) ] , @noop @noop link:\doibase 10.1103/physrevlett.103.061802 [ * * , ( ) ] , link:\doibase 10.1103/physrevd.79.072002 [ * * , ( ) ] , link:\doibase 10.1016/s0168 - 9002(03)01368 - 8 [ * * , ( ) ] link:\doibase 10.1140/epjc/ s10052 - 007 - 0382 - 8 [ * * , ( ) ] , link:\doibase 10.1103/physrevc.77.015209 [ * * , ( ) ] , link:\doibase 10.1103/physrevd.45.3906 [ * * , ( ) ] @noop link:\doibase 10.1103/physrev.137.b962 [ * * , ( ) ] link:\doibase 10.1103/physrev.179.1294 [ * * , ( ) ] link:\doibase 10.1016/0550 - 3213(72)90120 - 4 [ * * , ( ) ] @noop @noop * * , ( ) link:\doibase 10.1016/j.nuclphysbps.2004.11.225 [ * * , ( ) ] , link:\doibase 10.1016/j.nima.2008.10.028 [ * * , ( ) ] , @noop link:\doibase 10.1016/s0920 - 5632(02)01756 - 5 [ * * , ( ) ] , link:\doibase 10.1103/physrevd.83.012005 [ * * , ( ) ] , link:\doibase 10.1103/physrevlett.100.032301 [ * * , ( ) ] , link:\doibase 10.1103/physrevlett.98.231801 [ * * , ( ) ] , link:\doibase 10.1103/physrevlett.102.101802 [ * * , ( ) ] , link:\doibase 10.1103/physrevd.57.3873 [ * * , ( ) ] , link:\doibase 10.1103/physrevlett.52.1384 [ * * , ( ) ] link:\doibase 10.1016/0370 - 2693(84)90688 - 9 [ * * , ( ) ] link:\doibase 10.1103/physrevlett.107.011802 [ * * , ( ) ] ,

<S> the sciboone and miniboone collaborations report the results of a @xmath0 disappearance search in the @xmath1 region of @xmath2 . </S><S> the neutrino rate as measured by the sciboone tracking detectors is used to constrain the rate at the miniboone cherenkov detector in the first joint analysis of data from both collaborations . </S><S> two separate analyses of the combined data samples set 90% confidence level ( cl ) limits on @xmath0 disappearance in the @xmath2 @xmath1 region , with an improvement over previous experimental constraints between 10 and @xmath3 . </S>

1405.7665

recently the study of thermodynamics of black holes in ads spacetime has been generalized to the extended phase space , in which the cosmological constant is identified as a thermodynamic pressure and its variations are included in the first law of black hole thermodynamics @xcite .such studies are mainly motivated by the geometric derivation of the smarr relation @xcite , in which a term consisting of the pressure and its conjugate `` thermodynamic volume '' appears . once one takes the cosmological constant as thermodynamic pressure in the first law , the black hole mass should be explained as enthalpy rather than internal energy of the system . besides, the criticality associated with the pressure was discussed in the extended phase space @xcite . with the extended structure of the thermodynamic phase space , one can often find phase transitions and critical points of various kinds .for instance , an interesting phenomena of black hole reentrant phase transition is found recently in four - dimensional born - infeld - ads black hole @xcite , higher dimensional rotating ads black holes @xcite , kerr ads black holes @xcite , gb - ads black hole @xcite and the third order lovelock ads black holes @xcite , which is observed previously in multi - component fluids @xcite .it is also possible to take other parameters as novel dimensions of the thermodynamic phase space , e.g. the born - infeld parameter in the born - infeld - ads black holes @xcite and the gb parameter in gb - ads black holes @xcite .the reentrant phase transition includes both first and zeroth order phase transitions .however , the origin of zeroth order phase transition is still unclear .this makes it interesting to consider black hole systems containing only the zeroth order phase transition .the thermodynamics of the ads black holes in conformal gravity @xcite to be studied in this work provides precisely such an example . on the other hand , although taking the cosmological constant as a thermodynamic pressure is nowadays a common practice , such operations implicitly assume that gravitational theories which differ only in the values of the cosmological constants are considered to fall in the `` same class '' , with unified thermodynamic relations .the common excuse for doing this is that the classical theory of gravity may be an effective theory which follows from a yet unknown fundamental theory , in which all the presently physical constants " are actually moduli parameters that can run from place to place in the moduli space of the fundamental theory . given that the fundamental theory is yet unknown , it is preferable to consider the extended thermodynamics of gravitational theories involving only a single action , which requires that all variables included in the thermodynamical relations must be integration constants .it so happens that , in conformal gravity , the cosmological constant comes as an integration constant rather than as a bare parameter which appear explicitly in the classical action . in this paper, we consider the @xmath0 criticality of the static charged ads black hole in conformal gravity .there exists a single critical temperature , above which there are two zeroth order phase transition points . in the next section ,we first revisit the thermodynamics of static charged ads black hole in conformal gravity in the extended phase space .section 3 is devoted to the @xmath0 criticality , particular attention is paid toward the appearance of critical points and the zeroth order phase transitions .finally , some concluding remarks are given in the last section .we consider the static charged ads black hole in conformal gravity .the action of conformal gravity reads @xmath7 where the unusual sign in front of the maxwell term is inspired by critical gravity @xcite and is required by requiring that the einstein gravity emerges from conformal gravity in the infrared limit @xcite .the static black hole solution with ads asymptotics for conformal gravity is found in @xcite , which takes the form @xmath8 where @xmath9 and @xmath10 is the corresponding maxwell field .@xmath11 represents the line element of a @xmath12 maximally symmetric einstein space with constant curvature @xmath13 , with @xmath14 and @xmath15 , respectively .there are six parameters @xmath16 in the solution , five of which are integration constants , and the last one determines the spacial sectional geometry of the horizon .these parameters obey a constraint @xmath17 so there are actually only 5 independent parameters ( @xmath18 is to be considered to be determined by other parameters as in ) . except the discrete parameter @xmath19, the rest 4 parameters @xmath20 are related to conserved charges : electric charge , charge of massive spin-2 hair , enthalpy and pressure , respectively . however , at fixed charges @xmath21 , @xmath22 and @xmath19 , there still exist a discrete freedom in choosing the integration constant @xmath18 : @xmath23 under the point of view of taking the cosmological constant @xmath22 as a thermodynamic pressure @xmath24 the energy calculated by employing the noether charge associated with the time - like killing vector @xcite should be identified with the enthalpy @xmath25 of the gravitational system .it reads @xmath26 where @xmath27 denotes the largest real root of @xmath28 which corresponds to the event horizon black hole , and so there is no cosmological horizon in the solution .] , @xmath29 is overall coupling which is present in the action . because of the double - valuedness of the integration constant @xmath18, one immediately sees a double - valued behavior of the enthalpy .such behaviors are also present in the four dimensional charged rotating black hole @xcite and six dimensional static black holes @xcite of conformal gravity .one can expect that the temperature and gibbs free energy may also be double - valued .these double - valued variables must be all considered in order to have a holistic look at the thermodynamics in the extended phase space of the black hole .the thermodynamical conjugate of the pressure , i.e. the `` thermodynamic volume '' , is given by @xmath30 it can be seen that the sign of @xmath31 is determined by the sign of the parameter @xmath32 .besides @xmath6 and @xmath31 , all the other thermodynamic quantities are given in @xcite .the temperature is @xmath33 and its conjugate , i.e. the entropy is @xmath34 the electric charge and the conjugate potential are respectively @xmath35 the parameter @xmath36 is a massive spin-2 hair which is now taken as a novel dimension in the thermodynamic phase space .we label this novel dimension and its conjugate as @xmath37 : @xmath38 throughout this work , we will take the normalization @xmath39 .it can be checked that the first law of thermodynamics @xmath40 and the smarr relation @xmath41 hold in the extended thermodynamic phase space @xcite .the absence of @xmath42 and @xmath43 terms in the smarr relation can be explained by scaling arguments .the enthalpy can be viewed as a homogeneous function of the extensive variables @xmath44 , i.e. @xmath45 assuming each extensive variable has a scaling dimension which is denoted @xmath46 respectively . if the enthalpy @xmath25 itself has scaling dimension @xmath47 , then after a rescaling of the extensive variables we get @xmath48 taking the first derivative with respect to @xmath49 and then setting @xmath50 , we get @xmath51 after a little calculation , one can find that @xmath36 scales as [ length]@xmath52 ; @xmath18 scales as [ length]@xmath53 ; @xmath22 scales as [ length]@xmath54 which is the same with @xmath6 ; @xmath32 scales as [ length]@xmath55 ; @xmath56 scales as [ length]@xmath53 , which is the same with @xmath57 ; @xmath58 scales as [ length]@xmath53 ; @xmath25 scales as [ length]@xmath52 ; @xmath32 scales as [ length]@xmath55 , which results in the smarr relation , as expected .while considering critical behaviors , the gibbs free energy will play an important role .it is given as follows : @xmath59 on the other hand , the helmholtz free - energy can be obtained from the euclideanized action : @xmath60 a direct check yields @xmath61 this in turn justifies the explanation of @xmath25 as thermodynamic enthalpy . before proceeding , let us reveal a natural constraint of the black hole solutions . from @xmath62we get @xmath63 . on the other hand , from eq .( [ relation ] ) we get @xmath64 . inserting both into , one can obtain @xmath65 which leads to the constraint on @xmath5 and @xmath66 @xmath67 this constraint implies an upper bound of @xmath5 at fixed horizon radius @xmath66 , or an upper bound of @xmath66 at fixed temperature @xmath5 .to consider the @xmath0 criticality of the black hole , we should begin with the equation of state ( eos ) in @xmath0 plane at fixed conserved charges @xmath56 and @xmath36 .the eos arises from the expression for the temperature @xmath5 .however , to use as a reasonable eos , we need to eliminate the parameters @xmath18 and @xmath32 . assuming that both of these parameters are nonzero .then the condition @xmath62 yields @xmath68 inserting into eqs.([relation ] ) and ( [ temperature ] ) respectively , one gets @xmath69 then from eq.([eq2 ] ) , we find @xmath70 inserting this into eq.([eq1 ] ) , we get the eos @xmath71 solving this equation for @xmath6 we get @xmath72 and thanks to the condition ( [ tem ] ) , both branches of solutions should be considered physical . comparing the above expressions for the pressure with the van der waals equation @xmath73 we are tempted to use the variable @xmath74 as an effective specific volume for the black hole system under consideration .thus we rewrite the full eos as @xmath75 the critical points @xmath76 results from the conditions @xmath77 the partial derivative in and can be evaluated directly using , which read @xmath78 solving these two equations and substituting into , we get two sets of critical point parameters as follows . 1 . when @xmath79 , @xmath80 2 .when @xmath81 , @xmath82 in the above , @xmath83 can take two discrete values @xmath84 or @xmath85 : @xmath86 however , the physical critical point must obey the constraints @xmath87 , @xmath88 and @xmath89 , which lead to @xmath90 this exclude the choice @xmath81 and @xmath91 , and we are left with a single critical point which is characterized by the parameters @xmath92 in which @xmath93 .these parameters satisfy the relation @xmath94 which gives a pure number and is independent of all parameters .actually , this relation is very similar to the one of the van der waals system , which behaviors as @xmath95 at its critical point .however , if one replaces @xmath96 in eq.([pvt ] ) by the thermodynamic volume @xmath31 , the result of this relation will be no longer independent of parameters . in this sense, it is more natural to consider @xmath96 as the specific volume instead of the thermodynamic volume @xmath31 .note that the existence of critical point also exclude the possibility of @xmath97 .also , the square sum of the charge @xmath56 and the signature @xmath19 must also be nonzero .therefore , there is neither need to distinguish the case @xmath98 from the @xmath99 cases ( as long as @xmath100 ) nor need to distinguish the @xmath101 case from the @xmath102 cases ( as long as @xmath103 ) .one may also be curious about the cases with @xmath104 ( the bps black hole ) or @xmath105 , both of which will result in an eos of ideal gas after considering eq .( [ temperature ] ) directly .thus they are all out of our discussion .another degenerated case @xmath106 corresponds to the schwarzschild - ads black hole .in this case one can never find physical critical points as is known in @xcite .the isothermal plots at generic parameters @xmath107 are depicted in fig.[pv ] ( the right plot is a magnification of a single isotherm at the temperature @xmath108 ) .while creating the plots , we take @xmath109 and @xmath110 respectively as units for @xmath111 and @xmath2 .the pressure corresponding to the extremal specific volume is denoted @xmath112. it can be seen that on each isotherm there is an upper bound for the specific volume ( black hole radius ) @xmath113 . for @xmath114, the isotherm can be subdivided into two segments , i.e. the lower branch and the upper branch , which reflect the double - valuedness of the pressure .the difference between the @xmath115 and @xmath3 curves lies in that , each branch of the isotherm in the former case is monotonic with respect to the special volume @xmath96 , while the lower branch in the latter case is non - monotonic . consequently phase transitions will occur only in the @xmath3 regime .another way of subdividing the isotherms is according to the sign of @xmath116 , which is inversely proportional to the isothermal compressibility @xmath117 according to the sign of the isothermal compressibility , each isotherm with @xmath3 can be subdivided into 4 segments , two with positive isothermal compressibility and two with negative isothermal compressibility .let us take a closer look at the magnified plot given on the right diagram in fig.[pv ] .this is a curve corresponding to isotherm with @xmath108 .the lower and upper branches of the curve is joined together at the point d which corresponds to the extremal specific volume @xmath113 and the pressure @xmath118 .on the lower branch ( plotted in solid line ) one can see that there is a local maximum @xmath119 and local minimum @xmath120 for @xmath6 , the corresponding points on the isotherm are marked with b and c respectively .the 4 segments of the isotherm are @xmath121 and @xmath122 respectively . among these ,@xmath123 and @xmath124 have negative isothermal compressibilities which imply that black hole states falling in these segments may be unstable .the right plots in fig.[pv ] is typical for all @xmath3 . at such temperatures, we can subdivide the range of the pressure into 4 regimes : @xmath125 , @xmath126 , @xmath127 and @xmath128 . when @xmath129 , there is a single unstable black hole phase represented by the segment @xmath130 .due to its unstable nature , a perturbative increase in the pressure would result in an increase of the black hole radius until it reaches the state m , then through a phase transition it will enter the second regime for the pressure , @xmath126 . in this second regime for the pressure , there are 3 different black hole states for a single pressure value , among these , only the one with intermediate sized specific volume ( i.e. the state lying on the segment @xmath131 ) is stable .therefore , as the pressure increases , the black hole state will evolve from the state c until the state k is reached and then the pressure enters the next regime , @xmath127 .there are still 3 black hole states at each pressure values in this regime , however , two of these have positive isothermal compressibility ( i.e. the states lying on the @xmath132 and @xmath133 segments ) , so it is hard to tell which is more stable by looking at the eos alone .if the pressure enters the fourth regime , @xmath128 , then there is only a single stable phase which corresponds to states on the segment @xmath134 . from the above analysis, it is clear that there are two possible phase transition points , the first one occurs at @xmath135 , where the black hole will most probably transit from the state @xmath136 to the state @xmath137 which is more stable upon perturbation .the second phase transition point occurs either at @xmath118 ( if the gibbs free energy on the segment @xmath132 is higher than that on @xmath133 ) or at @xmath138 ( if the gibbs free energy on the segment @xmath132 is lower than that on @xmath133 ) . in the next subsection, it will be clear that the gibbs free energy on the segment @xmath132 is always lower than it is on @xmath133 , so the second phase transition point occurs at @xmath138 , where the black hole picks the state b instead of n , because the state @xmath139 has lower gibbs free energy than the state @xmath140 . in order to have a further look at the critical points , we need to plot the gibbs free energy versus pressure at fixed temperature .the gibbs free energy @xmath141 as a function of @xmath5 and @xmath6 is quite complicated and can not be given explicitly with ease .so we will try to present the @xmath142 relationship in terms of a pair of parametric equations .first we can invert the relation to get @xmath6 as a function of @xmath5 and other parameters .eliminating @xmath18 in the resulting expression by use of , we get @xmath143 where we have also replaced @xmath66 by @xmath144 . inserting this equation together with into, we get @xmath145 in and , @xmath32 can not be taken as a free parameter , because there is an extra constraint ( [ eq1 ] ) , which can be rewritten as @xmath146 this quadratic algebraic equation gives two solutions for @xmath32 as a function of @xmath147 .picking each solution and substituting into and , the resulting pair of equations can be taken as parametric definition for the function @xmath148 for constant @xmath5 .so , we can easily plot the @xmath142 diagram at constant @xmath5 , which is presented in fig.[gp ] . versus @xmath6 at constant temperatures .on the intersections with vertical axes , the temperature decreases from bottom to top , and the temperatures for each curve are in one to one correspondence to the isotherms given in fig.[pv ] . the right plot is a magnification of the lower - left corner of the ( red ) line with the highest temperature in the left plot .the marked points are the same as in the right plot of fig.[pv ] .the global minimum of the gibbs free energy in the right plot is highlighted by the thick line.,title="fig:",scaledwidth=45.0% ] versus @xmath6 at constant temperatures .on the intersections with vertical axes , the temperature decreases from bottom to top , and the temperatures for each curve are in one to one correspondence to the isotherms given in fig.[pv ] .the right plot is a magnification of the lower - left corner of the ( red ) line with the highest temperature in the left plot .the marked points are the same as in the right plot of fig.[pv ] .the global minimum of the gibbs free energy in the right plot is highlighted by the thick line.,title="fig:",scaledwidth=45.0% ] it can be seen in fig.[gp ] that , for all @xmath3 , a downcast swallow tail appears on each @xmath142 curve . from @xmath149 and downwards ,the swallow tail disappears , with @xmath149 corresponding to the critical point .this is different from that of a van der waals liquid - gas system , where @xmath115 is required for a phase equilibrium .the same phenomenon has also been observed in the study of criticality associated with the gb coupling constant @xcite in gauss - bonnet gravity .the downcast swallow tail is different from the usual upcast one , which corresponds to first order phase transitions @xcite .on the magnified plot given on the right plot of fig.[gp ] , it is clear that the gibbs free energy on the segment @xmath150 is lower than that on @xmath151 and @xmath152 . at @xmath135 and @xmath138 , there exist discontinuities for the gibbs free energy , indicating that there are zeroth order phase transitions at these two particular pressures .if one follows the red curve given in the left plot of fig.[gp ] , it would be clear that at sufficiently high pressure , the gibbs free energy on the dotted segment would eventually become lower than it is on the solid segment .this implies that the thermodynamics of the system favors large pressure states .on the other hand , if one looks at the gibbs free energy versus temperature plots at fixed pressure ( fig.[gt ] ) , one would see the similar downcast swallow tail at @xmath4 .moreover , following a single @xmath153 curve reveals that the system favors high temperature states if the pressure is kept fixed .in this paper , we considered the @xmath0 criticality of static charged ads black holes in conformal gravity . unlike the cases of einstein gravity ,the cosmological constant arises as an integration constant in conformal gravity , making the analysis for @xmath0 criticality more self - contained , e.g. without need to consider systems with different actions .the thermodynamics in the extended phase space for black hole in conformal gravity possesses several unusual features : * there exists only one critical point but there are two phase transition points , both of which corresponds to zeroth order phase transitions ; * there is no first order phase transition in the system ; * the phase transition can occur only when @xmath3 ( or @xmath4 ) but not the other way round ; * at fixed @xmath3 , the system favors large pressure states , whilst at fixed @xmath4 , the system favors high temperature states .we do not know of any other black hole or ordinary matter systems which exhibit similar thermodynamic behaviors. it would be interesting to find other examples which yield similar behaviors , because otherwise the system under study would seem to be too bizarre to understand .we would like to thank hong lu for suggesting this research .d. kastor , s. ray and j. traschen , `` enthalpy and the mechanics of ads black holes , '' class .* 26 * , 195011 ( 2009 ) [ ] .b. p. dolan , `` the cosmological constant and the black hole equation of state , '' class .grav . * 28 * , 125020 ( 2011 ) [ ] .b. p. dolan , `` pressure and volume in the first law of black hole thermodynamics , '' class .* 28 * , 235017 ( 2011 ) [ ] .b. p. dolan , `` compressibility of rotating black holes , '' phys .d * 84 * , 127503 ( 2011 ) [ ] .m. cvetic , g. w. gibbons , d. kubiznak and c. n. pope , `` black hole enthalpy and an entropy inequality for the thermodynamic volume , '' phys .d * 84 * , 024037 ( 2011 ) [ ] .b. p. dolan , d. kastor , d. kubiznak , r. b. mann and j. traschen , `` thermodynamic volumes and isoperimetric inequalities for de sitter black holes , '' phys .d * 87 * , no .10 , 104017 ( 2013 ) [ ] .d. kubiznak and r. b. mann , `` p - v criticality of charged ads black holes , '' jhep * 1207 * , 033 ( 2012 ) [ ] .s. gunasekaran , r. b. mann and d. kubiznak , `` extended phase space thermodynamics for charged and rotating black holes and born - infeld vacuum polarization , '' jhep * 1211 * , 110 ( 2012 ) [ ] .b. p. dolan , `` where is the pdv term in the fist law of black hole thermodynamics ? , ''a. belhaj , m. chabab , h. el moumni and m. b. sedra , `` on thermodynamics of ads black holes in arbitrary dimensions , '' chin .* 29 * , 100401 ( 2012 ) [ ] .s. h. hendi and m. h. vahidinia , `` p - v criticality of higher dimensional black holes with nonlinear source , '' phys .d * 88 * , 084045 ( 2013 ) [ ] .s. chen , x. liu , c. liu and j. jing , `` @xmath0 criticality of ads black hole in @xmath154 gravity , '' chin .lett . * 30 * , 060401 ( 2013 ) [ ] .r. zhao , h. -h .zhao , m. -s .ma and l. -c .zhang , `` on the critical phenomena and thermodynamics of charged topological dilaton ads black holes , '' eur .j. c * 73 * , 2645 ( 2013 ) [ ] .a. belhaj , m. chabab , h. el moumni and m. b. sedra , `` critical behaviors of 3d black holes with a scalar hair , '' [ ] . m. b. j. poshteh , b. mirza and z. sherkatghanad , `` phase transition , critical behavior , and critical exponents of myers - perry black holes , '' phys .d * 88 * , no . 2 , 024005 ( 2013 ) [ ] .n. altamirano , d. kubiznak and r. b. mann , `` reentrant phase transitions in rotating ads black holes , '' phys .d * 88 * , 101502 ( 2013 ) [ ] .r. -g .cai , l. -m .cao , l. li and r. -q .yang , `` p - v criticality in the extended phase space of gauss - bonnet black holes in ads space , '' jhep * 1309 * , 005 ( 2013 ) [ ] .a. belhaj , m. chabab , h. e. moumni , l. medari and m. b. sedra , `` the thermodynamical behaviors of kerrnewman ads black holes , '' chin .* 30 * , 090402 ( 2013 ) [ ] .n. altamirano , d. kubizk , r. b. mann and z. sherkatghanad , `` kerr - ads analogue of triple point and solid / liquid / gas phase transition , '' class .grav . *31 * , 042001 ( 2014 ) [ ] .n. altamirano , d. kubiznak , r. b. mann and z. sherkatghanad , `` thermodynamics of rotating black holes and black rings : phase transitions and thermodynamic volume , '' galaxies * 2 * , 89 ( 2014 ) [ ] .s. -w .wei and y. -x .liu , `` triple points and phase diagrams in the extended phase space of charged gauss - bonnet black holes in ads space , '' [ ] .d. kubiznak and r. b. mann , `` black hole chemistry , '' [ ] .d. -c .zou , s. -j .zhang and b. wang , `` critical behavior of born - infeld ads black holes in the extended phase space thermodynamics , '' phys .d * 89 * , 044002 ( 2014 ) [ ] .d. -c .zou , y. liu and b. wang , `` critical behavior of charged gauss - bonnet ads black holes in the grand canonical ensemble , '' [ ] . w. xu , h. xu and l. zhao , `` gauss - bonnet coupling constant as a free thermodynamical variable and the associated criticality , '' [ ] . h. xu , w. xu and l. zhao , `` extended phase space thermodynamics for third order lovelock black holes in diverse dimensions , '' [ ] j. li , h. -s .liu , h. lu and z. -l .wang , `` fermi surfaces and analytic green s functions from conformal gravity , '' jhep * 1302 * , 109 ( 2013 ) [ ] .h. lu and c. n. pope , `` critical gravity in four dimensions , '' phys .* 106 * , 181302 ( 2011 ) [ ] .h. lu , y. pang , c. n. pope and j. f. vazquez - poritz , `` ads and lifshitz black holes in conformal and einstein - weyl gravities , '' phys .d * 86 * , 044011 ( 2012 ) [ ] .h. -s .liu and h. lu , `` charged rotating ads black hole and its thermodynamics in conformal gravity , '' jhep * 1302 * , 139 ( 2013 ) [ ] .h. lu , y. pang and c. n. pope , `` black holes in six - dimensional conformal gravity , '' phys .d * 87 * , 104013 ( 2013 ) [ ] .

<S> the extended thermodynamics of static charged ads black holes in conformal gravity is analyzed . </S><S> the @xmath0 criticality of these black holes has some unusual features </S><S> . there exists a single critical point with critical temperature @xmath1 and critical pressure @xmath2 . at fixed @xmath3 ( or at fixed @xmath4 ) </S><S> , there are two zeroth order phase transition points but no first order phase transition points . </S><S> the systems favors large pressure states at constant @xmath5 , or high temperature states at constant @xmath6 . </S>

1506.06322

ranked set sampling ( rss ) is a powerful and cost - effective data collection technique that is often used to collect more representative samples from the underlying population when a small number of sampling units can be fairly accurately ordered without taking actual measurements on the variable of interest .rss is most effective when obtaining exact measurement on the variable of interest is very costly , but ranking the sampling units is relatively inexpensive .rss finds applications in industrial statistics , environmental and ecological studies as well as medical sciences . for recent overviews of the theory and applications of rss and its variationssee wolfe ( 2012 ) and chen et al .( 2004 ) . ranked set samplescan be either balanced or unbalanced .an unbalanced ranked set sample ( urss ) is one in which the ranked order statistics are not quantified the same number of times . to obtain an urss of size @xmath6 from the underlying population we proceed as follows .let @xmath6 sets of sampling units , each of size @xmath7 , be randomly chosen from the population using a simple random sampling ( srs ) technique .the units of each set are ranked by any means other than the actual quantification of the variable of interest .finally , one and only one unit in each ordered set with a pre - specified rank is measured .let @xmath8 be the number of measurements on units with rank @xmath9 , @xmath10 such that @xmath11 .suppose @xmath12 denotes the measurement on the @xmath13th unit with rank @xmath9 .the resulting urss of size @xmath6 from the underlying population is denoted by @xmath14 , where the elements of the @xmath9th row @xmath15 are independently and identically distributed ( i.i.d . ) from @xmath16 and @xmath17 is the df of the @xmath9th order statistic .moreover , @xmath12s are independent for @xmath18 and @xmath19 .note that if @xmath20 , @xmath18 , then urss reduces to the balanced rss .the df of urss is @xmath21 where @xmath22 and @xmath23 . as it is shown in chen et al .( 2004 ) , when @xmath24 , and @xmath25 , for @xmath26 , we have @xmath27 , where @xmath28 one can easily see that @xmath29 is not equal to the underlying df @xmath30 , unless @xmath31 , @xmath18 , showing that the edf based on the urss data does not provide a good estimate of the underlying distribution @xmath32 . the properties of the edf of the balanced and unbalanced rss are studied in stokes and sager ( 1988 ) as well as chen et al .( 2004 ) . in this paper, we use the empirical likelihood method as a nonparametric approach for estimating @xmath32 . to this end , we propose two methods to estimate @xmath32 using the exponentially tilted ( et ) technique .the proposed estimators can be used as standard tools for practitioners to estimate the standard error of any well - defined statistic based on rss or urss data and to make inferences about the characteristics of interest of the underlying population .another interesting problem in this direction is to develop efficient resampling techniques for urss data , as in many cases the exact or the asymptotic distribution of the statistics based on urss data are not available or they are very difficult to obtain ( e.g. , chen et al . , 2004 ) .akin to the methods of modarres et al .( 2006 ) and amiri et al .( 2014 ) , the new et estimators of @xmath32 are used to construct new resampling techniques for urss data .we study different properties of the proposed algorithms . for a hypothesis testing problem , about the underlying population mean , we show that the bootstrap tests based on the et estimators are asymptotically normal and exhibit a small bias of order @xmath0 which are desirable properties .the outline of the paper is as follows . in section [ et ] , we present et estimators of @xmath32 based on the urss data .section [ bt - urss ] considers two methods for resampling rss and urss data based on the et estimators of @xmath32 .we provide justifications for validity of these methods for a hypothesis testing problem about the population mean .section [ simulation ] describes a simulation study to compare the finite sampling properties of the proposed methods with parametric bootstrap and some existing resampling techniques for testing a hypothesis about the population mean .we consider both perfect and imperfect ranking scenarios , three different distributions and five rss designs .we compare the performance of our proposed methods with the one based on the empirical likelihood method studied in liu et al .( 2009 ) as well as baklizi ( 2009 ) . in section[ real data ] , we apply our methods for a testing hypothesis problem using a real data set consisting of the birth weight and seven - month weight of 224 lambs along with the mother s weight at time of mating .section [ concluding ] provides some concluding remarks .exponential tilting of an empirical likelihood is a powerful technique in nonparametric statistical inference .the impetus of this approach is the use of the estimated df subject to some constraints rather than the edf .et methods find applications in computation of bootstrap tail probabilities ( efron and tibshirani , 1993 ) , point estimation ( schennach , 2007 ) , estimation of the spatial quantile regression ( kostov , 2012 ) , bayesian treatment of quantile regression ( schennach , 2005 ) , small area estimation ( chaudhuri and ghosh , 2011 ) and calibration estimation ( kim , 2010 ) , among others .let @xmath33 be a generic sample of size @xmath6 from @xmath32 and suppose @xmath34 is the edf of @xmath35 which places empirical frequencies ( weights ) @xmath36 on each @xmath37 .consider an estimator @xmath38 of @xmath32 which assigns weights @xmath39 instead of @xmath36 to each @xmath37 . to obtain the et estimator of @xmath32 ,we minimize an aggregated distance between the empirical weights @xmath36 and @xmath40 subject to some constraints on the @xmath40 s . more specifically, one chooses a distance @xmath41 and minimizes @xmath42 subject to @xmath43 and some other constrains such as @xmath44 , using the following lagrangian multiplier method @xmath45 where @xmath46 is often imposed under the null hypothesis in a testing problem or any other conditions that one needs to account for in practice .note that the minimization in can also be done by minimizing the distance between @xmath47 and any target estimator @xmath48 other than the edf @xmath49 . the choice of the discrepancy function @xmath50 for the aggregated loss @xmath51 in leads to different et estimators of @xmath32 .since @xmath52 is the nonparametric maximum likelihood estimator of @xmath32 under the kullback - leibler distance subject to the restriction @xmath43 , one often uses @xmath53 we propose two et estimators of @xmath32 based on urss data with sample size @xmath54 where @xmath7 is the set size .the et estimators are then used to propose new bootstrapping algorithms from urss data . in this section ,we propose our first et estimator of @xmath32 which is later used to resample from within each row of @xmath55 .the idea behind the first et estimator of @xmath32 , for bootstrapping @xmath56 , is to find an estimator @xmath57 subject to the constraints @xmath58 where @xmath59 .[ lem1 ] let @xmath55 be a urss sample of size @xmath6 from the underlying population @xmath32 when the set size is @xmath7 and @xmath60 is the @xmath9-th order statistic in a simple random sample of size @xmath7 from @xmath32 .the optimum values of @xmath61 in under the constraints are given by @xmath62 where @xmath63 is obtained from @xmath64 .using the lagrange multipliers method , and by minimizing @xmath65 with respect to @xmath61 s , one can easily obtain the optimum values in .in section [ bt - urss ] , we use @xmath66 for bootstrapping @xmath67 instead of the commonly used empirical df .it is worth noting that for hypothesis testing problems about the underlying population mean @xmath68 involving the null hypothesis @xmath69 , minimization in is done subject to the condition @xmath70 . using the optimum weights @xmath71 from the et estimate of @xmath32, we also propose @xmath72 to estimate the population variance @xmath73 . by the structure of the urss data , @xmath56, we observe that @xmath74 are i.i.d .samples from @xmath75 , which is the distribution of the @xmath9-th order statistic in a simple random sample of size @xmath7 from @xmath32 . since @xmath76 the idea behind our next proposed et estimator of @xmath32 is to estimate each @xmath17 using @xmath74 , and construct an estimator of @xmath32 by averaging over these estimators using suitable weights obtained from the lagrange multipliers method under some constraints . to this end, we work with an estimator of @xmath32 of the form @xmath77 where @xmath78 is the edf of @xmath74 .[ lem2 ] let @xmath55 be a urss sample of size @xmath6 from @xmath32 where the set size is @xmath7 and @xmath79 are i.i.d .samples from @xmath17 the df of the @xmath9-th order statistic of a simple random sample of size @xmath7 from @xmath32 .then , an optimum estimator of @xmath32 in the form of under the constraints @xmath80 and @xmath81 , where @xmath82 , @xmath18 , is given by @xmath83 where @xmath63 is obtained from @xmath84 .the results easily follow using the lagrange multipliers method and minimizing @xmath85 with respect to @xmath86 .in section [ bt - urss ] , we use @xmath87 and propose a new bootstrapping algorithm to resample from @xmath67 instead of the commonly used empirical df . hereagain for hypothesis testing problems involving @xmath69 where @xmath68 is the population mean , minimization in is done subject to the condition @xmath88 . if for the observed urss data all the @xmath8s are large enough , then one can use et estimators of @xmath17 by simply treating @xmath12 s as a srs of size @xmath8 from @xmath17 and constructing the estimator @xmath89 for @xmath32 . here , @xmath90 and @xmath91sare obtained subject to constraints @xmath92 and @xmath93 , for @xmath18 , using the following lagrange multipliers problems : @xmath94in this section , we propose two new bootstrapping techniques to resample from a balanced or unbalanced ranked set sample of size @xmath6 .the first algorithm is based on the et estimator of @xmath32 in lemma [ lem1 ] to resample the entire urss while the second one uses the et estimator of @xmath32 in lemma [ lem2 ] to resample from within each row separately .we note that most of the bootstrap methods developed for rss are based on the edf and one can easily modify them using et estimators of @xmath32 .monte carlo simulation studies indicate that bootstrapping methods based on the et estimators of @xmath32 perform better than their counterparts using the edf . to resample from the et estimator of @xmath32 given by @xmath95 where @xmath96 is defined in we proceedas follows : * assign probability @xmath97 to each element @xmath12 of @xmath56 .* randomly draw @xmath98 from @xmath56 according to probabilities @xmath99 , order them as @xmath100 and retain @xmath101 . * repeat step 2 , for @xmath102 and @xmath103 to generate a bootstrap urss @xmath104 . * repeat steps 2 - 3 , @xmath105 times to obtain the bootstrap samples .one can easily validate the use of the et estimator of @xmath32 for different bootstrapping purposes .for example , suppose we want to carry out a bootstrap test for testing @xmath69 against @xmath106 , where @xmath68 is the unknown parameter of interest .using hall ( 1992 ) , the edgeworth expansion of the @xmath107-value for testing @xmath108 against @xmath109 based on a srs of size @xmath110 from the underlying population with the test statistic @xmath111 , is given by @xmath112 where @xmath113 is a quadratic function and @xmath114 and @xmath115 are the standard normal distribution and density functions , respectively .we consider the problem for a balanced rss case , as the following argument can also be applied to urss data with some modifications .let @xmath116 be a balanced ranked set sample of size @xmath110 from the underlying population with mean @xmath68 .we show that the et bootstrap approximation of the sampling distribution of @xmath117 is in error by only @xmath118 and the @xmath107-value obtained through the eat method has the desirable second order accuracy this is similar to results obtained in diciccio and romano ( 1990 ) . for more detailssee efron ( 1981 ) and feuerveger et al .( 1999 ) .suppose @xmath119 is a bootstrap sample generated from the eat algorithm .let @xmath120 be the bootstrap test for testing @xmath69 with p - value @xmath121 , where @xmath122 is the mean of the bootstrap sample obtained form the et estimator of @xmath32 and @xmath123 with @xmath124 .then , @xmath125 where @xmath126 , given by , is the p - value of the usual @xmath117-test based on a simple random sample of comparable size @xmath110 from the underlying population . for simplicity ,we write the resampled data as @xmath127 . in order to test @xmath69 , and to ensure that the null hypothesis is incorporated into the et estimator of @xmath32, we introduce the lagrange multipliers for the constraints @xmath128 and @xmath129 , where the weights @xmath130 are obtained as @xmath131 and @xmath132 is the coefficient calculated from @xmath133 .one can easily show that @xmath134s are generated from @xmath135 where @xmath136 . to obtain the et estimator of @xmath32 under the null hypothesis we must have @xmath137 therefore , one can use the bootstrap test statistic @xmath120 for testing @xmath69 where @xmath122 is the mean of the bootstrap sample obtained form the et estimator of @xmath32 and @xmath123 with @xmath124 . following hall ( 1992 ) and using the edgeworth expansion ,the @xmath107-value for testing @xmath69 against @xmath138 using the bootstrap test statistic @xmath139 is given by @xmath140 where @xmath141 is a quadratic function .now , the results follows from . the idea behind this method is to use the et estimator of @xmath32 given by @xmath142 where @xmath143 is defined in .to this end we proceed as follows : * assign probabilities @xmath143 to each row @xmath144 of @xmath56 , @xmath18 .* select a row randomly using @xmath143 and select an observation randomly from that row . *continue step 2 for @xmath7 times to obtain @xmath7 observations .* order them as @xmath145 and retain @xmath101 * perform steps 24 for @xmath8 and obtain @xmath146 . * perform steps 25 for @xmath102 . * repeat steps 26 , @xmath105 times to obtain the bootstrap samples .in this section , we compare the finite sample performance of out nonparametric eat and ear resampling methods with a parametric bootstrap ( pb ) procedure .the pb method uses a parametric test ( pt ) with an asymptotic normal distribution to test the hypothesis @xmath69 , where @xmath68 is the unknown parameter of interest and @xmath147 is a known constant .the resampling is performed using b=500 resamples and the entire experiment is then replicated 2000 times .we use several rss and urss designs with different sample sizes when the set size is chosen to be @xmath148 .we also conducted unreported simulation studies for other values of @xmath7 and we observed similar performance that we summarize below . the rss designs that we consider are written as @xmath149 with @xmath150 .for example , the first design is balanced with @xmath148 and @xmath151 observations per stratum , which is denoted by @xmath152 similarly , we define the following designs , @xmath153 we obtain samples from the normal(0,1 ) , logistic(1,1 ) and exponential(1 ) distributions . we first proceed with the following proposition . [ p - bt1 ] suppose @xmath32 is the df of the variable of interest in the underlying population with @xmath154 .let @xmath155 be the edf of the @xmath156 row of a balanced rss data and @xmath68 represent the population mean .then @xmath157 , with @xmath158 , converges in distribution to a multivariate normal distribution with the mean vector zero and the covariance matrix @xmath159 where @xmath160 and @xmath161 .this proposition suggests to use the following test statistic for testing the hypothesis @xmath69 @xmath162 where @xmath163 the test statistic @xmath164 , which is approximately normal@xmath165 for large @xmath7 , is referred to as the pt in the rest of the work .ahn et al .( 2014 ) consider the welch - type ( wt ) approximation to the distribution @xmath164 , where the degree of freedom of the test can be approximated using @xmath166 the nonparametric bootstrap tests using the eat and ear methods are conducted based on the following steps : * let x be an urss / rss sample from @xmath32 . *calculate @xmath167 , given in ( [ urssa5 ] ) , under the null hypothesis @xmath69 .* apply each of the resampling procedures on @xmath168 to obtain @xmath169 . *calculate @xmath170 , @xmath171 . *obtain the proportion of rejections via @xmath172 to estimate the @xmath107-value .we also performed the desired testing hypothesis using pb by generating urss samples from normal(0,1 ) , logistic(1,1 ) and exponential(1 ) distributions . to perform pb test we use the following steps ( for more details on pb method see efron and tibshirani ( 1993 ) ) : *let x be a urss sample from a distribution @xmath173 where @xmath174 is the unknown parameter and let @xmath175 .* calculate @xmath176 , under the null hypothesis @xmath69 .* estimate @xmath174 from x and take a urss from @xmath177 , @xmath169 . *calculate @xmath178 .* obtain the proportion of rejections via @xmath179 to estimate the @xmath107-value . to conduct the parametric bootstrap we estimated the population mean using the sample mean and used @xmath180 .subsequently , we generated samples from the n(@xmath181 , 1 ) , logistic(@xmath181 , 1 ) and exponential(@xmath181 ) distributions .2observed_alpha - normal ] displays the observed @xmath182 levels . the parametric bootstrap ( pb )method is accurate and the estimated @xmath182 levels are close to the nominal level 0.05 .the pt test is liberal and its approximated @xmath107-value is higher than the nominal level , specially under exponential distribution .we observe that the wt test is a bit conservative under the normal and logistic distributions , i.e. , the approximated @xmath107-values are lower than the nominal level . the observed @xmath182 levels for ear follow the pb method closely and they are less liberal than the pt under the exponential distribution ..observed @xmath182-levels of the proposed tests for testing @xmath183 under the normal distribution and @xmath184 for the exponential and logistic distributions .[ cols="<,^,^,^,^,^,^,^,^ " , ] [ 2observed_alpha - real2 ]we propose nonparametric estimators of the cumulative distribution of a continuous random variable using the et empirical likelihood method based on ranked set sampling designs .the et df estimators are used to construct new resampling techniques for urss data .we study different properties of the proposed algorithms . for a hypothesis testing problem ,we show that the bootstrap test based on exponential tilted estimators exhibit a small bias of order @xmath0 , which is a very desirable property .we compared the performance of our proposed techniques with those based on empirical likelihood .the latter are developed under the balanced rss assumption and they are not applicable for urss situation .the results of the simulation studies as well as a real data application show that the method based on et estimators of the df perform very well even for moderate or small sample sizes .we gratefully acknowledge the constructive comments of the referees and the associate editor .the research of m. jafari jozani was supported by the nserc of canada .the research of r. modarres was supported in part by the national institute of health , under the grant no .1r01gm092963 - 01a1 .amiri , s. and jafari jozani , m. , modarres , r. ( 2014 ) .resampling unbalanced ranked set sampling with application in testing hypothesis about the population mean . _ journal of agricultural , biological , and environmental statistics _ , 19 , 117 .chaudhuri , s. and ghosh , m. ( 2011 ) .empirical likelihood for small area estimation ._ biometrika _ , vol .98 , issue 2 , 473 - 480 .chen , z. , bai , z. and sinha , b.k .( 2004 ) . ranked set sampling : theory and applications .springer - verlag , new york .efron , b. ( 1981 ) .nonparametric standard errors and confidence intervals ( with discussion ) .j. statist ._ , 9 , 139172 .efron , b. and tibshirani , r. ( 1993 ) .an introduction to bootstrap .chapman & hall , new york .frey , j. , ozturk , o. & deshpande , j.v .nonparametric tests for perfect judgment rankings .102 , 708 - 717 .frey , j. , & wang , l. ( 2013 ) .most powerful rank tests for perfect rankings .computational statistics & data analysis , 60 , 157 - 168 .patil , g. p. , sinha , a. k. and taillie , c. ( 1999 ) . ranked set sampling : a bibliography . _ environmental and ecological statistics _, 6 , 9198 .schennach , s. m. ( 2005 ) .bayesian exponentially - tilted empirical likelihood ._ biometrika _ , 92 , 31 - 46 .schennach , s. m. ( 2007 ) .point estimation with exponentially tilted empirical likelihood ._ annals of statistics _ , vol .35 , no . 2 , 634 - 672 . stokes , s. l. , and sager , t. w. ( 1988 ) .characterization of a ranked - set sample with application to estimating distribution functions ._ journal of the american statistical association _, 83(402 ) , 374 - 381 .

<S> we study nonparametric estimation of the distribution function ( df ) of a continuous random variable based on a ranked set sampling design using the exponentially tilted ( et ) empirical likelihood method . </S><S> we propose et estimators of the df and use them to construct new resampling algorithms for unbalanced ranked set samples . </S><S> we explore the properties of the proposed algorithms . for a hypothesis testing problem about the underlying population mean , </S><S> we show that the bootstrap tests based on the et estimators of the df are asymptotically normal and exhibit a small bias of order @xmath0 . </S><S> we illustrate the methods and evaluate the finite sample performance of the algorithms under both perfect and imperfect ranking schemes using a real data set and several monte carlo simulation studies . </S><S> we compare the performance of the test statistics based on the et estimators with those based on the empirical likelihood estimators . </S><S> + draft of * exponentially titled empirical distribution function * + * for ranked set samples * saeid amiri@xmath1 , mohammad jafari jozani@xmath2 , and reza modarres@xmath3 _ @xmath4 department of statistics , university of nebraska - lincoln , lincoln , nebraska , usa _ _ @xmath5 department of statistics , university of manitoba , winnipeg , mb , canada , r3 t 2n2 _ _ @xmath3 department of statistics , the george washington university , washington dc , usa _ * keywords * : distribution function ; exponential tilting ; imperfect ranking ; ranked set sample . </S>

1210.5508

the importance of event - by - event fluctuations in the initial conditions context of heavy - ion collisions was first pointed out by miller and snellings who added nucleon position fluctuations to existing smooth ( i.e. ensemble - averaged ) glauber initial conditions @xcite .these fluctuations explain the experimental observation of non - vanishing anisotropic flow in central cu - cu and au - au collisions @xcite and of odd flow harmonics @xmath2 @xcite .recently attention has turned to a new source of event - by - event fluctuations , namely fluctuations in the transverse distribution of color charge within the colliding nucleons .these fluctuations are evidenced by large multiplicity fluctuations observed in minimum bias @xmath3 collisions and suggest that sub - nucleonic degrees of freedom may also play an important role in determining the event - by - event geometry of the initial state in nucleus - nucleus collisions @xcite .in this work , we develop a toy model for imprinting sub - nucleonic fluctuations on the transverse energy density profiles produced in relativistic heavy - ion collisions .the study is motivated by recent work of mller and schfer in which they calculate the mean normalized covariance function @xmath4 $ ] for the transverse energy density fluctuations of gluon fields in central au - au collisions at 200 gev @xcite .the authors approximate the collision system by two infinite slabs of nuclear matter with fixed gluon saturation momentum @xmath1 for which they take the value corresponding to the nuclear thickness function in the center of a central au - au collision .we texture a given transverse energy density profile with additional fluctuations using a turning band gaussian random field simulator ( tbsim ) @xcite that includes several configurable covariance functions .the mller - schfer covariance @xmath4 = ( \delta \epsilon(r)/ \epsilon_0)^2 $ ] is well described by the cauchy covariance included in tbsim , @xmath5=c(1+(r / a)^2)^{-b}\ ] ] with fit paramaters @xmath6 , @xmath7 fm and @xmath8 .using tbsim we generate a large gaussian random field ( grf ) with the desired covariance on a @xmath9 lattice with grid spacing @xmath10fm .the left panel in fig .[ f1 ] shows a small slice of this grf ; in the right panel its two - point correlation function is compared to the mller - schfer covariance and its cauchy fit .10fm section of the tbsim gaussian random field for the normalized energy density @xmath11 with zero mean and a cauchy covariance fitted to the mller - schfer covariance ._ right : _ two - point covariance of the grf as a function of distance @xmath12 at various azimuthal angles @xmath13 ( symbols ) , compared with the mller - schfer function ( solid ) and its cauchy covariance fit ( dotted ) .[ f1 ] , title="fig:",width=211 ] 10fm section of the tbsim gaussian random field for the normalized energy density @xmath11 with zero mean and a cauchy covariance fitted to the mller - schfer covariance ._ right : _ two - point covariance of the grf as a function of distance @xmath12 at various azimuthal angles @xmath13 ( symbols ) , compared with the mller - schfer function ( solid ) and its cauchy covariance fit ( dotted ) .[ f1 ] , title="fig:",width=249,height=181 ] due to the tail of the gaussian distribution , there is a @xmath14 chance that fluctuations about the mean fall into the unphysical region of negative energy density .these unphysical fluctuations can be eliminated by mapping the gaussian random variable onto a positive definite negative binomial distribution ( nbd ) .there is a long history of modeling fluctuations in @xmath3 collisions with negative binomial distributions ( for recent examples see @xcite ) .writing @xmath15 where @xmath16 is the sampled value , @xmath17 is its mean and @xmath18 controls its variance , we identify @xmath19 with @xmath20 . in the limit of large @xmath17 ( we take @xmath21 ) , the nbd becomes a continuous function @xmath22 of the reduced variable @xmath23 whose width parameter @xmath18 we adjust such that its variance @xmath24 agrees with the squared gaussian width @xmath25|_{r=0}{\,=\,}(\delta\epsilon(0)/\epsilon_0)^2{\,=\,}(0.684)^2 $ ] .the mapping is now achieved by replacing each value of @xmath26 from the gaussian random field by a new value @xmath27 such that the cumulative gaussian distribution function at @xmath28 coincides with the cumulative nbd distribution at @xmath29 : @xmath30 the resulting negative binomial random field @xmath31 is positive definite and retains the two - point covariance embedded in the original grf . a small section of this field and its two - point correlation function are shown in fig . [ f2 ] . ,but for the nbd random field .note that in the left panel the plotted quantity @xmath11 is now @xmath32 everywhere .[ f2 ] , title="fig:",width=211 ] , but for the nbd random field .note that in the left panel the plotted quantity @xmath11 is now @xmath32 everywhere .[ f2 ] , title="fig:",width=249,height=185 ] to imprint these energy density fluctuations onto the initial conditions for a heavy - ion collision , we take the transverse energy densities @xmath33 for events generated with the mc - kln model @xcite and multiply them with the nbd random @xmath31 field taken from randomly selected and appropriately sized sections of the final @xmath34 field grid : @xmath35this texturing procedure was applied to @xmath36 mc - kln au - au events at @xmath37gev ( using the mc - kln code with gaussian nucleons of width @xmath38fm ) , partitioned into equally sized bins in the number of participants of width @xmath39 . for both the textured and untextured events we compute the harmonic eccentricity coefficients @xmath40 using the definition @xmath41 andaverage them over the event ensemble .the ratios of these averages are plotted in fig .[ f3 ] for the harmonics @xmath42 as functions of @xmath43 . in @xmath44gev au - au events , for a correlation length @xmath45fmas used in @xcite ( left panel ) and a value @xmath46fm as large as the nucleon size ( right panel ) .[ f3 ] , title="fig:",width=226 ] in @xmath44gev au - au events , for a correlation length @xmath45fm as used in @xcite ( left panel ) and a value @xmath46fm as large as the nucleon size ( right panel ) .[ f3 ] , title="fig:",width=226 ] for the nominal correlation length @xmath45fm , shown in the left panel , we see that the gluon - field fluctuations induce only a small increase in the eccentricity harmonics @xmath40 , reaching 5 - 10% in central collisions and falling off in more peripheral ones .we should note , however , that our fluctuation texture assumes constant ( i.e. position - independent ) @xmath47 or @xmath48 , with a value expected ( on average ) in the center of central au - au collisions .more realistically , @xmath49 should vary inversely with the position - dependent nuclear thickness function ( which , on average , is largest in the fireball center ) : @xmath50 consequently , the mller - schfer texture with constant @xmath45fm provides an approximate _ lower _ bound on the eccentricity enhancement caused by sub - nucleonic fluctuations in @xmath44gev au - au collisions . to obtain an estimated upper bound on the eccentricity enhancement caused by sub - nucleonic fluctuations , we inflate the correlation length @xmath48 to the radius of our gaussian nucleons , @xmath51fm and repeat the texturing procedure .( this amounts to reducing @xmath47 by about a factor 2 . ) as seen in the right panel of fig .[ f3 ] , with the larger correlation length the gluon fluctuations increase the eccentricities @xmath40 by larger factors , reaching now 20 - 25% in central collisions ( and again falling off in peripheral ones ) . for both values of the correlation length , the sub - nucleonic fluctuation effects on @xmath40 are strongest in central collisions; in more peripheral collisions , fluctuations in the nucleon positions dominate the fluctuation effects on @xmath40 .to summarize , we generated a toy model to analyze the effects of sub - nucleonic color fluctuations on the centrality dependent eccentricty harmonics @xmath40 for mc - kln initial conditions .while we qualitatively confirm earlier findings @xcite that such fluctuations tend to increase the @xmath40 , only relatively small enhancements ( smaller than those reported in @xcite ) are found for realistic values of the gluon field correlation length .correlations over larger distances generate larger eccentricities . due to the assumption of a position - independent saturation momentum @xmath47 or correlation length @xmath48 ,our implementation of sub - nucleonic fluctuations is much less realistic than the one in the ip - glasma model of refs .@xcite ; it has , however , the advantage of allowing us to turn the sub - nucleonic fluctuations on and off at will and thus to study their effects on @xmath40 in isolation .* acknowledgements : * this work was supported by the u.s .department of energy under grants no .de - sc0004286 and ( within the framework of the jet collaboration ) de - sc0004104 , and by the ohio supercomputer center .b. alver and g. roland , phys .c * 81 * , 054905 ( 2010 ) .p. tribedy and r. venugopalan , nucl .a * 850 * , 136 ( 2011 ) ; phys .b * 710 * , 125 ( 2012 ) .b. schenke , p. tribedy and r. venugopalan , phys .lett . * 108 * , 252301 ( 2012 ) ; phys .c * 86 * , 034908 ( 2012 ) .a. dumitru , y. nara , phys .c * 85 * , 034907 ( 2012 ) .

<S> we have developed an algorithm to imprint quantum fluctuations onto the initial transverse energy density profile according to a given two - point covariance function . using as an example mc - kln initial conditions with added fluctuations satisfying the covariance function derived in @xcite </S><S> , we find that effects from sub - nucleonic gluon field fluctuations on the eccentricity harmonics @xmath0 vary strongly with the gluonic correlation length controlled by the saturation momentum @xmath1 . </S><S> varying @xmath1 over the range probed in au+au collisions at rhic , we find gluon fluctuation induced enhancements of the eccentricity coefficients ranging from 10 to 20% in central collisions . </S>

1008.5341

despite the pioneering work of @xcite on pulsar magnetospheres created by a rotating aligned dipole and of @xcite on relativistic winds both of which have been further explored in @xcite , there are few exact analytic solutions in relativistic magnetohydrodynamics ( mhd ) .prendergast s ( 2005 ) solution for a point explosion of force - free magnetic energy was the foundation of our recent work on that theme @xcite but recent advances have been mainly computational .the early axisymmetric static solution by @xcite has been followed by accurate computations of @xcite and @xcite which give solutions for the steady aligned pulsar while @xcite considered both time - dependent instabilities at the edge of the corotating zone and the non - aligned case .@xcite studied the time dependent relativistic problem of both the aligned and the oblique rotator ; in the case of the aligned rotator he confirmed the results of @xcite who have studied the time - independent problem , whereas for the oblique rotator he presented the three - dimensional structure of the magnetosphere .recent evidence that the pulses are generated near the light cylinder reinforce ideas put forward by @xcite who considered the orthogonally aligned cylindrical pulsar .numerical solutions more relevant to active galactic nuclei have been developed by @xcite and @xcite .more recently @xcite have studied the problem of a spinning monopole , and they found that the combined effect of rotation and the magnetic field provide sufficient conversion of magnetic energy to kinetic with profound applications to grbs but also to pulsar systems .force - free fields have been used as an approximation for pulsar magnetospheres . leaving a magnetic field configuration to relax, we take a final state of a force - free magnetic field .taking into account that the pulsar magnetosphere is dominated by the electromagnetic fields and considering that the effects of inertia are weaker makes the force - free assumption a valid starting point .the non - relativistic force - free approximation requires that the electric current flows along the magnetic field lines but neglects any effect of the electric field on the charges . in relativistic mhdthe electric field is no more negligible as the electric field , in non - resistive mhd , is given by the cross product of the magnetic field with the velocity scaled to the speed of light .thus the relativistic force - free approximation requires that the force of the magnetic field on the currents and that of the induced electric field on the charges add to zero . herewe explore the simplest exact relativistic force - free fields , those with cylindrical symmetry , and their coupling to rotating magnetospheres which corotate out to the light cylinder .while these solutions are very limited they allow us to answer questions such as : when one end of a cylinder of force - free field is rotated at relativistic speed , is its flux per unit area greater than or less than the static field that surrounds it in pressure equilibrium , i.e. if we start rotating one end of a tube of flux , does the twist generated in it lead to its expansion or does its cross - section remain the same or even contract ?what if anything happens as the rotation velocity of the tube approaches the speed of light ? how rapidly does the rotation advance along the tube ?for non - relativistic mhd this problem was treated in detail in @xcite but inevitably that has little bearing on light cylinders .we are studying idealized mhd in which the effects of resistivity are absent , therefore field lines that join the source of twist rotate while those that do not are left static .thus a discontinuity is no surprise . in section 2we give the basic equations and solve them to find the general steady cylindrical force - free fields .section 3 explores some simple special cases in greater detail , while section 4 gives the time - dependent exact solution for rotationally generated alfvn waves traveling down a uniform field .section 5 considers how a rotating star or a pulsar might generate such waves in the interstellar magnetic field .it also presents the conundrum of how a pulsar s field that extends beyond the light cylinder can connect to an anti - aligned interstellar field and suggests a possible resolution .we take the magnetic field and the velocity of the fluid to lie on cylinders and in cylindrical polars @xmath0 so @xmath1 and @xmath2\ , , \label{magnetic}\end{aligned}\ ] ] @xmath3\ , .\label{velocity}\end{aligned}\ ] ] the fluid is perfectly conducting so @xmath4\ , , \label{electric}\end{aligned}\ ] ] since @xmath5 is independent of @xmath6 and @xmath7 it follows that @xmath8 so @xmath9 is time independent .the charge density , @xmath10 , is given by @xmath11\ , .\label{charge}\end{aligned}\ ] ] provided @xmath12 is independent of time the current density , @xmath13 , is given by @xmath14 the force - free condition is @xmath15 using the expressions for @xmath10 and @xmath13 above , this becomes @xmath16^{2}-\frac{1}{2}\frac{\partial}{\partial r}b_{z}^{2}-\frac{1}{2r^{2}}\frac{\partial}{\partial r}(rb_{\phi})^{2}=0\ , .\label{force2}\end{aligned}\ ] ] we notice that only the component of @xmath12 perpendicular to @xmath9 plays any role in equation ( [ force2 ] ) .it is then convenient to write @xmath12 in the form @xmath17 , where @xmath18 is the unit vector along @xmath9 .we note that @xmath19 is not the component of @xmath12 along @xmath9 which contains also @xmath20 .thus any of our velocity fields may be considered as a ( differential ) rotation with flow along @xmath9 superposed .since the the flow along @xmath9 does not contribute to the @xmath21 term in ( [ force2 ] ) that equation becomes @xmath22 integrating the last term by parts and then dividing by @xmath23 we find @xmath24 where @xmath25 is the area average of @xmath26 up to @xmath27 .evidently if @xmath28 and @xmath29 are specified such that the right hand side of ( [ bf ] ) is non - negative , then ( [ bf ] ) determines @xmath30 ( up to a sign ) . with @xmath9 and @xmath31then known , @xmath32 follows from ( [ electric ] ) , @xmath10 from ( [ charge ] ) and @xmath13 from ( [ current ] ) .thus the whole electromagnetic configuration is determined but to find the flow we must also specify the velocity @xmath33 along @xmath9 .this is subject to the restriction that @xmath34 which gives @xmath35 that is @xmath36 , which can be satisfied by a real @xmath19 for example @xmath37 provided @xmath38 which is always satisfied since from ( [ bf ] ) the expression on the left is @xmath39 .notice there is no restriction that @xmath40 because the fluid can move along the lines of the force to counteract such a superluminal rotation leaving its total @xmath41 .this is true however large @xmath42 becomes !looking at equation ( [ bf ] ) there are several special cases that are worth noticing because they illustrate situations of particular simplicity .all the fields so far discussed are helical so it is useful here to define the pitch @xmath43 of a helical field line .the equation of such a field line is given by @xmath44 with @xmath27 constant .the pitch is defined as @xmath45 so @xmath46 .\(i ) if @xmath47 is independent of @xmath27 then @xmath48 so from ( [ bf ] ) @xmath49 .when @xmath50 , @xmath51 but as pointed out above there is no restriction that @xmath52 so we can have situations with @xmath53 at large @xmath27 .there is however a tendency for such solutions to be unstable . in all casesthe lines of force are given by @xmath54const and @xmath55 .the pitch of the resultant helix is @xmath56 so @xmath57 .( ii)writing equation ( [ bf ] ) in terms of the pitch @xmath43 we find @xmath58=<b_{z}^{2}>\ , , \label{bz}\end{aligned}\ ] ] a remarkable special case occurs when @xmath59 is independent of @xmath27 .this occurs when for example both the pitch and the angular velocity are both constant. then we may multiply equation ( [ bz ] ) by @xmath23 and differentiate the result with respect to @xmath23 ; this gives us @xmath26 on the right which cancels against a term on the left . dividing the resultant equation by @xmath23 yields @xmath60 + 2(\kappa^{2}-\omega^{2}/c^{2})b_{z}^{2}=0\,.\end{aligned}\ ] ]this integrates to give , on taking the square root @xmath61 notice that the field is uniform if @xmath62 .if now a rotating tube of force is surrounded by a static uniform field @xmath63 at equilibrium with it at @xmath64 @xmath65 where @xmath66 is the surface charge on the rotating cylinder and @xmath5 is the electric field component inside the boundary @xmath64 . by gauss s theorem @xmath67 and from ( [ electric ] )hence @xmath69 where @xmath70 .thus when @xmath71 the rotating uniform field has the same @xmath72 as the non - rotating field outside .more generally the flux along the cylinder of radius @xmath73 is @xmath74\ , .\end{array}\end{aligned}\ ] ] this function of @xmath75 maximises at @xmath76 when @xmath77 that is when @xmath78 . in all other cases the equilibrium twisted field configuration carries less flux over the circle @xmath79 than the straight bounding field @xmath80 .thus although some feel that the twisting of the field should cause it to contract under the influence of the hoop stresses so generated , nevertheless this is more than offset by the increase in field strength per unit longitudinal flux which is also caused by the twisting of the field .notice however we have only proved this when @xmath81 is independent of @xmath27 within the twisted region .however for @xmath82 between @xmath83 and @xmath84 the flux decrease is less than @xmath85 so there is only a @xmath86 increase in the radius of the twisted rotating cylinder in such circumstances .an untwisted @xmath9 field spinning about the axis of a cylinder whose top speed is @xmath87 of the velocity of light only increases its radius by about about @xmath88 !\(iii ) now suppose that the rotation is uniform out to @xmath89 but thereafter takes the keplerian profile @xmath90 .we substitute this velocity profile in equation ( [ bf ] ) @xmath91 now the alfvn waves generated by such a rotation will travel at the speed light in a force - free field ( see section 4 ) so the pitch generated is given by @xmath92 so @xmath93 so the first and last terms of equation ( [ bf2 ] ) cancel so we are left with @xmath48 so the z - component of the field is uniform despite its differential rotation i.e. we are back to ( i ) .since the solutions that leave the flux through the cylinder unchanged have a uniform @xmath47 we now look for time dependent solutions of the form @xmath94 $ ] so @xmath1 @xmath95+v_{b } \bm{\hat{b}}\,,\end{aligned}\ ] ] then , @xmath96 @xmath97 @xmath98 @xmath99 @xmath100 @xmath101 from the force - free condition ( [ force ] ) we have @xmath102+\nonumber \\\big(\frac{\dot{\omega}}{c^{2}}b_{0}r-\frac{\partial b_{\phi}}{\partial z}\big)(b_{\phi}\bm{\hat{z}}-b_{0}\bm{\hat{\phi}})=0\ , . \label{force3}\end{aligned}\ ] ] the @xmath103 and @xmath104 components of this equation are satisfied provided @xmath105 but by ( [ curle ] ) @xmath106 hence @xmath107 which wave equation is satisfied by @xmath108 $ ] .then @xmath109 and putting this back into ( [ force3 ] ) the @xmath110 component is satisfied identically , so we have an exact solution for any wave form @xmath111 .of course we could superpose any @xmath112 but to solve the problem of a perfectly conducting disc at @xmath113 forced to rotate at a given time dependent rate @xmath114 we need the wave that travels out from the origin .then the magnetic field at @xmath115 , @xmath79 rotates at @xmath111 and that at @xmath116 , @xmath79 rotates at @xmath112 , see fig . 1 .thus we have alfvn waves generated by a rotating disc . at @xmath64there is a surface density of charge @xmath117 ; @xmath118 , which accounts for the electric field outside @xmath64 .the changing surface density of charge involves a sheet current along the cylinder given by integrating ( [ force3 ] ) across the discontinuity in @xmath30 at @xmath64 thus @xmath119 and @xmath120 as required by charge conservation .note that the net charge of the rotating cylinder is zero , as the volume charge density given by ( [ charge ] ) and the surface charge density when integrated over the cylindrical volume and the discontinuity surface respectively add up to zero , so the electric field outside the cylinder is zero .@xcite in the context of blandford - znajek mechanism has investigated similar solutions .this is also the light cylinder.,scaledwidth=50.0% ]so far our solutions have been exact but abstract .we now take a step nearer to the real world but the application is somewhat forced and the solution inexact .consider a star with a dipolar magnetic field with axis along @xmath121 .let it be embedded in an interstellar magnetic field oriented in the same direction .then the field is as shown in fig . 2 where the separatrix has been emphasised .the closed field lines near the star will lie in meridional planes and rotate with the star. the outermost field lines outside any separatrix will not rotate at all just like the outer field lines in fig . 1 .the remaining field lines that join the star to the interstellar gas will carry the torque that generates the alfvn wave that propagates along the interstellar field to great distances from the star . in our force - free approximationthese alfvn waves travel at the velocity of light but in reality the fields will probably be too weak well away from the star so there the rotational disturbance will travel at the alfvn speed given by @xmath122 where @xmath123 is the gas density . in fig . 2 we have assumed that the light cylinder lies beyond the asymptotic radius of the separatrix . for a strong stellar dipole or a weaker interstellar fieldthis will not be true .we shall consider that situation afterwards .we write the field in the form @xmath124 then for a dipole @xmath125 in a straight external field we find @xmath126 the separatrix @xmath127 which for large @xmath7 gives @xmath128 . thus all the flux with @xmath129 links the star to the interstellar magnetic field . at large @xmath7we have from section 3 ( i ) and under equation ( 27 ) @xmath130 , so since near the star we need @xmath131 to carry the torque away from the star we set @xmath132 for @xmath133 , @xmath134 and @xmath135 for @xmath7 large since that gives the field found in section 4 .we therefore choose as our approximation @xmath136 whenever @xmath115 and @xmath134 . for @xmath137 , @xmath138 , see fig 3 .since we have used the @xmath139 of the non - rotating configuration which will in practice be modified by the rotating charges this is only approximate ; however we note that in our rotating cylinder the @xmath47 remained at @xmath63 even in the presence of the rotation so at large @xmath7 the proposed solution is exact .now we consider the case in which the light cylinder is of smaller radius than the asymptotic radius , @xmath73 , of the separatrix of the static field .close to the light cylinder the interstellar magnetic field is far too weak to have any effect so for the aligned pulsar , the work of @xcite as modified by @xcite and most recently treated @xcite is appropriate . however in those treatments the far field has a weak poloidal component and spirals off to infinity with that component not far from radial .there the field will be radically modified if the pulsar resides in an asymptotic field from the interstellar medium because asymptotically the field from the pulsar must align along the cylinders and produce alfvn waves propagating along them .@xmath140 will be much greater than @xmath141 .4 illustrates the type of the field configuration that we envisage .all the pulsar field lines that do not close within the light cylinder end up winding around the interstellar field and their @xmath142 is still that of the pulsar ! the flux function for the pure dipolar field is @xmath143 so the flux that fails to reach the equator within the light cylinder in such a field would be @xmath144 . in realitythe field is distorted by the rotation to be more like @xmath145 for @xmath146 which gives @xmath147 . a very rough model that gets the fluxes right ( but is not force - free ) is given by the @xmath139 above for @xmath148 , @xmath149 , for @xmath150 where @xmath151 and @xmath152 for @xmath153 . as before the toroidal fieldis given as in equation ( 28 ) with @xmath154 for @xmath155 and zero elsewhere .to see the orders of magnitude consider a pulsar whose polar field is @xmath156 and suppose the radius of the polar cap whose field does not close within the light cylinder is @xmath157 and the inter - stellar field is @xmath63 .then the asymptotic radius of the field attached to the pulsar will be @xmath158 suitable values might be @xmath159 , @xmath160 and @xmath161 giving @xmath162 .the value of @xmath157 is linked to the pulsars rotation rate @xmath142 and radius @xmath163 by @xmath164 which gives @xmath161 for @xmath165 for a pulsar with a period of @xmath166 .more generally @xmath167 where @xmath168 is the pulsar period . for the numbers above but general @xmath168 , we find @xmath169 which is many times the speed of light .if the alfvn wave advances at c then in the interstellar gas @xmath170 in the idealised model and this is increased by the factor @xmath171 if the alfvn speed is @xmath172 .however such highly wound magnetic fields are likely to be highly unstable . in the ideal casethe plasma has to move along the field so that @xmath173 .the field winds through the plasma like an archimidean screw winds up the water .while these estimates will be complicated by instabilities they do indicate that pulsars generate much enhanced fields in the interstellar medium albeit over tubes of force with diameters similar to that of the solar system .a different but interesting situation occurs when the interstellar field is aligned oppositely to magnetic moment of the pulsar .the total flux of the pulsar outside its light cylinder , @xmath174 , will be unable to counteract the flux of the interstellar field equatorially all the way to infinity since the latter grows like @xmath23 .furthermore the pulsar s flux can not cross its equatorial plane outside the light cylinder as by symmetry it would have to cross perpendicularly and the plasma attached would then move faster than light!however the flux from the pulsar has to reach infinity since it carries the flow of plasma pushed by the archimedian screw of the whirling field .the only place where the flux from the pulsar can deform the field over large distances is close to the axis , indeed in the non - rotating case there would be a neutral point on axis @xcite , see fig .5 . as the pulsar rotates the equatorial diameter of the field structure around it grows and this neutral pointis sent to infinity along the axis .an oppositely directed poloidal field is generated near the axis and is accompanied by the @xmath30 and the plasma flow due to the pulsar s rotation .we illustrate both the weak spin case when all the star s field lies within the light cylinder and rotates with the star fig . 5 and the pulsar case just described in fig .corresponding to any solution such as that shown in fig .there is an other solution with the field reversed outside the separatrix . if the longitudinal field inside the separatrix is small compared to the toroidal field there , the field pressures and tension still balance along the separatrix .the argument given above leads us to believe that this is the field configuration of the steady anti - aligned pulsar .in this paper we have studied analytical solutions of rotating magnetic fields emerging from dipoles or discs embedded in uniform magnetic fields .this study is based on the assumption that the dipole field is either aligned or counter - aligned to the external field .fields at an angle @xmath175 to the dipole may be considered by taking the solutions above with @xmath80 replaced by @xmath176 and superposing a lateral field parallel to the plane of the disk of magnitude @xmath177 , but this gives gives only orders of magnitude and is no substitute for a full non - axially symmetric treatment .we find that the spinning field will form alfvn waves in the interstellar gas .there is no need to confine these alfvn waves inside the light cylinder , as the plasma can have a velocity component parallel to the magnetic field and not exceed the speed of light and we find that for typical pulsar values they will occupy cylinders whose cross sectional areas are of the size of the solar system .these field configurations with tightly wound fields outside the light cylinder are probably very unstable , in particular the pressure of the whole tube along the axis is very large so any departure from axial symmetry will be enhanced by an instability analogous to the fire - hose instability .however there remains the interesting possibility that the magnetic fields generated over small regions by the thrashing field lines might be surprisingly large although smaller than the maxima estimated via these axially symmetric models .

<S> exact relativistic force free fields with cylindrical symmetry are explored . </S><S> such fields are generated in the interstellar gas via their connection to pulsar magnetospheres both inside and outside their light cylinders . </S><S> the possibility of much enhanced interstellar fields wound on cylinders of solar system dimensions is discussed but these are most likely unstable . </S><S> [ firstpage ] mhd methods : analytical stars : magnetic field stars : neutron stars : pulsars . </S>

hep-lat9709138

lattice nrqcd has been widely used in the past few years to calculate a variety of phenomenologically interesting quantities .the first step in the nrqcd program , naturally , is the calculation of the spectrum of the @xmath3 system . in the simulation presented herewe use sesam s large sample of dynamical wilson - fermion gauge configurations to study both the spin - independent as well as the spin - dependent spectrum of the @xmath4 .our strategy , in searching for sea - quark effects , will be to compare our final dynamical results to that of a quenched simulation at equivalent lattice spacing . with our three sea - quark masseswe can also study the dependence of mass - splittings on the light sea - quark mass .recent simulations have exposed the sensitivity of spin - independent splittings to the details of the action .therefore , in the spirit of @xcite , we have implemented the nrqcd action including spin - dependent corrections of @xmath1 and we tadpole improve using the mean link calculated in landau gauge . with these ingredients we hope to clarify the effect of unquenching in the spin - dependent splittings .in addition , using the quenched configurations we investigate : ( i ) the effect of changing tadpole prescriptions ; ( ii ) the effect of using an @xmath5 correct action compared to the @xmath1 corrected one .the non - relativistic ( euclidean ) lattice hamiltonian to @xmath6 consists of @xcite : the kinetic energy operator , @xmath7 which is of order @xmath8 ; relativistic corrections of order @xmath9 , @xmath10 and spin - sensitive corrections of order @xmath11 , @xmath12 the @xmath13 and @xmath14 fields are represented by the standard clover term or , where appropriate , by an improved version to remove @xmath15 discretization errors .following @xcite the quark greens functions are calculated from the evolution equation @xmath16 with @xmath17 @xcite .we rely on tapole improvement choosing the mean link in landau gauge as our improvement scheme .the coefficients @xmath18 are then set to their tree level value of 1 .the lattice parameters we have used are displayed in table [ simulation ] .details concerning the generation of our dynamical configurations and issues surrounding autocorrelations are discussed in ref .we exploit configurations more than once by starting the propagator evolution both at different spatial source points and on different timeslices .a binning procedure confirms that our 4 measurements per configuration are indeed independent .throughout , we fix the bare heavy quark mass to a value @xmath19 .cccc + @xmath20 & 0.156 & 0.1570 & 0.1575 + # configurations & 200 & 200 & 200 + measurements & 800 & 800 & 800 + + + +for details concerning the choice of operators , smearing techniques and fitting procedures we refer to @xcite . we extrapolate energy level splittings linearly in the quark mass : @xmath21 where @xmath22 denotes the bare light quark masses(see figure [ chiral1 ] for an example ) .the relevant scale we pick is given by @xmath23 .the value of the strange quark mass is taken from our recent light quark mass calculation @xcite and is such that @xmath24 lies close to our lightest sea quark mass ..we use the average lattice spacing from the @xmath25 and @xmath26 to convert our results to physical units .@xmath27 is to be compared to the experimental value of 1.28.[tab_spacings ] [ cols="^,^,^",options="header " , ] ( 130,140)(10,930 ) ( 15,935)(0,1)125 ( 13,950)(0,50)3(1,0)4 ( 14,950)(0,10)10(1,0)2 ( 12,950)(0,0)[r]9.5 ( 12,1000)(0,0)[r]10.0 ( 12,1050)(0,0)[r]10.5 ( 12,1070)(0,0)[r]gev ( 15,935)(1,0)150 ( 40,930)(0,0)[t]@xmath28 ( 25 , 942.9 ) ( 25 , 942.9)(0,1)0.1 ( 25 , 942.9)(0,-1)0.1 ( 25,1002.8 ) ( 25,1002.8)(0,1)2.1 ( 25,1002.8)(0,-1)1.6 ( 35,1001.8 ) ( 35,1001.8)(0,1)2.6 ( 35,1001.8)(0,-1)2.4 ( 35 , 942.7 ) ( 35 , 942.7)(0,1)0.2 ( 35 , 942.7)(0,-1)0.1 ( 80,930)(0,0)[t]@xmath29 ( 90,940)(0,0)1s ( 65,946)(3,0)9(1,0)2 ( 90,990)(0,0)2s ( 65,1002)(3,0)9(1,0)2 ( 90,1026)(0,0)3s ( 65,1036)(3,0)9(1,0)2 ( 70,946 ) ( 70,1005.8 ) ( 70,1005.8)(0,1)1.6 ( 70,1005.8)(0,-1)2.0 ( 70,1061.4 ) ( 70,1061.4)(0,1)7.5 ( 70,1061.4)(0,-1)7.5 ( 75,946 ) ( 75,1001.3 ) ( 75,1001.3)(0,1)2.5 ( 75,1001.3)(0,-1)2.4 ( 75,1042.3 ) ( 75,1042.3)(0,1)11.6 ( 75,1042.3)(0,-1)16.2 ( 120,930)(0,0)[t]@xmath30 ( 140,990)(0,0)1p ( 100,990)(3,0)9(1,0)2 ( 140,1026)(0,0)2p ( 100,1026)(3,0)9(1,0)2 ( 105 , 987.6 ) ( 105 , 987.6)(0,1)1.3 ( 105 , 987.6)(0,-1)0.9 ( 105,1039.6 ) ( 105,1039.6)(0,1)7.3 ( 105,1039.6)(0,-1)5.6 ( 115 , 991.0 ) ( 115 , 991.0)(0,1)2.2 ( 115 , 991.0)(0,-1)2.0 ( 115,1031.2 ) ( 115,1031.2)(0,1)9.1 ( 115,1031.2)(0,-1)7.1 figure [ fig : spectrum ] and [ fig : finestructure ] summarize our results for 2 and 0 flavours .unquenching effects are clearly visible in the spin - independent part of the @xmath0 spectrum .this is also evident from table [ tab_spacings ] : @xmath31 disagrees with the experimental value @xmath32 in the quenched case but coincides with it when dynamical fermions are included .( 100,80)(15,-50 ) ( 15,-50)(0,1)80 ( 13,-40)(0,20)4(1,0)4 ( 14,-40)(0,10)7(1,0)2 ( 12,-40)(0,0)[r]@xmath33 ( 12,-20)(0,0)[r]@xmath34 ( 12,0)(0,0)[r]@xmath35 ( 12,20)(0,0)[r]@xmath36 ( 12,40)(0,0)[r]mev ( 15,-50)(1,0)130 ( 28,0)(3,0)9(1,0)2 ( 60,2)(0,0)[t]@xmath4 ( 60,-34)(0,0)[t]@xmath37 ( 35,0 ) ( 35 , -31.3 ) ( 35 , -31.3)(0,1)1.2 ( 35 , -31.3)(0,-1)1.0 ( 45,0)(0,0 ) ( 45 , -33.4 ) ( 45 , -33.4)(0,1)1.9 ( 45 , -33.4)(0,-1)1.5 ( 68,0)(3,0)10(1,0)2 ( 100,0)(0,0)[l]@xmath38 ( 73 , -2.8 ) ( 73 , -2.8)(0,1)1.7 ( 73 , -2.8)(0,-1)2.1 ( 83 , 0.9 ) ( 83 , 0.9)(0,1)2.9 ( 83 , 0.9)(0,-1)2.8 ( 110,-40)(3,0)10(1,0)2 ( 140,-40)(0,0)[l]@xmath39 ( 110,-8)(3,0)10(1,0)2 ( 140,-8)(0,0)[l]@xmath40 ( 110,13)(3,0)10(1,0)2 ( 140,13)(0,0)[l]@xmath41 ( 115 , -27.1 ) ( 115 , -27.1)(0,1)4.1 ( 115 , -27.1)(0,-1)3.8 ( 115 , -6.7 ) ( 115 , -6.7)(0,1)2.3 ( 115 , -6.7)(0,-1)3.2 ( 115 , 9.4 ) ( 115 , 9.4)(0,1)1.7 ( 115 , 9.4)(0,-1)2.5 ( 125 , -28.3 ) ( 125 , -28.3)(0,1)6.9 ( 125 , -28.3)(0,-1)6.3 ( 125 , -4.0 ) ( 125 , -4.0)(0,1)3.7 ( 125 , -4.0)(0,-1)2.7 ( 125 , 8.0 ) ( 125 , 8.0)(0,1)3.1 ( 125 , 8.0)(0,-1)2.7 we do not observe any significant impact of unquenching on the hyperfine splittings . in particular , the p - hyperfine splittings seem to be underestimated for both @xmath42 and @xmath43 . clearly , this result needs to be corroborated with higher statistics .errors on our hyperfine splittings encompass the statistical error as well as the uncertainty in the fitting range ; the latter is essentially of the same size as the statistical error so that we can expect some improvement with higher statistics .we expect the inclusion of @xmath44 terms to effect the hyperfine splittings on the 10 % level , since @xmath45 . performinga relatively inexpensive @xmath46 quenched simulation at @xmath47 ( with 200 measurements ) we find this naive expectation to be well satisfied for the @xmath48 splitting as is shown in figure [ figsystematics ] .however , using the plaquette prescription for @xmath49 ( result from the nrqcd - collab ., also included in figure [ figsystematics ] ) shifts the value on the 10% level in the opposite direction as adding @xmath1 corrections .overall , nrqcd has proven to work very well in the spin - independent sector , in particular in giving a precise determination of the lattice spacing @xmath50 .it appears worthwhile to push the scale determinations to higher statistics using our configurations generated by t@xmath51l at @xmath52 on @xmath53 lattices ( @xmath2 ) .this would enable a reliable extrapolation to 3 or 4 flavours .the dominant source of error in such a lattice scale determination is most likely due to the remaining lattice discretisation errors in the gauge configurations .although higher statistics are highly desirable , it seems unlikely that progress in the spin - dependent sector will come from this approach alone .it is worrying to find the spin - dependent corrections of @xmath54 as large as 10 % even for the @xmath3 , decreasing the fine - splittings relative to the lower order estimates .a perturbative or non - perturbative calculation of the coefficients @xmath18 is badly needed .

<S> we present results for the @xmath0 spectrum obtained using an @xmath1-correct non - relativistic lattice qcd action . </S><S> propagators are evaluated on sesam s three sets of dynamical gauge configurations generated with two flavours of wilson fermions at @xmath2 . </S><S> compared to a quenched simulation at equivalent lattice spacing we find better agreement of our dynamical data with experimental results in the spin - independent sector but observe no unquenching effects in hyperfine - splittings . to pin down the systematic errors we have also compared quenched results in different `` tadpole '' schemes and used a lower order action . </S>

quant-ph0401158

molecules interacting with electromagnetic fields foster a variety of dynamical phenomena .specific wave packets of the internuclear vibration have been excited by shaped fs laser pulses @xcite .such vibrational quantum states have then been reconstructed by molecular emission tomography @xcite. furthermore , control of the internal molecular quantum state could be achieved @xcite , for eventually enhancing chemical reactions . besides these applications of pulsed fields , also interactions with cw laser fields or static fields reveal interesting effects .for instance , the axis of paramagnetic molecules can be aligned in pendular states by applying magnetic fields @xcite .furthermore , it has been shown how electric fields align polar molecules @xcite .non - polar molecules , on the other hand , seem to be rather isolated from the influence of electric fields .they are infrared inactive and thus their rotation is essentially undamped and isolated from the electromagnetic vacuum .clearly this does not hold for the internuclear vibration and , in addition , the rotational statistics determines via ro - vibrational coupling its relaxation and decoherence properties @xcite .nevertheless , a permanent dipole moment being absent , non - polar molecules still can be polarised by electrical fields and thus feel an effective alignment force when an anisotropic polarisability prevails .the quantum - mechanical eigenstates and energies have been discussed in ref .@xcite , where it has been shown that the dependence of the eigenenergies on the field intensity can be used for providing a dipole trapping force .recent experiments have implemented such a dipole trap for molecules @xcite . in general the topic of producing @xcite , cooling @xcite , and trapping of ultra - cold molecules@xcite became a major focus of research in recent years .given the ultra - low temperatures of a condensed atomic or possibly molecular gas , the molecular rotation , representing the lowest energy scale apart from centre - of - mass motion , then should play an important role . for sufficiently cold molecules one might expect decoherence first to appear on the rotational energy scale before affecting the energetically higher degrees of freedom , such as the internuclear vibration .moreover , the molecular rotation provides a toolbox for implementing various textbook examples of mechanics , such as spherical pendula and the motion of rigid bodies .in particular , we will focus here on homonuclear , diatomic molecules in a far - detuned linear - polarised laser field .this is a setup relevant also for trapping of photo - associated ultra - cold molecules in a dipole trap @xcite . in sec .[ sec:2 ] a brief review is given on the classical description of the motion of the molecular axis .a link to a quantum description in terms of stimulated two - photon raman processes is then presented in sec .[ sec:3 ] .moreover , the theory is extended there by including spontaneous raman processes in form of a quantum master equation .the effects due to these spontaneous processes on the molecular rotation are evaluated in sec .[ sec:4 ] where we provide a phase - space picture by using a wigner function for the molecular rotation .consider a diatomic molecule in an electric field of amplitude @xmath0 pointing towards the @xmath1 direction .the potential energy in such a field is given by @xmath2 where @xmath3 is the permanent dipole moment of the molecule and @xmath4 denotes the anisotropy of the polarisability with respect to fields parallel and perpendicular to the molecular axis .the direction of this axis with respect to the electric field is specified by the enclosing angle @xmath5 , cf .[ fig : geometry ] .the molecular axis specifies the @xmath6-axis of a body - fixed coordinate system , whose orientation with respect to the laboratory frame is specified by euler angles . from the lagrange equations one obtains the equations of motion for these as @xmath7 where @xmath8 is the moment of inertia perpendicular to the molecular axis .moreover , the conservation of energy @xmath9 reads @xmath10 .\ ] ] note that for our case the relevant euler angles @xmath5 and @xmath11 coincide with the spherical angles as depicted in fig .[ fig : geometry ] . the angular - momentum @xmath12 can be expressed in the laboratory frame by the angles and their time derivatives as @xmath13 from eqs ( [ eq : lz - const ] ) and ( [ eq : l - vec ] ) it is seen that the component of the angular momentum in direction of the applied field is a constant of motion : @xmath14 .thus the tip of the angular - momentum moves in a plane normal to the electric - field direction . using the constant of motion @xmath15 , eq .( [ eq : e - const ] ) can be rewritten as @xmath16 where the new potential reveals a centrifugal - type barrier @xmath17 for later purposes we have introduced here the potential depths @xmath18 , @xmath19 and the @xmath1-part of the kinetic energy @xmath20 . a molecule with permanent dipole moment , @xmath21 , omitting its anisotropic polarisability , @xmath22 , reveals a potential of a spherical pendulum , as shown in fig . [fig : d - potential ] .the motion in @xmath11 with constant angular momentum @xmath15 is superimposed by a nutation in @xmath5 , which is equivalently described by a heavy symmetrical top with vanishing moment of inertia along the ( molecular ) body axis .the harmonically approximated nutation frequency is in this case @xmath23 .however , for the case of non - polar ( @xmath24 ) but polarisable molecules ( @xmath25 ) , which is the focus of this paper , the anisotropic polarisability produces a @xmath26 potential , cf .[ fig : alpha - potential ] , which for @xmath27 can not be cast into the form of a spherical pendulum .nevertheless , a spherical - pendulum - type motion can be expected , since also for this case the solution @xmath28 can be obtained in terms of elliptic integrals . expressed in terms of jacobian elliptic functions @xcite it reads , @xmath29 where @xmath30 is the initial angle at time @xmath31 .the constants appearing in eq .( [ eq : theta - solution ] ) are given as @xmath32 where the scaled energies @xmath33 and @xmath34 have been used . in the asymptotic limit of large energies@xmath35 , @xmath36 and @xmath37 so that @xmath38 results , which corresponds to the free rotator case with angular velocity @xmath39 . in this limit all three vector components of the angular momentum are conserved quantities . for lower energiesthe potential barriers lead to modifications , supporting several types of motion . for @xmath40 unbound motionoccurs for @xmath41 where the amplitude of the oscillation of @xmath5 attains the full @xmath42 range , see solid curve in fig .[ fig : kappaz=0 ] . given an initial angular momentum only in @xmath43 direction ,the molecular axis then continuously rotates around the @xmath43 axis so that while @xmath44 is modulated due to the potential barriers , it does not change its sign , see solid curve in fig .[ fig : ly - kappaz=0 ] . on the other hand , for the same case of @xmath40 , for energies @xmath45 the angle @xmath5 is subject to a reflection at the light - induced potential barrier .thus the range of @xmath5 values is restricted , as seen for the dashed line in fig .[ fig : kappaz=0 ] .in addition , since @xmath46 changes its sign at the reflecting barrier , the angular momentum @xmath44 correspondingly crosses zero and thus rotations with positive and negative orientations subsequently interchange , cf .dashed line in fig .[ fig : ly - kappaz=0 ] .note that for @xmath47 the motion of @xmath5 is additionally bound by reflections at the centrifugal barriers near @xmath48 and @xmath42 .for implementing the dynamics described so far , not only static electric fields can be used . also linear polarised cw optical fields , that are far detuned from electronic resonances ,have been shown to produce a similar interaction @xcite . in this way one can take advantage of large polarisabilities of non - polar molecules to attain deep potential wells for the nutation angle .the specific excitation scheme for a dimer , depicting only the two lowest electronic potential surfaces that provide a @xmath49 dipole transition , is shown in fig .[ fig : level - scheme ] . in order to avoid resonant vibronic transitions and to be left with the sought two - photon processes , a sufficiently large detuning @xmath50 from the bare electronic resonance frequency is required .given an optical field of spectral width much smaller than vibrational frequencies @xmath51 , stimulated raman transitions occur that only affect the rotational degree of freedom .however , the relative rate of spontaneous vs stimulated raman transitions , @xmath52 , with @xmath53 being the bare electronic linewidth of the dipole transition , shows that even for large detuning , @xmath54 , a certain number of spontaneous processes occur .these processes not only lead to an incoherent effect on the molecular rotation , such as relaxation and quantum decoherence , but also excite the internuclear vibration via franck condon transitions .nevertheless , as long as the laser remains detuned from resonant vibronic transitions , the electronic and vibrational degrees of freedom can be eliminated to obtain equations of motion for the molecular rotation alone .this is achieved by consistently eliminating the electronic coherences and electronic excited - state populations , keeping only terms up to second order in the molecule - light interaction . in franck condon approximation the vibrational degree of freedom can then be traced over @xcite .the requirement of avoiding resonant vibronic transitions leads to a maximum interaction time on which the molecule stays vibrationally sufficiently cold , which can be estimated as @xmath55 here @xmath56 is the raman rabi frequency with @xmath57 being the electric - dipole matrix element between electronic ground and excited states .the linear polarised electric field is specified by @xmath58 with @xmath59 being an optical frequency far - detuned from vibronic resonance and @xmath60 being the unit vector in @xmath1 direction .using these parameters the light - induced potential depth results as @xmath61 .the parameter @xmath62 is the ratio of the difference of internuclear equilibrium distances in ground and excited electronic states to the spatial extent of the vibrational ground - state wave packet . since @xmath63 for alkali dimers , in the weak - field regime [ @xmath64 this time corresponds to a very large number of free rotational periods . under these conditionsthe equations of motion for the orientation of the molecular axis in the far detuned optical field can be cast into a master equation of lindblad form @xcite : @xmath65 + \sum_{i = x , y , z } \left ( \hat{s}_i \ , \hat{\sigma } \ , \hat{s}^\dagger_i - { \textstyle\frac{1}{2 } } \ , \hat{s}^\dagger_i \hat{s}_i \ , \hat{\sigma } - { \textstyle\frac{1}{2 } } \ , \hat{\sigma } \ , \hat{s}^\dagger_i \hat{s}_i \right ) .\end{aligned}\ ] ] here @xmath66 is the reduced rotational density operator of the molecule and the free kinetic energy is given by @xcite @xmath67 the stimulated rotational raman transitions then provide the sought potential in the form @xmath68 where in terms of spherical angles the unit - vector operator @xmath60 reads @xmath69 , so that @xmath70 recovers the potential as discussed for the classical treatment .the spontaneous rotational raman transitions are given in terms of the operators @xmath71 , with @xmath72 denoting the polarisation of the spontaneously emitted photon .they read @xmath73 with @xmath74 ( @xmath72 ) being unit vectors pointing into the corresponding directions . in the representation of spherical anglesthese jump operators can be cast in vector form to : @xmath75 \langle \vartheta , \varphi| .\ ] ] for vanishing bare electronic linewidth , @xmath76 , a unitary evolution is recovered from the master equation ( [ eq : master ] ) that can be reformulated in terms of the schrdinger equation @xmath77 with the hamiltonian given by eqs ( [ eq : ham - mol ] ) and ( [ eq : raman - ham ] ) . in the @xmath78 representationthe evolution of the wavefunction @xmath79 reads @xmath80 + u_\alpha \cos^2\vartheta \right\ } \, \psi(\vartheta , \varphi ) .\ ] ] using the ansatz @xmath81 , with @xmath82 being a constant of motion , and @xmath83 , the energy eigenstates are obtained from the angular oblate spheroidal wave equation @xcite @xmath84- \frac{m^2}{1 \!-\ ! z^2 } \right ] + u_\alpha z^2 + e_{lm } \right\ } s_{lm}(z ) = 0 .\ ] ] the corresponding eigenenergies @xmath85 can be expanded in powers of the interaction potential @xmath86 as @xmath87 + \ldots \ , .\ ] ] they manifest a partial removal of the degeneracy with respect to @xmath88 , as shown in fig .[ fig : e - levels ] .the properties of these eigenstates and their angular confinement have been discussed in detail in ref .@xcite .the quantum dynamics of a system prepared quite analogously as discussed for the classical treatment , i.e. with an initial average angular momentum pointing only in @xmath43 direction , is shown in fig .[ fig : lx - dynamics ] . herean initial coherent angular - momentum state @xcite of the form @xmath89^{j+m } \big[\cos ( \alpha/2 ) \big]^{j - m } \ , e^{-i(j + m)\beta } |jm\rangle , \ ] ] for @xmath90 , @xmath91 , and @xmath92 has been used .classically that would correspond to the case @xmath40 where reflections in the motion of @xmath5 can only occur due to the light - induced potential . but from eq .( [ eq : coh - state ] ) it is clear that we deal with an initial state containing states @xmath93 with all possible values for @xmath88 : @xmath94 thus part of the wave packet in @xmath5 will also be reflected by centrifugal barriers . in fig .[ fig : lx - dynamics ] an oscillation of @xmath95 between positive and negative values is observed that is characteristic for reflections of the wavefunction at potential barriers .however , the transient behaviour at the crossing of zero indicates that part of the wave packet in @xmath5 is transmitted through the barrier , and that reflected and transmitted parts interfere for the short time where they spatially overlap . when spontaneous processes are included in the dynamics , i.e for @xmath96 , the full master equation ( [ eq : master ] ) must be solved . due to the scaling of complexity of this numerical problem with the number of required rotational levels , it is advantageous to employ a quantum trajectory method. then individual realisations of state vectors can be calculated and finally an ensemble average over the set of realisation reproduces the sought density matrix .individual trajectories are comprised of non - unitary evolutions with the effective hamiltonian @xmath97 describing the conditioned evolution when no spontaneous processes occur , intermitted by spontaneous raman processes described by the application of the jump operators @xmath71 . as can be seen from their definition ( [ eq : def - s2 ] ) , these operators may excite the motion in the angle @xmath11 , by producing via @xmath98 and @xmath99 a weighting of the corresponding probability amplitude @xmath100 in @xmath11 . thus motion in @xmath11 andconsequently the angular momentum component @xmath101 are getting excited by spontaneous raman processes .this can be observed in fig .[ fig : lz2 ] , where the variance of @xmath101 is plotted over time . for @xmath102 a monotonous increase of the varianceis observed , indicating that the kinetic energy is subject to heating .of course the spontaneous raman scattering will also lead to a suppression of coherent processes and the oscillations of the average angular momentum will be damped , cf .solid curve in fig .[ fig : lx - dynamics ] .in the same way as the angular - momentum coherent states are a generalisation of the familiar harmonic - oscillator coherent states , also the concept of phase - space quasi - probability distributions can be defined on the angular - momentum phase space @xcite . herewe employ a wigner function , which in the @xmath103-dimensional hilbert space of angular momentum @xmath104 , is defined in terms of the density operator @xmath66 and the spherical harmonics , @xmath105 , as defined in refs @xcite @xmath106 the multipole operators are defined by @xmath107 with @xmath108 being the wigner 3j symbol . the factor in front of eq .( [ eq : w - fn ] ) ensures normalisation to unity .as the laser interaction will result in a change of the rotational quantum number @xmath104 of the molecule we consider a sum of wigner functions over all possible values of @xmath104 : @xmath109 albeit the sum over @xmath104 , this phase - space distribution is not a complete description of the rotational quantum state .however , it proves useful for illustrating the laser - induced dynamics . for alternative approaches defining phase - space distributions independent of the particular value of @xmath104 ,see refs @xcite .moreover , it should be noted that even though the angular - momentum coherent states are analogous to the harmonic - oscillator ones in many respects , the wigner function of angular - momentum coherent states is weakly negative .this negativity decreases with increasing value for @xmath104 and reaches zero only in the ( classical ) limit @xmath110 .figure [ fig : w - fn_coh - state ] shows the time evolution of the system starting initially from the state @xmath111 , cf .( [ eq : coh - state2 ] ) .the phase - space distribution is shown for a scaled time @xmath112 after the first full period of oscillation of @xmath95 ( upper left plot ) , and then at later times in steps of quarter periods .one can see that the system almost returns to its initial coherent state after one period , when the spontaneous emission is neglected ( left column ) .the peak in phase space has then returned to its initial position at @xmath113 and @xmath114 . in the course of time the peak splits and after a half period , at @xmath115 , two peaks appear at @xmath113 , @xmath116 .that corresponds to an average angular momentum pointing now in negative @xmath43 direction . note that between half and full cycles , at @xmath117 and @xmath118 , the phase - space distribution develops several smaller peaks and also negative regions of the wigner function appear . including the spontaneous raman processes , the structures in phase - space generally are smoothed and negativities are suppressedthis can be seen from the right column of plots in fig .[ fig : w - fn_coh - state ] . as another example , in fig .[ fig : w - fn_sup - state ] a superposition of two coherent angular - momentum states , @xmath119 has been taken as the initial quantum state .this state has a strong negative region of the wigner function centred at @xmath120 , @xmath121 .the plots of the wigner function for this initial state show the same oscillatory behaviour as fig .[ fig : w - fn_coh - state ] , with the negative peak being shifted to @xmath122 at half periods . however , comparing the time evolution of these two cases one can see that the difference between the unitary and non - unitary evolution is much more pronounced in the case of the coherent state . especially at the times betweencompleted half and full cycles ( at @xmath123 and @xmath118 ) , when also the wigner function of the time - evolved coherent state shows relative strong negative regions , one observes that these negativities have become weaker in the presence of spontaneous emission .the populations in the @xmath104-level manifolds for the final interaction time of figs [ fig : w - fn_coh - state ] and [ fig : w - fn_sup - state ] are shown in fig .[ fig : jpops ] .it can be seen that for both the coherent state and the superposition of coherent states the populations in @xmath124 have increased by a few percent at the expense of the population in the initial manifold @xmath90 .that the coherent state in this example actually is much more sensitive to decoherence may seem counter - intuitive at first sight , but one should keep in mind that the angular - momentum coherent state is not classical in the usual sense of a harmonic - oscillator coherent state .that is , as has been seen above its wigner function in general is not positive everywhere .furthermore , angular - momentum coherent states apparently are not in general the most robust states against decoherence , as are harmonic oscillator coherent states for linearly coupled reservoirs .the substantial difference in how these two cases are affected by decoherence due to spontaneous processes can also be observed in fig .[ fig : purity ] , where the time evolution of the purity of the density operator is shown .it can be seen that the superposition state ( [ eq : sup - state ] ) develops much slower into a mixed state compared to the coherent state . .[ fig : purity ] the type of decoherence mechanism is in general determined by the form of the operators @xmath71 , see eqs ( [ eq : def - s ] ) and ( [ eq : def - s2 ] ) . in our casethese operators differ from raising and lowering operators @xmath125 , and thus our decoherence mechanism itself is different from that discussed in refs @xcite , though for special choices of initial conditions particular features may be similar .we have considered a diatomic homonuclear molecule in a far - detuned , linear - polarised light field . first the classical motion of the molecule was studied . for the case of a molecule with a permanent dipole ,the potential was seen to be that of a spherical pendulum . for non - polar molecules , however , the potential is of a different form , but the solution for the nutation angle still could be found in terms of elliptic integrals .the motion was then treated quantum mechanically , where stimulated and spontaneous raman processes were taken into account by a master equation . solving this equation both for an angular - momentum coherent state and a superposition state , the effects of the decoherence induced by the spontaneous processes where compared .it was seen that , in this particular example , the superposition state was more robust to decoherence than the coherent state .this research was supported by deutsche forschungsgemeinschaft .99 mckenzie c , hecker denschlag j , hffner h , browaeys a , de araujo l e e , fatemi f k , jones k m , simsarian j e , cho d , simoni a , tiesinga e , julienne p s , helmerson k , lett p d , rolston s l and phillips w d 2002 _ phys . rev . lett . _ * 88 * 120403 abramowitz m and stegun i a 1972 _ handbook of mathematical functions _ ( dover publications , 9th ed ) : for jacobian elliptic functions see chap .16 , p. 567, for oblate spheroidal wave equations see eq . 21.6.4 , p. 753 .

<S> the dynamics of non - polar diatomic molecules interacting with a far - detuned narrow - band laser field , that only may drive rotational transitions , is studied . </S><S> the rotation of the molecule is considered both classically and quantum mechanically , providing links to features known from the heavy symmetric top . in particular , </S><S> quantum decoherence in the molecular rotation , being induced by spontaneous raman processes , is addressed . </S><S> it is shown how this decoherence modifies the rotational dynamics in phase space . </S>

1511.03883

the slice genus of a positive braid knot equals the ordinary genus by a theorem of kronheimer und mrowka @xcite .however , the story is very different for the _ topological 4genus _ @xmath0 of positive braid knots @xmath1 , i.e. the minimal genus among surfaces which are locally flatly embedded in the 4ball and have @xmath1 as boundary .a first example is due to rudolph @xcite : for the torus knot @xmath2 , the inequality @xmath3 holds .more recently , a large proportional difference @xmath4 with respect to @xmath5 was found for all torus knots with non maximal signature @xmath6 in @xcite . on the other hand, there exists a lower bound , due to kauffman and taylor , for the topological 4genus of knots : @xmath7 holds for any knot @xmath1 @xcite .we show that for positive braid knots , this bound is in fact the only obstruction to non maximal topological 4genus , i.e. @xmath3 .[ equivalenz ] for a positive braid knot @xmath1 , the equality @xmath8 holds exactly if @xmath9 . combining this result with baader s classification of prime positive braid links of maximal signature @xcite, we immediately get a full description of all prime positive braid knots of maximal topological 4genus : they are exactly the torus knots of maximal signature . [ maximal ] the torus knots @xmath10 , @xmath11 and @xmath12 are the only prime positive braid knots @xmath1 with @xmath8 .our proof of theorem [ equivalenz ] uses two main ingredients .the first one is a homological criterion from @xcite using freedman s disc theorem @xcite , allowing us to conclude @xmath3 for certain positive braids .the second one is that _ genus defect _@xmath13 is inherited from _ surface minors _ , i.e. incompressible subsurfaces .similar to baader s four surface minors @xmath14 , @xmath15 , @xmath16 and @xmath17 obstructing maximal signature for positive braid links , we use enriched versions @xmath18 , @xmath19 , @xmath20 and @xmath21 to obstruct maximal topological 4genus for positive braid knots .theorem [ equivalenz ] also allows us to compute the topological 4genus for positive braid knots @xmath1 with @xmath22 . combining the lower bound of kauffman and taylor with @xmath23 yields the exact result @xmath24 .this suffices to compute the topological 4genus for prime positive braids knots with up to 12 crossings .table 1 lists all these knots , except for the torus knots @xmath10 , @xmath11 and @xmath12 , which have maximal topological 4genus. .small positive braid knots . [ cols="^,^,^,^,^",options="header " , ] this list is created with the help of the software knotinfo @xcite .previously , the values of the topological 4genus for all these examples except @xmath25 were marked as unknown .* acknowledgements . *i warmly thank sebastian baader , peter feller and lukas lewark for many inspiring discussions and ideas that found their way into this article .a _ positive braid knot _ is a knot that can be obtained from a positive braid via the closure operation , an important example being torus knots .a _ positive braid on @xmath26 strands _ is a finite word in positive powers of the braid generators @xmath27 . by a theorem of stallings ,positive braid knots are fibred with the standard seifert surface as fibre @xcite .as baader did in @xcite , we use _ brick diagrams _ to visualise the fibre surface of positive braid knots : each horizontal bar corresponds to a braid generator @xmath28 and each _ brick _ , i.e. each rectangle , corresponds to a positive hopf band in the plumbing construction of the fibre surface .if two bricks link , it means that the core curves of the corresponding positive hopf bands intersect once , see figure [ linking ] .let the _ linking pattern _ be the the plane graph obtained by putting a vertex into every brick and an edge between two vertices exactly if the corresponding bricks link .it can be easily seen that if the intersection pattern of a positive braid @xmath29 is not connected , then the positive braid link @xmath30 is not prime .in fact , the converse is also true since positive braids are visually prime by a theorem of cromwell @xcite .let us for a moment consider the case where the linking pattern is a tree .there are many brick diagrams that yield the same tree as linking pattern .since closures of positive braids corresponding to different brick diagrams might be equivalent as links in @xmath31 , it is natural to ask whether the plane tree of the linking pattern uniquely determines the positive braid link up to ambient isotopy .as we will see in the following remark , this is indeed the case .[ monodromiedeterminiert ] _ the fibre surface @xmath32 of a positive braid @xmath29 retracts to its brick diagram .since for a successive plumbing of positive hopf bands , the monodromy is conjugate to the product ( in the succession of plumbing ) of positive dehn twists along the core curves of the hopf bands @xcite , the conjugacy class of the monodromy ( and therefore also the corresponding fibred link @xmath30 ) is completely determined by the plane tree given by the linking pattern of the brick diagram . _ furthermore , if the linking pattern is a tree , a matrix for the seifert form of the corresponding fibre surface @xmath33 is particularly easy to describe : as a basis of @xmath34 take the core curves @xmath35 $ ] of the positive hopf bands corresponding to the bricks .a matrix @xmath36 for the seifert form is then given by @xmath37 and @xmath38 if @xmath39 and the curves @xmath40 and @xmath41 intersect ( i.e. if the corresponding vertices of the linking pattern are connected by an edge ) .all other entries are equal to zero . in order to detect genus defect for a positive braid knot @xmath30 , we search for minors @xmath42 or @xmath43 in the fibre surfacethis is always based on the fact that the linking pattern of @xmath29 contains the tree corresponding to @xmath42 or @xmath43 via deleting vertices and contracting edges .one can then see that also @xmath32 contains @xmath42 or @xmath43 , respectively , as a surface minor , implying @xmath44 .for example , figure [ h - tilde ] shows how the tree corresponding to @xmath20 is contained in the linking pattern of the positive braid @xmath45 .before we prove it for the case of positive braid knots , we show an analogue of theorem [ equivalenz ] for knots obtained as _ plumbing of positive hopf bands along a tree @xmath46_. this notion generalises knots corresponding to brick diagrams having some plane tree as linking pattern . starting from any finite plane tree @xmath46 , we plumb positive hopf bands ( which are in one - to - one correspondence with the vertices of the tree ) such that their core curves intersect once exactly if the corresponding vertices of @xmath46 are connected by an edge .otherwise , they do not intersect .furthermore , they respect the circular ordering of the vertices given by the plane tree structure of @xmath46 . by the argument given in remark [ monodromiedeterminiert ] ,there is , up to ambient isotopy , only one way to do this .this construction is strictly more general than positive braid knots with a plane tree as linking pattern : vertices of a tree can have arbitrary valency , while for linking patterns associated with positive braid knots , this valency is bounded from above by @xmath47 .[ trees_prop ] for a knot @xmath1 obtained by plumbing positive hopf bands along a plane tree @xmath46 , the equality @xmath8 holds exactly if @xmath9 .if @xmath9 , then @xmath8 follows from the signature bound of kauffman and taylor @xcite .if @xmath48 , we distinguish three different cases . if @xmath46 has at least three vertices of degree at least three , then the corresponding fibre surface contains @xmath49 as a minor and thus @xmath23 . if @xmath46 has two vertices of degree at least three , then at least one branch emanating from such a vertex is of length at least two , since otherwise @xmath1 can not be a knot .again the corresponding fibre surface contains @xmath49 as a minor .if @xmath46 has only one vertex of degree at least three , then @xmath48 holds if and only if @xmath46 contains the linking pattern of @xmath50 or @xmath17 as an induced subgraph .again , for @xmath1 to be a knot , at least one of the branches of the linking pattern of @xmath50 or @xmath17 has to be extended and @xmath46 in fact contains the linking pattern of @xmath42 or @xmath43 as an induced subgraph .hence , the corresponding fibre surface contains @xmath42 or @xmath43 as a surface minor .the proof of theorem 1 for positive braid knots @xmath1 is divided into two parts , depending on the _ positive braid index _ of @xmath1 , i.e. the minimal index of a positive braid @xmath29 representing the knot @xmath1 . for @xmath1 of positive index at most three , we can essentially reduce the problem to proposition [ trees_prop ] . for @xmath1 of positive index at least four ,we show that the strict inequality @xmath23 always holds .[ index3_prop ] for a knot @xmath1 obtained as the closure of a positive @xmath51braid @xmath29 , the equality @xmath8 holds exactly if @xmath9 .we assume to have applied all possible braid relations @xmath52 to the braid @xmath29 , so , up to cyclic permutation , @xmath29 can be assumed to be of the form @xmath53 , where @xmath54 and @xmath55 . if @xmath56 , the linking pattern of the braid is a plane tree and we are done by proposition [ trees_prop ] .we now show that in the other cases we already have @xmath44 . for this ,let @xmath57 and remark that at least one of the @xmath58 has to be odd and hence at least three , otherwise the permutation given by the braid leaves the third strand invariant and @xmath30 is not a knot ._ case 1 : @xmath59_. up to cyclic permutation , the braid @xmath29 contains the word @xmath45 as a _ subword _, i.e. via reducing powers of occurrences of generators , and thus the fibre surface of @xmath30 contains @xmath20 as a minor , implying @xmath44 ._ case 2 : @xmath60 , @xmath61_. if , up to cyclic permutation , @xmath62 equals @xmath63 , the seifert form of @xmath30 is positive definite .if @xmath62 equals @xmath64 or @xmath65 , the second strand is left invariant by the permutation given by the braid , so we can assume that one of the @xmath58 is at least four .furthermore , since one of the @xmath58 has to be odd , @xmath62 can be assumed to be at least @xmath66 or @xmath67 with respect to the product order . in both cases , @xmath29 contains the word @xmath68 as a subword and thus the fibre surface of @xmath30 contains @xmath18 as a minor , implying @xmath44 ._ case 3 : @xmath60 , at least one @xmath69_. as before , one of the @xmath58 has to be at least three , say @xmath70. if @xmath71 or @xmath72 is at least two , then @xmath29 contains , up to cyclic permutation , @xmath45 as a subword and thus the fibre surface of @xmath30 contains @xmath20 as a minor .now assume @xmath73 and @xmath74 .we also assume @xmath75 , otherwise we are , up to cyclic permutation , in the case we already dealt with .note that the permutation given by a braid of the form @xmath76 leaves the second strand invariant , so @xmath77 needs to be at least three in order for @xmath30 to be a knot .now , up to cyclic permutation , @xmath29 must contain the word @xmath78 and the fibre surface of @xmath30 contains @xmath18 as a minor , implying @xmath44 .[ minimalitylemma ] let @xmath29 be a positive braid of index @xmath79 . if for some @xmath80 the linking pattern of the subword of @xmath29 induced by the generators @xmath28 and @xmath81 is a path , then @xmath29 is not of minimal positive index. we can assume the subword of @xmath29 induced by the generators @xmath28 and @xmath81 to be @xmath82 , for some positive numbers @xmath83 and @xmath84 .this can be achieved by cyclic permutation and possibly reversing the order of the word @xmath29 , operations that do not change the associated fibre surface .similarly , we can assume that all occurrences of generators with index smaller than @xmath80 come before the last occurrence of @xmath28 and , likewise , all occurrences of generators with index greater than @xmath85 come after the first occurrence of @xmath81 .the situation is schematically depicted in figure [ nichtminimal ] on the left .now consider the brick diagram obtained by merging the two columns corresponding to the generators @xmath28 and @xmath81 as indicated in figure [ nichtminimal ] on the right . by definition , the corresponding positive braid @xmath86 has fewer strands than @xmath29 . to show that the closures of @xmath29 and @xmath86 are ambient isotopic in @xmath31, we study the corresponding monodromy homeomorphism of their fibre surfaces @xmath32 and @xmath87 . since the linking pattern of the two brick diagrams are equal , the corresponding monodromies are conjugate by the argument used in remark [ monodromiedeterminiert ] .therefore also the corresponding mapping tori are homeomorphic and the closures of @xmath29 and @xmath86 are ambient isotopic .[ index4_prop ] if @xmath1 is a prime knot obtained as the closure of a positive braid @xmath29 of minimal positive index @xmath88 , then @xmath23 .let @xmath29 be a positive braid of minimal braid index @xmath89 whose closure @xmath30 is a prime knot .we assume to have applied all possible braid relations @xmath90 to @xmath29 .in other words , the crossings of @xmath29 are as far to the right as possible .we can furthermore assume that @xmath29 still contains , up to cyclic permutation , the subword @xmath91 , since otherwise @xmath29 would not be of minimal index .we first delete , without disconnecting the linking pattern , a minimal amount of occurrences of @xmath92 so that the induced subword of @xmath29 in the first two generators is , after a possible cyclic permutation , of the form @xmath93 , where @xmath70 and @xmath94 are greater or equal to two .for example , if the induced subword of @xmath29 in the first two generators is @xmath95 , we delete one occurrence of @xmath92 ( to the power two ) , yielding , after a possible cyclic permutation , @xmath96 .note that in case @xmath73 , no generators @xmath92 have to be deleted to achieve the desired form ._ case 1 : @xmath97 @xmath98_. in this case , we did not have to delete any occurrence of @xmath92 and the induced subword of @xmath29 in the first two generators is exactly @xmath91 .both occurrences of @xmath92 have to be split by an occurrence of @xmath99 , since otherwise the permutation given by @xmath29 would leave the first or second strand invariant , see figure [ invariantstrand ] . furthermore , these occurrences have to be to the power at least two , since we ruled out the possibility of a braid relation @xmath100 .if one of the occurrences of @xmath99 is to some power at least three , @xmath29 contains , up to cyclic permutation , the subword @xmath101 and thus the fibre surface of @xmath30 contains the minor @xmath49 , implying @xmath102 .if the power of both occurrences of @xmath99 is equal to two , we repeat the same argument . eventually , some splitting occurrence has to be of power at least three and @xmath29 contains , up to cyclic permutation , the subword @xmath103 ._ case 2 : @xmath97 @xmath104 @xmath105_. in this case , we did not have to delete any occurrence of @xmath92 and the induced subword of @xmath29 in the first two generators is exactly @xmath106 . as in case 1 , the second occurrence of @xmath92 has to be split by an occurrence of @xmath99 ( otherwise the permutation given by @xmath29 would leave the second strand invariant ) , so @xmath29 must contain a subword of the form @xmath107 . note that @xmath108 must be greater or equal to two , since we applied all possible braid relations @xmath109 .figure [ nichtprim ] depicts the brick diagram and intersecting pattern of this subword for @xmath110 and @xmath111 .since the intersection pattern is not connected , there has to be another occurrence of @xmath99 in @xmath29 , otherwise the closure @xmath30 would not be prime .what are the possibilities for the other occurrences of @xmath99 ?if the first occurrence of @xmath92 is split by an occurrence of @xmath99 , again the occurrence of @xmath99 has to be to the power at least two .hence , @xmath29 contains , up to reversing order and cyclic permutation , the subword @xmath112 and the fibre surface of @xmath30 contains the minor @xmath49 , implying @xmath102 .similarly , if @xmath29 contains , up to reversing order and cyclic permutation , the subword @xmath113 , again the fibre surface of @xmath30 contains the minor @xmath49 , implying @xmath102 .if we exclude these cases , the only two possibilities for the induced subword of @xmath29 in the first three generators are @xmath114 and @xmath115 , which are , up to cyclic permutation , reverse to each other .if @xmath116 is greater or equal to two , the fibre surface of @xmath30 again contains the minor @xmath49 , implying @xmath102 , so we assume the induced subword of @xmath29 in the first three generators to be , up to reversing order and cyclic permutation , @xmath117 .but in this case , @xmath29 restricted to the second and third generator has a path as linking pattern and is not minimal by lemma [ minimalitylemma ] ._ case 3 : @xmath97 @xmath118_. the only possibility not considered in case 2 is the following : @xmath29 contains , up to cyclic permutation , the subword @xmath119 , thus also @xmath120 and the fibre surface of @xmath30 contains @xmath20 as a minor , implying @xmath102 .however , when reconsidering our discussion of case 2 , the powers of @xmath92 appearing could be greater , so we get @xmath121 as possibilities for the induced subword of @xmath29 in the first three generators , where @xmath122 . again, note that if @xmath116 or @xmath123 is greater or equal to two , the fibre surface of @xmath30 contains the minor @xmath49 , implying @xmath102 , so we assume the induced subword of @xmath29 in the first three generators to be , up to reversing order and cyclic permutation , @xmath124 .again , @xmath29 restricted to the second and third generator has a path as linking pattern and is not minimal ._ case 4 : @xmath125 @xmath126_. we can apply the same arguments as in the cases above . from thisit follows that if the fibre surface of @xmath30 contains no minor @xmath20 , then the induced subword in the first three generators is , after the described process of deleting some generators @xmath92 , either @xmath127 or @xmath128 .as before , these two words are , up to cyclic permutation , reverse to each other .but since we might have deleted some generators @xmath92 to obtain them , we should consider them separately .again as before , if @xmath116 or @xmath123 is greater or equal to two , the fibre surface of @xmath30 contains the minor @xmath49 , implying @xmath102 .if we restrict @xmath129 to the second and third generator , the linking pattern is a path . note that reinserting the deleted generators @xmath92 would split @xmath130 or @xmath131 . in any case, the linking pattern of @xmath29 restricted to the second and third generator is still a path and @xmath29 is not minimal .this does not necessarily hold for the other possibility @xmath132 .however , note that if @xmath133 is greater or equal to two , then @xmath134 contains the subword @xmath135 and the fibre surface of @xmath30 contains the minor @xmath49 , implying @xmath102 .so we are left with the possibility @xmath136 .if all the deleted occurrences of @xmath92 appeared before the first occurrence of @xmath99 in @xmath134 , after a cyclic permutation the linking pattern of @xmath29 restricted to the second and third generator again is a path and @xmath29 is not minimal .if some deleted occurrence of @xmath92 appeared after the first occurrence of @xmath99 in @xmath134 , then @xmath29 contains the word @xmath135 as a subword and , as before , @xmath102 ._ case 5 : @xmath125 @xmath98_. in this case , there is one last new possibility : as in case 1 , the word @xmath137 could be a subword of @xmath29 ( without directly yielding @xmath112 as a subword ) . again , since we applied all possible braid relations @xmath100 , @xmath108 and @xmath116 are greater or equal to two . if @xmath29 should , up to cyclic permutation , neither contain @xmath112 nor @xmath101 as subword , then @xmath108 and @xmath116 are both equal to two and the induced subword of @xmath29 in the first two generators is exactly @xmath138 .in particular , we again did not have to delete any occurrence of @xmath92 in the deletion process described above .if the induced subword of @xmath29 in the first three generators was @xmath139 , the permutation given by @xmath29 would leave the third strand invariant and @xmath30 would not be a knot . thus , there has to be at least one more occurrence of a generator @xmath99 .this gives the last two possibilities of induced subwords of @xmath29 in the first three generators : @xmath140 and @xmath141 , which are , up to cyclic permutation , reverse to each other .if @xmath29 is of index four , then actually @xmath29 would have to equal @xmath142 .but the closure @xmath143 can never be a knot , since the last two strands get permuted among themselves independently of @xmath144 and @xmath145 .now let @xmath29 be of index at least five .if @xmath142 is the induced subword of @xmath29 in the first three generators , one of the occurrences of @xmath146 has to be separated by an occurrence of @xmath147 to the power at least two ( recall that we ruled out the possibilty of a braid relation @xmath148 ) , since otherwise the first two strands would get permuted among themselves by @xmath29 and @xmath30 would not be a knot .one can then see that @xmath29 contains , up to reversing order and cyclic permutation , one of the subwords @xmath149 or @xmath150 , each of which guarantees the existence of a minor @xmath20 in the fibre surface of @xmath30 , implying @xmath102 .in the previous section , we established a characterisation of positive braid knots with maximal topological 4genus by the forbidden surface minors @xmath42 and @xmath43 .another way to think of this is that for positive braid knots , genus defect @xmath151 is characterised by these four forbidden minors .a natural question to ask is whether a similar result holds for larger genus defect .as noted by baader and dehornoy @xcite , this is implied by higman s lemma if we restrict ourselves to positive braids of index bounded by some natural number @xmath154 .a possible way of dealing with question [ characterisation ] in the unbounded case could be to give a positive answer to the following question , which can be thought of as a strengthening of proposition [ index4_prop ] .99 s. baader : _ positive braids of maximal signature _ , enseign .* 59 * ( 2013 ) , no .34 , 351358 .s. baader , p. dehornoy : _ minor theory for surfaces and divides of maximal signature _ ,s. baader , p. feller , l. lewark , l. liechti : _ on the topological 4genus of torus knots _ , http://arxiv.org/abs/1509.07634 .j. c. cha , c. livingston : _ knotinfo : table of knot invariants _ , http://www.indiana.edu/~knotinfo , 12.11.2015 .r. cromwell : _ positive braids are visually prime _ , proc .london math .( 3 ) * 67 * ( 1993 ) , no . 2 , 384424 . m. h. freedman : _ the topology of four - dimensional manifolds _ , j. differential geom .* 17 * ( 1982 ) , no . 3 , 357453 . l. h. kauffman , l. r. taylor : _ signature of links _ , trans .* 216 * ( 1976 ) , 351365 .b. kronheimer , t. s. mrowka : _ the genus of embedded surfaces in the projective plane _ , math .* 1 * ( 1994 ) , no . 6 , 797808 .l. rudolph : _ some topologically locally flat surfaces in the complex projective plane _, comment .* 59 * ( 1984 ) , no . 4 , 592599 .j. stallings : _ constructions of fibred knots and links _ , algebraic and geometric topology , 5560 , proc .pure math .* 32 * ( 1978 ) amer .soc . , providence , r.i .

<S> we show that a positive braid knot has maximal topological 4genus exactly if it has maximal signature invariant . as an application , we determine all positive braid knots with maximal topological 4genus and compute the topological 4genus for all positive braid knots with up to 12 crossings . </S>

1412.3042

graphene , a two - dimensional crystal of carbon atoms arranged in a honeycomb lattice , owes its extraordinary optical and electronic properties to the presence of dirac points in its bandstructure .the strong linear @xcite and nonlinear interaction @xcite of the massless dirac fermions ( mdfs ) with light , together with a tunable conductivity and broadband response , make graphene an attractive material for atomically - thin active devices operating at optical and terahertz frequencies , such as saturable absorbers @xcite , modulators @xcite , metamaterial devices @xcite , photo - detectors @xcite , and sensing applications @xcite .graphene s strong interaction with light is epitomized by its ability to support plasmons that are bound to the mono - atomic sheet of carbon @xcite .these collective excitations of the two - dimensional electron gas in form of charge density waves feature a strong spatial confinement of the electromagnetic energy to typically @xmath0 of the freespace wavelength or less @xcite , group velocities several hundred times lower than the vacuum speed of light @xcite , and tunable propagation characteristics controllable by chemical doping ( i.e. , using ionic gels ) or application of a gate voltage ( dc doping ) @xcite .although these properties are very attractive from an application perspective , graphene plasmons suffer from high losses at infrared wavelengths @xcite attributed to the presence of multiple damping pathways @xcite , such as collisions with impurities and phonons , as well as particle / hole generation via interband damping . arguably , the success of graphene as material for plasmonic applications depends on the development of strategies to control or compensate plasmonic losses .one such strategy is to supply gain via optical pumping .when graphene is excited by a short optical pulse , a hot non - equilibrium particle / hole distribution is created , that thermalizes rapidly within the bands to form an inverted carrier plasma in quasi - equilibrium , which provides gain over a wide range of frequencies @xcite . as a result of the inverted carrier state , plasmons are now not only subjected to interband absorption but can also experience amplification via stimulated emission @xcite .just as with interband absorption , the stimulated plasmon emission is equally enhanced by the concentration of field energy at the sheet that greatly increases the coupling of the plasmons to the particle / hole plasma .the possibility of overcoming plasmon losses at terahertz and infrared frequencies in photo - inverted graphene has been explored in refs .these studies offer first theoretical insight into the interplay of plasmons with particle / hole excitations via stimulated emission , but employ approximative expressions for the polarizability ( or conductivity ) , which do not establish the proper plasmon dispersion of inverted graphene .as both , the plasmon gain and emission rates depend critically on the plasmon dispersion and the derived density of states , it remains a matter of debate whether plasmon amplification is achievable under realistic conditions , i.e. under consideration of collision loss , doping , and temperature .the stimulated emission of plasmons in photo - inverted graphene is necessarily accompanied by the spontaneous emission of plasmons .contrary to plasmon amplification , spontaneous emission of plasmons is a broadband phenomenon that involves incoherent emission into all available modes .the concentration of the plasmon field energy to small volumes , their low group velocity , and the broadband gain all impact on the local density of optical states at the graphene sheet ( purcell factor ) , that causes the acceleration of the spontaneous plasmon emission processes . indeed , theoretical studies suggest that spontaneous plasmon emission provides an ultrafast channel for carrier recombination @xcite , that influences the non - equilibrium dynamics of hot carriers @xcite .only recently , transient carrier inversion has been observed in graphene by time- and angular - resolved photo - emission spectroscopy ( tr - arpes ) @xcite , and pump - probe experiments @xcite .these experiments reveal that while the photo - excited hot carriers ( i.e. , dirac particles and holes ) thermalize quickly ( on a @xmath1 timescale ) within the conduction and valence band , they simultaneously undergo ultrafast recombination processes , limiting the life - time of the inversion to the @xmath2 timescale .as these recombination rates are too fast to be explained by optical phonon emission , which occurs on @xmath3 timescales , other recombination processes are likely to be responsible for the rapid loss of inversion .a candidate for this is the auger recombination of carriers , where the energy of a recombined particle / hole pair is transferred to a particle ( or hole ) that is then lifted into a higher state within its band .however , the rates for auger recombination depend critically on the model for the screened coulomb interaction between the carriers and thus remain subject to discussion @xcite .carrier recombination due to plasmon emission is another mechanism that affects the carrier life - times @xcite .first theoretical calculations predict ultrafast recombination rates in the @xmath4 range , depending on temperature and doping @xcite . the wide range of timescales and their relevance for the carrier relaxation dynamics motivate the refined calculations on the plasmon emission rates presented in this work under consideration of the exact non - equilibrium plasmon dispersion . the relevance of plasmons for both non - equilibrium carrier dynamics and many - body effects in graphene can not be overstated .plasmons do not only contribute to the spontaneous recombination of carriers , but also to the self - energy of the carriers @xcite .furthermore , plasmons are associated with the poles of the screened coulomb potential , which in turn affects the interaction processes between charged particles , such as such as carrier - carrier scattering @xcite , auger recombination and impact ionization , as well as scattering with optical phonons @xcite and charged impurities @xcite . while the plasmon dispersion has been extensively studied in thermal equilibrium @xcite ,there is , until now , no general formalism that allows for the efficient calculation of the plasmon dispersion ( or the screening ) for _ arbitrary _ carrier distributions far from thermal equilibrium , although such a theory would be important to accurately calculate the interaction processes in hot carrier distributions created by ultra - short optical excitation . in this articlewe investigate the properties of the plasmons in photo - inverted , gapless ( i.e. , free - standing ) graphene .the presented study can be broadly split into two parts .the first part seeks to clarify how carrier inversion affects the non - equilibrium plasmon dispersion and decay rate , and establishes the conditions under which coherent plasmon amplification becomes possible .the second part of this work concerns the incoherent , spontaneous emission of plasmons , which competes with the plasmon amplification for gain . as a basis for our theoretical analysiswe introduce a general theoretical framework for the non - equilibrium polarizability that is applicable to arbitrary carrier distributions in graphene ( section [ sec : noneqplas ] ) . in section [ sec : plasinverted ] this theory is used to calculate the exact ( complex - frequency ) plasmon dispersion for photo - inverted intrinsic graphene and compare the it to the plasmon dispersion in thermal equilibrium .in particular , we analyze how the stimulated excitation / de - excitation processes impact on the gain / loss spectrum ( see fig . [fig : gainspectrum ] ) and then how the doping level of the particle / hole plasmas affects the plasmon dispersion curves and the associated gain spectra .the initial studies for a zero temperature and collision - free particle / hole plasma are followed by calculations that incorporate temperature and collision loss ( see section [ sec : collandtemp ] ) .the results exemplify that plasmon amplification is possible under realistic assumptions of temperature and collision loss , even for relatively low levels of inversion . in section [ sec : plasgainspec ]we evaluate the plasmon emission spectra and spontaneous emission rates and compare the exact results with approximative results obtained from fermi s golden rule ( fgr ) . as part of this analysis we study the impact of collision loss and temperature on the spontaneous emission rates , and establish that spontaneous plasmon emission is significantly faster than previous estimates suggest .plasmon decay rate @xmath5 over plasmon energy @xmath6 for inverted intrinsic graphene . at energies below @xmath7 ( where @xmath8 is the chemical potential of particles and holes )plasmons undergo stimulated emission processes , accompanied by interband recombination of carriers . above energies of @xmath7stimulated absorption processes associated with inter- and intraband recombination of particle / hole pairs dominate . ]graphene supports two types of plasmons , associated with the longitudinal density - density response and the transverse current - current response of the mdf plasma @xcite . within the field of plasmonics , these collective excitationsare frequently referred to as tm ( transverse magnetic ) and te ( transverse electric ) plasmons according to the polarization of the electromagnetic fields .in contrast to the conventional strongly bound tm plasmons , te plasmons are weakly bound and can only exist under certain restrictive conditions , such as low temperatures or a high - permittivity environment @xcite . for the present study on plasmon gain ,we only consider tm - polarized longitudinal plasmons , as these plasmons couple more strongly to the particle / hole plasma @xcite and thus constitute the dominant channel for stimulated and spontaneous emission @xcite .the dispersion of tm graphene plasmons can be obtained by solving maxwell s equations for bound tm waves .for graphene suspended in air , neglecting wave retardation ( i.e. , @xmath9 ) , this gives @xcite @xmath10 where @xmath11 , the non - local sheet conductivity , describes the linear current response to an applied electric field with in - plane wavevector @xmath12 and frequency @xmath13 .the same plasmon dispersion equation can be obtained from linear response theory of 2d electron gases @xcite where longitudinal plasmons emerge as collective charge density waves , whose dispersion is determined by the zeros of the dielectric function @xcite , @xmath14 introducing @xmath15 as the bare 2d coulomb potential and @xmath16 , the irreducible polarizability . both eqs .( [ eq : plasdispsigma ] ) and ( [ eq : epsilonequalzero ] ) are assuming a linear response of the carriers to an external excitation .thus , the non - local linear response functions @xmath11 and @xmath16 relate to each other via @xmath17 , as a comparison of eqs .( [ eq : plasdispsigma ] ) and ( [ eq : epsilonequalzero ] ) shows . assuming a weakly interacting 2d electron gas , the irreducible polarizability is , in random - phase approximation ( rpa ) , replaced by the leading order term .we note that the rpa is a commonly employed ( see e.g. , ref .@xcite ) and surprisingly accurate approximation for graphene @xcite . in the following we lay out the theory for calculating the exact complex - frequency plasmon dispersion , first for the equilibrium system , and then for arbitrary non - equilibrium carrier distributions .one of the main objectives of this work is to determine the gain spectrum of plasmons in photo - inverted graphene .this involves consideration of regimes where plasmons are strongly amplified or damped as they couple to the particle / hole plasma via stimulated emission and absorption processes .we therefore do not make the usual low - loss approximation , which treats the plasmon frequency as a purely real variable and extracts the gain / decay rate perturbatively , but instead seek the _ exact _ complex - frequency plasmon dispersion ( cfpd ) , as explained in the following .our starting point is eq .( [ eq : epsilonequalzero ] ) . in rpa , the irreducible polarizability @xmath16 is replaced by its leading order term , the polarizability of the non - interacting particle / hole plasma , which we subsequently refer to as , simply , _ the polarizability_. assuming an arbitrary ( non - equilibrium ) distribution @xmath18 of dirac fermions , the polarizability is obtained from @xcite @xmath19(q,\omega)=\frac{g}{a}\sum_{s , s'=\pm}\sum_{\mathbf{k } } \frac{m_{\mathbf{k},\mathbf{k}+\mathbf{q}}^{ss ' } \left[n(\epsilon_{\mathbf{k}}^{s})-n(\epsilon_{\mathbf{k}+\mathbf{q}}^{s'})\right ] } { \epsilon_{\mathbf{k}}^{s}-\epsilon_{\mathbf{k}+\mathbf{q}}^{s'}+\hbar\omega+i\times0 } \:,\ ] ] which describes the bare ( unscreened ) response of the particle / hole plasma to density fluctuations induced by an external disturbance .the expression is a weighted sum over all intra- and interband transitions @xmath20 where @xmath21 ( @xmath22 ) labels the conduction ( valence ) band and includes the spin / valley degeneracy in the prefactor @xmath23 .close to the dirac point , the energy dispersion of the conduction and valence bands is approximately linear ( @xmath24 ) and the square of the transition matrix element is @xmath25/2 $ ] @xcite .the notation @xmath26(q,\omega)$ ] indicates that the polarizability is a functional of the distribution function . in thermal equilibrium ,the distribution function @xmath18 is given by the fermi - dirac distribution , @xmath27 + 1)$ ] , parametrized by the chemical potential @xmath8 and temperature @xmath28 ; and we define @xmath29 $ ] for brevity . at zero temperature ,closed - form expressions have been derived for the equilibrium polarizability @xcite , while for finite temperatures it reduces to a semi - analytical form @xcite . for real wavevectors@xmath12 , the complex frequency zeros of eq .( [ eq : epsilonequalzero ] ) define the cfpd @xmath30 , whose real part @xmath31 is the frequency dispersion and imaginary part @xmath32 is the temporal decay rate . per definitiona negative decay rate implies plasmon gain .finding the complex - frequency solution requires a polarizability function that is well - defined on the complex frequency plane .the expressions given in refs .@xcite , for example , are restricted to real @xmath12 and @xmath13 as they are defined piecewise or contain heaviside functions which have no unique complex representation . in regimes where plasmons do not couple to the particle / hole plasma, this is not a problem as one can assume @xmath33 and then solve eq .( [ eq : epsilonequalzero ] ) approximately by extrapolating around values on the real frequency axis @xcite . using a first order taylor expansion the approximate frequency dispersion @xmath31 is obtained by solving @xmath34\approx0 \:,\ ] ] while the corresponding imaginary part , the decay rate @xcite , emerges as @xmath35}{\frac{\partial\mathrm{re}[\pi(q,\omega)]}{\partial\omega}}\right|_{\omega=\omega_{\mathrm{pl}}(q ) } \:.\ ] ] in section [ sec : plasgainspec ] we show that the rhs is in fact equivalent to the fgr expression for the net stimulated absorption rate of plasmons . in regimes where @xmath36=0 $ ] the decay rate @xmath5 is zero and the solutions of eq .( [ eq : reepsrpaequalzero ] ) are exact . for the purpose of finding the exact plasmon gain / loss spectra, we require an expression for the polarizability that applies to complex frequency values . in ref .@xcite an equation for the zero - temperature equilibrium polarizability is reported , which , for the case of gapless graphene , can be written in compact form , @xmath37 where @xmath38 is the dimensionless polarizability function , and @xmath39 . here, the chemical potential is assumed to be positive and one imposes @xmath40 to reflect particle / hole symmetry . for @xmath41this equation takes the same values as the equations in refs .@xcite assuming that the branch - cuts of @xmath42 are ( as usual ) oriented along the negative real axis @xcite . however , in contrast to other formulations , eq .( [ eq : pyatkovsky ] ) is analytic in the entire upper complex - frequency half - plane ( @xmath43>0 $ ] ) and allows for analytic continuation into the lower half - plane ( @xmath43<0 $ ] ) , where plasmons experience loss . the dielectric function ( [ eq : epsilonequalzero ] ) associated with eq .( [ eq : pyatkovsky ] ) has a scale invariance that becomes apparent when expressing frequency and wavevector in units of the chemical potential @xmath44 and fermi - wavevector @xmath45 .introducing dimensionless variables @xmath46 and @xmath47 one obtains @xmath48 where @xmath49 is the effective fine - structure constant of graphene in air @xcite . as the solutions of the dispersion relation ( [ eq : epsilonrpa-1 ] )no longer explicitly depend on the chemical potential , they can be represented by a single plasmon dispersion curve @xmath50 .this is a result of the conical dirac dispersion which remains invariant when rescaling energy and momentum variables by the same factor . to obtain the cfpd we solve eq .( [ eq : epsilonequalzero ] ) numerically using a complex root finding algorithm ( newton - raphson ) . before tracing the dispersion curves we rotate the branch - cuts by @xmath51 making them point down into the lower frequency half - plane .starting from @xmath52 and @xmath53 we find the next point of the dispersion by solving eq .( [ eq : epsilonequalzero ] ) . whenever @xmath13 enters the half - space the branch - cuts occupy , we rotate each branch - cut by @xmath54 , depending from which side its branch - point is passed , so that they now lie in the opposite half - space .this procedure is repeated as we scan through @xmath12 and ensures that the complex - frequency dispersion curve @xmath55 never crosses a branch - cut and thus retains a continuous , physical behavior throughout the entire wavevector regime . in the followingwe explain how we can generalize the polarizability to finite temperatures and non - equilibrium carrier distributions .the dynamic polarizability @xmath26 $ ] of an arbitrary non - equilibrium carrier distribution @xmath18 can be represented as the sum of the intrinsic zero - temperature polarizability @xmath56 and polarizabilities of the particle and hole plasmas @xmath57 $ ] and @xmath58 $ ] , where @xmath59 . ]the polarizability [ eq . ( [ eq : lindhardformula ] ) ]depends on the distribution of carriers and captures their response to electromagnetic excitation .it also enters the expression for the screened coulomb potential @xmath60 @xcite , which in turn affects the interaction processes between charged particles , such as carrier - carrier scattering , auger recombination , impact ionization and optical phonon scattering . over recent years, intense efforts have been made to calculate the polarizability of graphene , first in the limit of zero doping and zero temperature @xcite , then for finite doping @xcite , at finite temperatures @xcite , and in thermal quasi - equilibrium @xcite . here , we present transformations that allow us to calculate polarizabilities associated with _arbitrary _ non - equilibrium carrier distributions .these transformations are then used to calculate the cfpd of photo - inverted graphene in thermal quasi - equilibrium .let us first consider the general case where the carrier distribution is not in thermal equilibrium ( see fig .[ fig : noneqpol ] ) , but of arbitrary form @xmath18 . as the polarizability @xmath61 $ ] , eq .( [ eq : lindhardformula ] ) , is a linear functional of the non - equilibrium carrier distribution function @xmath18 it fulfills the property @xmath62=\int_{-\infty}^{+\infty}\mathrm{d}\epsilon\,\pi[\delta(\epsilon-\circ)]n(\epsilon ) \:.\ ] ] where ` @xmath63 ' marks the variable to which the functional applies . for brevitywe here omit the wavevector / frequency arguments [ see eq .( [ eq : lindhardformula ] ) ] .the dirac - delta function @xmath64 can be represented as the derivative of the fermi - dirac distribution @xmath65 in the zero - temperature limit .this allows us to deduce the generic formula , @xmath66=\int_{-\infty}^{+\infty}\mathrm{d}\epsilon\,\frac{\partial\pi|_{\mu=\epsilon}^{t=0}}{\partial\epsilon}n(\epsilon)\:,\label{eq : noneqpol}\ ] ] for the non - equilibrium polarizability , where @xmath67 $ ] denotes the zero - temperature polarizability associated with the equilibrium distribution of carriers with chemical potential @xmath8 .alternatively , using integration by parts , one may cast eq .( [ eq : noneqpol ] ) into the form @xmath68=-\int_{-\infty}^{+\infty}\mathrm{d}\epsilon\,\pi|_{\mu=\epsilon}^{t=0}\frac{\partial n(\epsilon)}{\partial\epsilon } \:,\ ] ] where we assumed that @xmath69 ( no occupied states at infinite energy ) and @xmath70 .we justify the latter by inspection of eq .( [ eq : lindhardformula ] ) . as @xmath71the distribution functions necessarily vanish as the ( finite ) valence band becomes empty .this limit is not reflected in the closed - form expression for the zero - temperature polarizability [ eq . ( [ eq : pyatkovsky ] ) ] , where the integration over energy states is assumed to extend to infinity and has been carried out before the limit @xmath71 is applied . as the derivation of eq .( [ eq : noneqpol ] ) is solely based on the linearity of the response function @xmath26 $ ] it is universally applicable to any bandstructure and carrier distribution .it should be noted , that the linearity only holds within the rpa and corrections due to self - interactions , which alter the electronic band structure , are not included . for graphene , the combination of eqs .( [ eq : pyatkovsky ] ) and ( [ eq : noneqpol ] ) proves particularly useful as it enables the numerical evaluation of finite - temperature and non - equilibrium polarizabilities for complex frequencies via analytic continuation . owing to graphene s vanishing band - gap , eq .( [ eq : noneqpol ] ) can be separated into integrals over positive and negative energies . introducing the hole distribution function @xmath59 and exploiting particle / hole symmetry, we cast eq .( [ eq : noneqpol ] ) into the following elegant form , @xmath72 & = \pi|_{\mu=0}^{t=0}+\\ & \quad\int_{0}^{\infty}\mathrm{d}\epsilon\,\left[\frac{\partial\pi|_{\mu=\epsilon}^{t=0}}{\partial\epsilon}n(\epsilon)+\frac{\partial\pi|_{\mu=\epsilon}^{t=0}}{\partial\epsilon}\bar{n}(\epsilon)\right]\\ & = \pi|_{\mu=0}^{t=0}+\pi^{(e)}[n]+\pi^{(h)}[\bar{n}]\ : .\end{split } \label{eq : noneqpol-2}\ ] ] the non - equilibrium polarizability can thus be represented as the sum of the zero - temperature intrinsic polarizability @xmath73 and contributions @xmath57 $ ] and @xmath58 $ ] for the particles and holes as depicted in fig .[ fig : noneqpol ] .this formulation is used later in this work when calculating the plasmon dispersion of photo - inverted graphene at finite temperatures .the theory presented in this section provides a simple yet powerful tool for the evaluation of the dielectric function for an arbitrary non - equilibrium carrier distribution @xmath18 .although we apply it here to find the plasmon dispersion of graphene when the mdf plasma is in an inverted quasi - equilibrium state , it may equally aid the evaluation of the screened coulomb interaction in mdf plasmas far from thermal equilibrium .the zero - temperature cfpd of extrinsic ( doped ) graphene in thermal equilibrium has been previously presented in ref .@xcite , albeit only inside the dirac cone ( @xmath74 ) . for photo - inverted graphenethe cfpd has , to the best of our knowledge , not yet been studied .for the rest of this paper we assume a particular non - equilibrium state , known as quasi - equilibrium .the quasi - equilibrium is an approximation that is commonly applied when describing band - gap semiconductors that are pumped into a state of inversion @xcite . assuming that carrier - carrier scattering is significantly faster than interband recombination processes , the carriers are able to thermalize within their bands to separate fermi distributions , each with their own chemical potential .despite the fact that recombination can be ultrafast in graphene @xcite , carrier - carrier scattering times are still one or two orders of magnitude faster ( typically tens of femtoseconds ) @xcite . due to this separation of time - scales ,quasi - equilibrium is established some tens of femtoseconds after pulsed excitation . in this context , the derived plasmon dispersion and emission rates are momentary quantities that , together with @xmath75 , @xmath76 , and @xmath28 , evolve in time as the carrier system returns back to equilibrium .owing to particle / hole symmetry , carriers thermalize to a common temperature , and the quasi - equilibrium distribution function takes the form @xmath77 where @xmath78 ( @xmath79 ) denotes the chemical potential of the dirac particles ( holes ) . in order to assess the differences between the plasmon dispersion in equilibrium and the inverted state , we first concentrate on the zero - temperature case , where the expressions for the polarizabilities take closed - form . inserting the zero - temperature quasi - equilibrium distribution function @xmath80 into eq .( [ eq : noneqpol-1 ] ) yields the expression @xmath81 \ : , \end{split}\ ] ] where the zero - temperature equilibrium case can be recovered by setting @xmath82 . in the following , we first study the plasmon dispersion of photo - inverted intrinsic graphene ( @xmath83 ) andcompare the result with the equilibrium plasmon dispersion of extrinsic graphene ( @xmath84 ) .the plasmon dispersion in the more general case of photo - inverted extrinsic graphene ( @xmath85 ) is examined thereafter .consider the case where intrinsic graphene is optically excited . as particles/ holes are generated in pairs , their quasi - equilibrium chemical potentials will be identical ( i.e. , @xmath86 ) once the carrier plasmas are thermalized within their respective bands . the zero - temperature quasi - equilibrium polarizability [ see eq .( [ eq : zerotpolquasieq ] ) ] thus simplifies to @xmath87 we insert this expression into the dielectric function ( [ eq : epsilonequalzero ] ) and introduce dimensionless variables to remove the explicit dependency on the chemical potential .this allows us to plot a single representative plasmon dispersion curve for the photo - inverted intrinsic case . using analytic continuation we calculated the cfpd for both the ( doped ) equilibrium ( @xmath88 ) and photo - inverted intrinsic case ( @xmath89 ) by tracing the complex - frequency roots of the dielectric function [( [ eq : epsilonequalzero ] ) ] .the corresponding results are shown in fig .[ fig : equilibvsinverted](a ) and ( b ) , respectively . for comparison ,we also show in each figure the real - frequency dispersion obtained in the low - loss approximation ( dashed black lines ) ; the dispersion in the classical limit using the drude conductivity ( dotted black lines ) ; and the dispersion in the local approximation ( @xmath90 ) using the optical conductivity ( dash - dotted black lines ) .frequency and wavevector variables are rescaled to dimensionless quantities @xmath91 and @xmath92 . for small @xmath93the solutions converge to the classical limit ( dotted black line ) , as expected . to highlight the differences between the cfpd and the approximative dispersion curves , we first inspect the plasmon dispersion in equilibrium .as predicted , the cfpd ( red line ) matches the real - frequency dispersion ( dashed black line ) within the loss - free regime ii .when the cfpd enters regime iii it becomes complex - valued with an imaginary part arising from the interband generation of particle / hole pairs ( landau - damping @xcite ) and thus begins to deviate from the real - frequency solution .note , that owing to the @xmath94 singularity in the polarizability [ see eq .( [ eq : pyatkovsky ] ) ] the real - frequency dispersion ( dashed black line ) can not cross the dirac cone but asymptotically approaches @xmath95 as @xmath96 .the cfpd , in contrast , eludes the dirac cone singularity due to its lossy character and crosses smoothly from regime iii into the intraband excitation regime iv . after reaching its peak within the intraband regime iv ,the decay rate starts to decrease as the phase - space for intraband excitation processes shrinks and eventually becomes zero at the point where the dispersion enters the loss - free region ( v ) . forthe photo - inverted intrinsic case [ see fig .[ fig : equilibvsinverted](b ) ] the plasmon dispersion ( solid red line ) initially follows the classical limit ( dotted black line ) .in contrast to the equilibrium case , however , plasmons with small @xmath93 induce particle / hole recombination processes via stimulated emission and thus experience a negative damping , i.e. , amplification . within region ( i ) , the cfpd experiences gain and thus begins to diverge from the other approximative solutions . after reaching its peak value ,the gain steadily decreases and eventually becomes zero at @xmath97 . above thisthreshold interband recombination processes are no longer possible at @xmath98 due to a lack of carrier inversion .the plasmon dispersion then enters the interband excitation regime iii where plasmons are damped due to generation of particle / hole pairs .again , the cfpd crosses the dirac cone , passes through the intraband excitation regime iv and exits into the loss - free region ( v ) , where neither inter- nor intraband excitation can take place as energy and momentum conservation can not be fulfilled simultaneously .we note , that the real - frequency solution already deviates from the cfpd within region ( i ) in particular when crossing @xmath99 where the associated decay rate shows a distinct dip .the dispersion derived from the optical conductivity predicts gain in the long - wavelength region ( @xmath90 ) but fails to display a decrease in gain and does not reproduce the loss in region ( iv ) .this is because intraband processes are forbidden in the applied local limit that only allows vertical transitions . in this sectionwe studied the plasmon dispersion of photo - inverted intrinsic graphene at zero temperature and compared the result with graphene in equilibrium .we briefly summarize the main findings : ( 1 ) within the rpa the calculated cfpd curves represents the _ exact _ plasmon dispersion , with an imaginary part that reflects the plasmon loss and gain rate ; ( 2 ) in contrast to the real - frequency ( low - loss ) approximation it crosses through both inter- _ and _ intraband excitation regimes , where plasmons are damped due to stimulated absorption processes ; and ( 3 ) for photo - inverted graphene plasmons with frequencies lower than @xmath100 can become amplified due to stimulated emission .we next study the impact of carrier imbalance ( i.e. , doping ) on the plasmon dispersion of photo - inverted extrinsic graphene . chemical doping or the application of a gate voltage changes the equilibrium carrier density . when pumped into a inverted state , the chemical potentials of particles and holes that characterize the quasi - equilibrium will necessarily differ , i.e. , @xmath85 . assuming @xmath101 , eq .( [ eq : zerotpolquasieq ] ) simplifies to @xmath102 as before , the dispersion equation can be rescaled by introducing appropriate dimensionless variables using the sum of chemical potentials @xmath103 as a scale parameter .a simple substitution of arguments allows us to recast eq .( [ eq : zerotpolinv ] ) into the form @xmath104 where @xmath105 is the carrier imbalance parameter .the expression for the so - defined dimensionless polarizability @xmath106 contains the equilibrium polarizability @xmath107 as defined in eq .( [ eq : pyatkovsky ] ) .the last term is the intrinsic polarizability , which , as a consequence of the scaling , now appears in the limit @xmath108 .the parameter @xmath109 quantifies the relative difference of the chemical potentials of particles and holes .a value of @xmath110 represents the photo - inverted intrinsic case ; @xmath111 the doped equilibrium cases where only one plasma component is excited ; and values in between , the general case .owing to particle / hole symmetry , eq .( [ eq : zerotpolinvrescaled ] ) only depends on @xmath112 and we can restrict ourselves to positive values of @xmath109 without loss of generality .figure [ fig : dopedcase](a ) depicts the calculated frequency dispersion curves @xmath31 for varying values of @xmath109 .as the plasmon frequency dispersion curves are strictly monotonic , we can express the loss dispersion @xmath113 as function of frequency , i.e. , @xmath114 , as shown in fig .[ fig : dopedcase](b ) . when varying @xmath109 , all loss spectra pass through zero at @xmath115 , where the cfpd curves pass from the gain into the loss region . for small frequencies values ( @xmath116 )the dispersion curves form a single bundle [ see fig .[ fig : dopedcase](a ) ] , in the sense that an infinitesimal change of @xmath109 will lead to an infinitesimal variation of @xmath117 .one may expect that the solutions undergo a continuous variation when varying the inversion parameter @xmath109 .instead , the plasmon dispersion curves split into two bundles at @xmath118 , cross - over and become well separated at larger frequencies / wavevectors . the first bundle ( solid red lines )contains dispersion curves with @xmath119 , where @xmath120 .dispersion curves in this bundle make a direct transition from the trapezoidal gain region ( i ) into the lossy interband excitation region ( iii ) .the second bundle ( solid black lines ) , comprising dispersion curves with @xmath121 , is associated with a large imbalance in particle / hole numbers .these curves start in the gain region ( i ) but pass through the loss - free region ( ii ) before entering the loss region ( iii ) .a qualitative difference between the low-@xmath109 and the high-@xmath109 bundle is that curves in the latter enter region ( iii ) with a zero imaginary part , and therefore glance the branch point which separates the two regions ( yellow line ) , whereas curves in the former bundle avoid it , as prescribed in section [ sub : cfd ] .the split - up of the bundle occurs at the intersection point of regions ( i ) , ( ii ) , and ( iii ) [ filled magenta circle in fig .[ fig : dopedcase](a ) ] where the polarizability has a degenerate branch - singularity .the critical value @xmath122 is calculated by finding the dispersion curve that passes through this branch - singularity , i.e. , from the condition @xmath123 note , that while the plasmon dispersion curves are continuous ( and smooth ) in @xmath93 , a change in @xmath109 does not necessarily cause a continuous variation of the curves .adjacent curves can flank the branch - singularity from opposite sides and thus end up on different riemann sheets of the dielectric function ( [ eq : epsilonequalzero ] ) giving rise to the observed splitting of the bundle .this is most evident from fig .[ fig : dopedcase](b ) where the loss dispersion curves are separated into two bundles just above @xmath118 and then follow distinct trends . the plasmon decay rate @xmath124 [ see fig . [fig : dopedcase](b ) ] strongly depends on both the plasmon frequency @xmath125 and the carrier imbalance parameter @xmath109 . for frequencies below ( above ) @xmath118 , plasmon amplification ( damping ) occurs as indicated by red ( blue ) curves in fig .[ fig : dopedcase](c ) . at @xmath110 ( intrinsic inverted case )the gain spectrum features a single broad peak .as @xmath109 increases this peak first red shifts and then splits into two peaks , which , for even larger values of @xmath109 , become separated by a loss - free frequency region . for values of @xmath109 larger than @xmath122 the second peak vanishes and the loss - free region extends in frequency , reducing the gain available at low frequencies until the loss - free region ( ii ) covers the whole frequency range @xmath116 when @xmath126 ( doped equilibrium case ) . the discontinuity in @xmath109 stretches out from the branch - singularity at @xmath127 towards frequencies @xmath128 ( dashed line ) .it should be pointed out that this discontinuity is rooted in the singularities of the polarizability , which are physical and can not be lifted . as the polarizability appears in the helmholtz free energy @xcite the discontinuity may relate to a thermodynamic phase - transition .this clearly requires a more in depth analysis , which goes beyond the scope of this paper and will be explored in a follow - up study . concluding this section, we briefly summarize the main findings : ( 1 ) the complex - frequency solutions of the dispersion equation deliver a consistent picture of the frequency and loss dispersion of plasmons ; ( 2 ) plasmons are amplified in photo - inverted graphene at frequencies @xmath129 due to stimulated emission ; and ( 3 ) the plasmon gain spectrum is strongly influenced by the carrier imbalance parameter @xmath130 , exhibiting discontinuous behavior at @xmath131 when the plasmon dispersion passes through a singularity .impact of temperature and collision loss on plasmon decay rate and frequency dispersion .( a ) plasmon decay rate at @xmath132 for collision loss rates ranging from @xmath133 to @xmath134 , where @xmath135 .( b ) plasmon decay rate at @xmath132 for temperatures ranging from @xmath136 ( red lines ) to @xmath137 ( black lines ) , where @xmath138 .insets display the corresponding frequency dispersion curves for @xmath132 ( red lines ) and @xmath139 ( blue lines ) . ]so far , all calculations have been carried out under the idealized assumptions of zero temperature and without the inclusion of collision loss . in this sectionwe aim to clarify how finite - temperatures and collision losses impact on the spectral characteristics of plasmon gain in photo - inverted graphene . without the inclusion of collision losses , equilibrium plasmons are completely loss - free within regions ( ii ) and ( v ) of figs .[ fig : noneqpol ] and [ fig : dopedcase ] , where the landau damping is suppressed . in reality , collisions with impurities , acoustic phonons and optical phonons @xcite limit the life - time and propagation length of graphene plasmons , with a collision time @xmath140 that can range from tens to hundreds of femtoseconds depending on the quality of the sample @xcite and substrate used @xcite . to theoretically assess the effect of carrier collisions ( quantified by the phenomenological rate @xmath141 ) on the plasmon dispersion and loss we impose the transformation @xcite@xmath142 on the polarizability @xmath143 , which , in contrast to a simple replacement @xmath144 , conserves particle numbers locally . in cases where @xmath143 only accepts real frequency values , a taylor expansion @xmath145 needs to be carried out , effectively limiting the validity of eq .( [ eq : mermin-1 ] ) to small collision rates @xmath141 @xcite .this limitation does not exist when using eq .( [ eq : pyatkovsky ] ) in conjunction with the non - equilibrium polarizabilities , such as eq .( [ eq : zerotpolinv ] ) .their validity extends to the complex frequency plane and thus permits a direct application of the transformation ( [ eq : mermin-1 ] ) . in solving eq .( [ eq : epsilonequalzero ] ) together with eq .( [ eq : mermin-1 ] ) we directly obtain the cfpd for the photo - inverted case with collision losses .figure [ fig : collandtemp](a ) depicts the plasmon gain spectrum for a carrier imbalance of @xmath132 .the collision rates are varied from @xmath133 to @xmath134 ( @xmath135 ) .the maximum value corresponds to a collision time of only @xmath146 at an inversion of @xmath147 .we observe that the collision time only moderately affects the shape of the loss dispersion curve and that , at each given frequency , the decay rate @xmath148 is proportional to @xmath141 [ see fig . [fig : collandtemp](a ) ] . at the highest collision rate of @xmath134there is still a frequency region where plasmons are amplified ( i.e. , have a negative decay rate ) .eventually , for even higher collision rates , interband gain will no longer be able to compensate the collision loss and the plasmons will become lossy across the entire frequency spectrum .surprisingly , the frequency dispersion is almost unaffected by the introduction of collision loss as can be seen from the bundle of dispersion curves @xmath117 , which come to lie on top of each other when varying the collision rate [ see red lines in inset of fig .[ fig : collandtemp](a ) ] .we finally note , that the splitting of the dispersion bundle observed in section [ sec : plasinverted ] is robust against the introduction of collision loss and occurs roughly at the same critical value of @xmath120 . to give evidence for this behavior we show the dispersion curves associated with a value of @xmath139 ( blue lines ) in the inset of fig .[ fig : collandtemp](a ) , which clearly belong to the high-@xmath109 bundle and thus follow a different path than those for @xmath132 ( red lines ) .the influence of temperature on the plasmon dispersion of graphene has first been analyzed in ref .@xcite , where a semi - analytical formula for the finite - temperature polarizability has been derived . here, we seek a formulation that applies to photo - inverted graphene in thermal quasi - equilibrium and is valid for complex frequencies .we employ eq .( [ eq : noneqpol-2 ] ) , which , owing to the linear character of the functionals , can be cast into the form @xmath149+\pi^{(h)}[\delta f|_{\mu_{h}}^{t } ] \:,\ ] ] where @xmath150 in this expression the zero - temperature quasi - equilibrium @xmath151 is augmented by corrections that capture the change of the polarizability due to smearing of the fermi - edge as quantified by @xmath152 .this corrective approach enables us to trace the cfpd at finite temperatures with high accuracy .the evaluation of eq .( [ eq : finitetemppol ] ) requires the derivative of the polarizability ( [ eq : pyatkovsky ] ) , which is given by the following closed - form expression @xmath153 where @xmath154 and @xmath155 .just as eq .( [ eq : pyatkovsky ] ) , the equation above is analytic on the upper frequency half - plane and thus reproduces the correct values for real frequencies when using the prescription @xmath41 . note , that the positions of the branch cuts of @xmath156 vary with the integration variable ( i.e. , the chemical potential @xmath8 ) in eq .( [ eq : noneqpol-2 ] ) .the analytic continuation needs to be carried out inside the integrals as the cfpd traverses the complex plane . solving eq .( [ eq : epsilonequalzero ] ) together with eqs .( [ eq : finitetemppol])-([eq : rescaleddiffpyatkovsky ] ) gives the finite - temperature cfpd curves , plotted in fig .[ fig : collandtemp](b ) for @xmath132 and temperatures @xmath138 ranging from @xmath136 ( red line ) to @xmath137 ( black line ) .assuming @xmath147 this translates to a temperature range of @xmath157 .we first note that the loss curves @xmath148 associated with different temperatures all pass through zero at @xmath115 . with temperatures @xmath158 the smearing of the fermi - edgesdoes seemingly not impact the point where stimulated emission and absorption processes are balanced . for frequencies below @xmath115 the plasmon gain is strongly affected by temperature , displaying a distinctive blue - shift of the gain peak with increasing temperature .in addition , the region of zero gain / loss that occurs for @xmath136 at around @xmath159 vanishes for finite temperature leading to a strongly broadened gain spectrum .although the shape of the gain spectrum is strongly affected by temperature , the frequency dispersion @xmath117 remains fairly unaffected [ see red lines in [ fig : collandtemp](b ) ] , just as in the case of collision loss . the splitting behavior observed in section [ sec : plasinverted ] for the idealized case ( zero - temperature and zero collision loss ) proves to be remarkably robust against temperature as can be seen from the second set of dispersion curves ( [ fig : collandtemp]b ; blue lines ) , which for @xmath139 follow a different trajectory .the critical value @xmath122 at which the splitting occurs shifts slightly towards larger values and is found to be @xmath160 for @xmath137 compared to @xmath120 at @xmath136 . for a more realistic description of the plasmon dispersion, one may wish to include both finite collision loss and finite temperature .the procedure for this is straight - forward : first one applies eq .( [ eq : mermin-1 ] ) to obtain the equilibrium polarizability with finite collision loss and thereafter the transformation ( [ eq : finitetemppol ] ) to generalize for a finite temperature .the calculated cfpd curves account for both the frequency and loss dispersion of the plasmons . the losses and gain in regimes i , iii and iv of fig . [ fig : equilibvsinverted ] , due to single particle excitation ,can be related to the three fundamental processes of light - matter interaction : absorption , stimulated emission and spontaneous emission .in photo - inverted graphene , at frequencies below @xmath161 , plasmons experience gain due to stimulated emission processes and thus acquire a negative decay rate @xmath5 ( region ( i ) in fig . [fig : equilibvsinverted ] ) .considering the situation where all plasmon modes are in their ground state , the rate of spontaneous plasmon emission into a frequency interval @xmath162 $ ] is equal to the stimulated emission rate weighted with the plasmon density of states . in the following ,we extract the spontaneous plasmon emission spectra from the cfpds , compare them with first order approximative results obtained from fermi s golden rule ( fgr ) , and calculate the total spontaneous carrier recombination rates . for the following interpretationit is instructive to draw the connection between fgr for plasmon emission and eq .( [ eq : lossfunction ] ) . within the semiclassical frameworkthe decay of plasmon population due to stimulated processes can be expressed as @xmath163 where the net stimulated absorption rate @xmath164 is the stimulated absorption minus the stimulated emission rate . as @xmath164 is an intensity related quantity ,it is twice the plasmon decay rate , i.e. , @xmath165 in cases where @xmath33 the plasmon decay rate is approximated to first order by eq .( [ eq : lossfunction ] ) . inserting the imaginary part of the polarizability ( [ eq : lindhardformula ] ) @xmath166= & \frac{\pi}{\hbar}\frac{g}{a}\sum_{s , s'=\pm}\sum_{\mathbf{k}}\hbar\delta(\hbar\omega+\epsilon_{\mathbf{k}}^{s}-\epsilon_{\mathbf{k}+\mathbf{q}}^{s'})\\ & \times m_{\mathbf{k},\mathbf{k}+\mathbf{q}}^{ss'}\left[f(\epsilon_{\mathbf{k}}^{s})-f(\epsilon_{\mathbf{k}+\mathbf{q}}^{s'})\right ] \end{split } \label{eq : impi}\ ] ] into eq .( [ eq : lossfunction ] ) reproduces the fgr expressions for the stimulated intra- and interband processes ( see also ref .@xcite ) . in applying the identity @xmath167 to the difference of fermi functions in eq .( [ eq : impi ] ) , one can then split the net stimulated rate into rates for the absorption and emission processes . focusing on interband emission processes we thus identify the fgr expression associated with the emission of plasmons @xmath168}{\partial\omega}}\right|_{\omega=\omega_{\mathrm{pl}}(q)}\ : , \end{split } \label{eq : emratefgr}\ ] ] where @xmath169 . to evaluate this expression , the sum over @xmath170-states is replaced by an integration over momentum space . after performing a series of algebraic transformations , detailed in appendix [ sub : derivationofemrate ] , we find the following semi - analytical equation for the plasmon emission rate : @xmath171}{\partial\omega}}\biggr|_{\omega=\omega_{\mathrm{pl}}(q)}\:.\end{split } \label{eq : emratefgrsemian}\ ] ] the function @xmath172 , a measure for the phase - space associated with the emission processes , is defined as @xmath173 it is clear from the calculation above that fgr ( [ eq : emratefgr ] ) is an first order approximation in @xmath174 that is accurate as long as the plasmon loss / gain rates are much smaller than the plasmon frequency. from the plasmon emission rate @xmath175 one can derive the plasmon emission spectrum @xmath176 as follows : summation of @xmath175 over all wavevector states defines the spontaneous plasmon emission rate @xmath177 which in turn is obtained by integrating the plasmon emission spectrum @xmath176 over frequency . from the equation above it is clear that the plasmon emission spectrum @xmath176 is just @xmath178 where the plasmon density of states @xmath179 quantifies how many plasmons per unit area and time are spontaneously emitted into the ( infinitesimal ) frequency - interval @xmath162 $ ] . dividing the spontaneous plasmon emission rate @xmath180 by the particle area density yields the recombination rate @xmath181 where @xmath182 is the area density of mdfs in the conduction band .carrier recombination rates due to plasmon emission have been previously calculated in ref .@xcite using fgr with the limitation that only the intraband contribution of the polarizability to the plasmon dispersion had been taken into account .the reported rates suggest that plasmon emission is an ultrafast channel for carrier recombination that needs to be considered when analyzing the hot carrier dynamics in graphene @xcite . in the following ,we determine the exact plasmon emission spectra and recombination rates directly from the cfpd and compare the result with approximative solutions obtained from fgr . with the exact solution as a benchmark , we are able to assess the limitations of fgr , which as we will show , depends on the accuracy of the plasmon dispersion used for its evaluation .spontaneous plasmon emission spectra @xmath183 ( a ) and electron recombination rates @xmath184 ( b ) at zero temperature and in dependence of carrier imbalance @xmath109 .the respective dimensionless quantities are defined as @xmath185 and @xmath186 ( see appendix [ sec : rescalingofgain ] ) .exact results ( i ) derived from the exact complex-@xmath13 plasmon dispersion curves , together with approximative results obtained from fermi s golden rule in conjunction with ( ii ) the classical ( drude ) dispersion , ( iii ) the complex-@xmath13 dispersion , and ( iv ) the real-@xmath13 dispersion . ] at zero temperature the fermi - surfaces of the particle / hole plasmas are sharply defined .depending on the frequency , the graphene plasmons will either experience gain ( @xmath161 ) or loss ( @xmath187 ) as shown in fig . [fig : dopedcase ] .it is also clear from fig .[ fig : dopedcase ] that for @xmath188 plasmons only exist left of the dirac cone where intraband processes can not occur .this implies that for @xmath161 plasmons are only subjected to emission processes and we can write @xmath189 for the plasmon emission rate . as @xmath190we can determine the plasmon emission rate @xmath175 directly from the cfpd curves ( see fig .[ fig : dopedcase ] ) and then calculate the exact plasmon emission spectrum @xmath176 with the help of eq . ([ eq : emspectrum ] ) .this also puts us into the position to assess the accuracy of fgr when using different approximations to the plasmon frequency dispersion with the exact result as a reference . in fig .[ fig : emspectra](a ) we plot the plasmon emission spectra @xmath183 of photo - inverted extrinsic graphene at zero temperature in dimensionless scaling ( see appendix [ sec : rescalingofgain ] ) .the exact result [ ( i ) , red lines ] is shown together with approximative results , all obtained from fgr [ see eq .( [ eq : emratefgrsemian ] ) ] using the following different solutions for the frequency dispersion @xmath31 : ( ii ) the approximate dispersion obtained in the classical ( drude ) limit of the conductivity ( dotted line ) , ( iii ) the exact dispersion obtained from the real part of the cfpd ( dash - dotted line ) ; and ( iv ) the real - frequency ( low - loss ) approximation of the dispersion(dashed line ) . of these three emission spectra ,the classical approximation is worst , as it severely underestimates the plasmon emission rate , since the plasmon density of states that enters eq .( [ eq : emspectrum ] ) is about a factor of 10 lower in the classical limit .solution ( ii ) reproduces the shape of the emission spectrum apart from small deviations ( dips ) that occur when the plasmon dispersion crosses the boundaries of the regions depicted in fig .[ fig : dopedcase ] . solution ( iii ) features pronounced spikes , which are induced by the first - order approximation [ see eqs .( [ eq : reepsrpaequalzero ] ) and ( [ eq : lossfunction ] ) ] . integrating the emission spectrum @xmath183 over all frequenciesgives the spontaneous carrier recombination rate @xmath184 , which we plot as a function of carrier imbalance @xmath109 .we first note , that while @xmath191 holds due to particle / hole symmetry , the particle recombination rate @xmath192 itself is not symmetric in @xmath109 as the recombination rates are obtained from eq .( [ eq : recrate ] ) by dividing the emission rate @xmath193 by the particle density . as a result , the curves in fig .[ fig : emspectra](b ) are skewed towards negative @xmath109-values , indicating that particle / hole recombination is faster for the p - doped case , where holes are the majority carriers .using the classical approximation ( ii ) for the dispersion in conjunction with fgr ( dotted black line ) underestimates the rates by a factor of 5 at m=0 . the approximative results ( iii ) and ( iv ) reproduce the shape and magnitude of the exact emission rate ( i ) quite well , apart from a small shift in @xmath109 towards negative values .this means that the first - order approximation ( iv ) does provide a good estimate for the carrier recombination rate , despite not reproducing the plasmon emission spectra correctly .the scale - free representation implies that the emission spectrum @xmath176 scales with @xmath194 and the recombination rate @xmath195 with @xmath196 .the actual values for the carrier recombination rates can be directly extracted from fig .[ fig : emspectra](b ) .assuming for example an inversion of @xmath147 ( and @xmath110 ) the calculated spontaneous recombination times are @xmath197 for the exact solution ( i ) and @xmath198 when using the drude approximation ( ii ) .the latter is in good agreement with those presented in ref .@xcite ( see fig . 5 therein ) , which are based on fgr and a plasmon dispersion for which only intraband contributions to the polarizability have been considered .this result implies that spontaneous plasmon emission is significantly faster than previously assumed and constitutes an important recombination channel that needs to be considered together with auger recombination @xcite and optical phonon emission @xcite .impact of collision loss and temperature on the plasmon gain spectra .( a ) plasmon gain spectra @xmath199 for @xmath110 and collision loss rates ranging from @xmath133 to @xmath134 .( b ) plasmon gain spectra for @xmath110 and temperatures ranging from @xmath136 to @xmath137 .( c ) particle recombination rates over carrier imbalance parameter @xmath109 for temperatures from @xmath136 to @xmath137 .the dimensionless spectral gain and carrier recombination rates are defined as @xmath185 and @xmath186 . ]we next examine the influence of collision loss and temperature on the plasmon emission spectra and the carrier recombination rates .in contrast to the ideal case ( zero - temperature , collision - free ) it is not possible to determine the exact emission spectra for finite collision loss or temperature as relation ( [ eq : emratezerotexact ] ) no longer holds under these conditions . however , in the first part of this section we concluded that the combination of fgr and complex - frequency dispersion [ ( i ) and ( iii ) in fig .[ fig : emspectra ] ] provides a good approximation for both the emission spectra and recombination rates ( see fig .[ fig : emspectra ] ) . as the collision loss does not enter the fgr expression ( [ eq : emratefgrsemian ] ) explicitly andonly weakly affects the plasmon dispersion ( as shown in section [ sec : collandtemp ] ) , we expect the spontaneous emission spectrum to have a weak dependence on the collision loss rate .indeed , fig .[ fig : speccollandtemp](a ) shows that an increase of the collision loss only causes a small increase of the emission spectrum that is most prominent around the emission peak . as a result, we do not observe a significant change of the carrier recombination rate with increasing collision rate , and therefore do not present this .for finite temperature , the situation is different ; the fermi distribution of the particle / hole plasmas that enter eq .( [ eq : kfun ] ) are smeared out . with increasing temperature, the peak gain decreases due to a reduction of the carrier occupation below @xmath115 [ see fig . [fig : speccollandtemp](b ) ] . on the other hand ,emission processes are now possible above @xmath115 , which was the zero - temperature limit for spontaneous emission . as a consequence , the spectral gain[ see fig .[ fig : speccollandtemp](b ) ] becomes increasingly broadened for increasing temperature , showing both a characteristic blue - shift and reduction of the peak gain .integrating the emission spectrum gives the total carrier recombination rate , which we show in fig .[ fig : speccollandtemp](c ) as a function of the carrier imbalance @xmath109 .we observe the recombination rate rises with temperature .the effect is most pronounced around @xmath200 where the particles are minority carriers and the temperature has a strong influence on the number of particles in the conduction band .the peak recombination rate shifts towards more negative values of @xmath109 as the temperature increases . at @xmath110 ,the recombination rates are least dependent of the value of @xmath28 due to the symmetry in the particle / hole population . as interpretation of fig .[ fig : speccollandtemp ] we give dimensional quantities for @xmath201 . at zero temperature ,the spectral gain peaks at @xmath202 with a value of @xmath203 and cuts off sharply at @xmath204 .the curve with @xmath137 corresponds to @xmath205 .the carrier recombination times of @xmath206 for @xmath110 , i.e. @xmath207 at @xmath98 , drop to @xmath208 for @xmath209 , i.e. @xmath210 and @xmath211 , for @xmath205 .we briefly summarize the main results of this section : ( 1 ) a comparison of the plasmon emission spectra at zero temperature shows that the accuracy of fermi s golden rule depends critically on the plasmon dispersion with best results when the exact dispersion is inserted ; ( 2 ) the calculated carrier recombination rates for plasmon emission are more than a factor of 5 larger than those obtained in the classical limit ; ( 3 ) collision loss has no influence on the plasmon emission spectra and carrier recombination rates ; and ( 4 ) an increase in temperature leads to a broadening and blue - shift of the emission spectrum and increases the carrier recombination rates , particularly at a high carrier imbalance , i.e. , when @xmath212 .in calculating the gain spectra of photo - inverted graphene self - consistently from the exact complex - frequency dispersion curves , this work provides evidence that graphene can , under realistic conditions , support plasmons with gain .as the dispersion crosses through regimes where plasmons couple to the particle / hole plasma via stimulated emission and absorption processes , it acquires an imaginary part that represents the gain and loss spectrum .based on a comprehensive theory for the non - equilibrium polarizability , we systematically studied the influence of doping , collision loss and temperature on both the plasmon dispersion and the gain / loss spectrum . while doping and temperature affect the shape of the emission spectrum , collision loss leads to a reduction of gain that is proportional to the collision rate .the frequency dispersion curves , in turn , are robust against collision loss and temperature but are distinctly affected by doping .when the imbalance in the particle / hole chemical potentials reaches a critical value , the plasmon dispersion passes through a singularity and undergoes a sudden change .our results show that plasmon amplification is possible under assumption of realistic collision loss and temperature .carrier inversion does not only enable plasmon amplification via stimulated emission but also leads to spontaneous emission of plasmons . to investigate this incoherent channel, we extracted the spontaneous plasmon emission spectra and associated carrier recombination rates directly from the complex - frequency dispersion and by application of fermi s golden rule .we found that the emission spectra are weakly dependent on the collision rate , but strongly influenced by doping and temperature .our results suggest that spontaneous plasmon emission is a significant channel for particle / hole recombination in photo - excited graphene , with rates that exceed those previously reported by a factor of 5 . in the light of these results, it appears evident that spontaneous plasmon emission plays an important role for the relaxation of the photo - excited plasma back to equilibrium , as observed in pump - probe and tr - arpes experiments .all calculations were carried out for free - standing graphene , disregarding the hybridization of plasmons with optical surface phonons that occurs when graphene is placed on a substrate .this aspect and the implications of the observed splitting of dispersion curves that occurs when varying the carrier imbalance , will be explored in future works .we believe that this work casts new light onto the nature of non - equilibrium plasmons and may offer an explanation why the carrier inversion observed in pump - probe and arpes experiments is typically short - lived .the authors acknowledge helpful discussions with andreas pusch , jorge bravo - abad , and francisco garca - vidal .the authors also acknowledge the financial support provided by the epsrc , the german research foundation ( dfg ) and the leverhulme trust .in first order approximation , the plasmon decay rate ( [ eq : lossfunction ] ) is proportional to the imaginary part of the polarizability ( [ eq : lindhardformula ] ) . by replacing the sum over @xmath170 with an integrationone finds that the imaginary part of the polarizability eq .( [ eq : impi ] ) can be transformed into @xmath166= & \zeta\int\limits _{ -1}^{+1}\mathrm{d}u\int\limits _ { + 1}^{+\infty}\mathrm{d}v\:\bigl[i_{++}(u , v)\\ & \qquad+i_{--}(u ,- v)+i_{-+}(v , u)\bigr ] \ : , \end{split}\ ] ] where @xmath213 .the intraband contributions of the conduction and valence bands ( @xmath214 and @xmath215 ) as well as the interband contribution ( @xmath216 ) are obtained from @xmath217 \ : .\end{split}\ ] ] here , @xmath218 ( @xmath219 ) are the chemical potentials of the particles in the conduction ( valence ) band .this result is in agreement with the equations for the finite - temperature polarizability reported in ref .@xcite . for the followingwe only consider the interband contribution associated with the @xmath216 term .further , using the identity @xmath220 , we split off the contribution that relates to plasmon emission @xmath221 where @xmath222 evaluation of the delta - function finally yields @xmath223 as @xmath224 and @xmath225 . inserting this result into @xmath226}{\partial\omega}}\right|_{\omega=\omega_{\mathrm{pl}}(q)}\label{eq : spectralstimrate-1}\ ] ] yields eq .( [ eq : emratefgrsemian ] ) for the plasmon emission rate .the calculated spectra @xmath176 and spontaneous emission rates @xmath195 are rescaled to dimensionless quantities . for this purposewe first introduce dimensionless variables @xmath92 , @xmath91 and @xmath138 where @xmath227 . as a starting pointwe rescale the carrier recombination rate to an dimensionless quantity , @xmath228 we next replace the variables in the carrier density @xmath229 with their dimensionless counterparts .this gives @xmath230 where @xmath231 using the eqs .( [ eq : recratescaled ] ) and ( [ eq : densityscaled ] ) allows us to write @xmath232 for eq .( [ eq : recrate ] ) with @xmath233 on the other hand , @xmath234 introducing the rescaled dimensionless plasmon emission spectrum @xmath235 which is defined as @xmath236 with the density of states @xmath237 and @xmath238 . in fig .[ fig : emspectra ] and fig .[ fig : speccollandtemp ] we applied ( [ eq : gainscaled ] ) and ( [ eq : recratescaled ] ) to display the plasmon emission spectrum and recombination rate in dimensionless form .these results are scale - free and can be used to extract the plasmon emission spectrum and recombination rate for arbitrary @xmath196 .108ifxundefined [ 1 ] ifx#1 ifnum [ 1 ] # 1firstoftwo secondoftwo ifx [ 1 ] # 1firstoftwo secondoftwo `` `` # 1''''@noop [ 0]secondoftwosanitize@url [ 0 ] + 12$12 & 12#1212_12%12@startlink[1]@endlink[0]@bib@innerbibempty link:\doibase 10.1103/physrevb.76.153410 [ * * , ( ) ] link:\doibase 10.1103/physrevlett.100.117401 [ * * , ( ) ] link:\doibase 10.1126/science.1156965 [ * * , ( ) ] link:\doibase 10.1103/physrevlett.101.196405 [ * * , ( ) ] link:\doibase 10.1209/0295 - 5075/79/27002 [ * * , ( ) ] link:\doibase 10.1103/physrevlett.105.097401 [ * * , ( ) ] link:\doibase 10.1103/physrevb.82.201402 [ * * , ( ) ] link:\doibase 10.1002/lpor.201200079 [ * * , ( ) ] link:\doibase 10.1002/adfm.200901007 [ * * , ( ) ] link:\doibase 10.1038/nature10067 [ * * , ( ) ] link:\doibase 10.1038/nmat3433 [ * * , ( ) ] link:\doibase 10.1103/physrevb.85.075434 [ * * , ( ) ] link:\doibase 10.1126/science.1239501 [ * * , ( ) ] link:\doibase 10.1038/nphoton.2010.40 [ * * , ( ) ] http://pubs.acs.org/doi/abs/10.1021/nn4005704 [ * * , ( ) ] link:\doibase 10.1038/nnano.2014.31 [ * * , ( ) ] http://pubs.acs.org/doi/abs/10.1021/nn1010842 [ * * , ( ) ] http://pubs.acs.org/doi/abs/10.1021/nl300948c [ * * , ( ) ] link:\doibase 10.1063/1.3630230 [ * * , ( ) ] link:\doibase 10.1038/nature11254 [ * * , ( ) ] link:\doibase 10.1038/nature11253 [ * * , ( ) ] http://dx.doi.org/10.1038/nphoton.2012.262 [ * * , ( ) ] link:\doibase 10.1021/nn300989 g [ * * , ( ) ] http://pubs.acs.org/doi/abs/10.1021/nn406627u [ * * , ( ) ] link:\doibase 10.1088/0953 - 8984/26/12/123201 [ * * , ( ) ] http://pubs.acs.org/doi/abs/10.1021/ph400147y [ * * ( ) ] link:\doibase 10.1021/nl201771h [ * * , ( ) ] link:\doibase 10.1103/revmodphys.81.109 [ * * , ( ) ] link:\doibase 10.1103/revmodphys.83.407 [ * * , ( ) ] link:\doibase 10.1021/nl101559n [ * * , ( ) ] link:\doibase 10.1039/c0jm02922j [ * * , ( ) ] link:\doibase 10.1038/nnano.2011.146 [ * * , ( ) ] link:\doibase 10.1038/nphoton.2012.27 [ * * , ( ) ] link:\doibase 10.1038/nphoton.2013.57 [ * * , ( ) ] link:\doibase 10.1063/1.2717566 [ * * , ( ) ] link:\doibase 10.1103/physrevlett.108.167401 [ * * , ( ) ] link:\doibase 10.1103/physrevb.87.165413 [ * * , ( ) ] @noop * * , ( ) link:\doibase 10.1103/physrevb.84.045437 [ * * , ( ) ] http://stacks.iop.org/0953-8984/23/i=14/a=145302 [ * * , ( ) ] link:\doibase 10.1103/physrevb.86.195437 [ * * , ( ) ] link:\doibase 10.1021/nl8019399 [ * * , ( ) ] link:\doibase 10.1063/1.3615286 [ * * , ( ) ] link:\doibase 10.1103/physrevb.84.205406 [ * * , ( ) ] link:\doibase 10.1103/physrevb.84.075449 [ * * , ( ) ] link:\doibase 10.1103/physrevb.85.241404 [ * * , ( ) ] link:\doibase 10.1103/physrevb.85.125413 [ * * , ( ) ] link:\doibase 10.1103/physrevb.88.035430 [ * * , ( ) ] link:\doibase 10.1088/1367 - 2630/15/8/083038 [ * * , ( ) ] link:\doibase 10.1038/nphys477 [ * * , ( ) ] link:\doibase 10.1038/nmat3757 [ * * , ( ) ] link:\doibase 10.1103/physrevlett.111.027403 [ * * , ( ) ] link:\doibase 10.1039/c4fd00020j [ * * , ( ) ] link:\doibase 10.1063/1.2837539 [ * * , ( ) ] link:\doibase 10.1063/1.3122348 [ * * , ( ) ] link:\doibase 10.1103/physrevlett.102.086809 [ * * , ( ) ] link:\doibase 10.1103/physrevb.83.153410 [ * * , ( ) ] link:\doibase 10.1088/1367 - 2630/15/1/015009 [ * * , ( ) ] link:\doibase 10.1021/nl502740 g [ * * , ( ) ] link:\doibase 10.1103/physrevb.76.155431 [ * * , ( ) ] http://scitation.aip.org/content/aip/journal/jap/112/9/10.1063/1.4761995 [ * * , ( ) ] link:\doibase 10.1038/ncomms2987 [ * * , ( ) ] link:\doibase 10.1103/physrevb.77.081411 [ * * , ( ) ] link:\doibase 10.1103/physrevlett.110.146801 [ * * , ( ) ] link:\doibase 10.1103/physrevb.80.155417 [ * * , ( ) ] link:\doibase 10.1103/physrevlett.105.055501 [ * * , ( ) ] link:\doibase 10.1063/1.2809413 [ * * , ( ) ] link:\doibase 10.1103/physrevb.79.115447 [ * * , ( ) ] link:\doibase 10.1063/1.3291615 [ * * , ( ) ] link:\doibase 10.1103/physrevb.79.165404 [ * * , ( ) ] link:\doibase 10.1103/physrevlett.97.266406 [ * * , ( ) ] link:\doibase 10.1088/1367 - 2630/8/12/318 [ * * , ( ) ] link:\doibase 10.1103/physrevb.75.205418 [ * * , ( ) ] link:\doibase 10.1103/physrevlett.99.016803 [ * * , ( ) ] link:\doibase 10.1103/physrevb.75.033408 [ * * , ( ) ] link:\doibase 10.1088/0953 - 8984/21/2/025506 [ * * ( ) , 10.1088/0953 - 8984/21/2/025506 ] link:\doibase 10.1088/1751 - 8113/42/21/214015 [ * * , ( ) ] link:\doibase 10.1103/physrevb.80.245435 [ * * , ( ) ] link:\doibase 10.1103/physrevb.83.235409 [ * * , ( ) ] link:\doibase 10.1103/physrevb.86.195424 [ * * , ( ) ] link:\doibase 10.1088/2040 - 8978/15/11/114005 [ * * , ( ) ] link:\doibase 10.1103/physrevb.80.075418 [ * * , ( ) ] link:\doibase 10.1103/physrevb.82.155412 [ * * , ( ) ] link:\doibase 10.1103/physrevb.84.195407 [ * * , ( ) ] link:\doibase 10.1126/science.1186489 [ * * , ( ) ] link:\doibase 10.1103/physrevb.84.195446 [ * * , ( ) ] link:\doibase 10.1103/physrevb.85.075410 [ * * , ( ) ] http://adsabs.harvard.edu/abs/1967phrvl..18..546s [ * * , ( ) ] http://books.google.co.uk/books?id=kfkikrfgupsc[__ ] ( , ) http://journals.jps.jp/doi/abs/10.1143/jpsj.27.1393 [ * * , ( ) ] http://books.google.co.uk/books?id=cktubaaaqbaj[__ ] , oxford graduate texts ( , ) link:\doibase 10.1103/physrevlett.113.105502 [ * * , ( ) ] link:\doibase 10.1103/revmodphys.84.1067 [ * * , ( ) ] http://books.google.co.uk/books?id=t5_dagaaqbaj[__ ] , dover books on physics ( , ) link:\doibase 10.1103/physrevb.88.195405 [ * * , ( ) ] link:\doibase 10.1103/physrevlett.101.146805 [ * * , ( ) ] link:\doibase 10.1209/0295 - 5075/87/27005 [ * * , ( ) ] http://www.sciencedirect.com/science/article/pii/0550321394904103 [ * * , ( ) ] link:\doibase 10.1063/1.98515 [ * * , ( ) ] http://journals.aps.org/pra/abstract/10.1103/physreva.39.1887 [ * * , ( ) ] link:\doibase 10.1103/physrevlett.107.237401 [ * * , ( ) ] http://journals.aps.org/pr/abstract/10.1103/physrev.186.409 [ * * , ( ) ] link:\doibase 10.1103/physrevlett.98.216401 [ * * , ( ) ] link:\doibase 10.1103/physrevb.85.045410 [ * * , ( ) ] link:\doibase 10.1063/1.2990753 [ * * , ( ) ] link:\doibase 10.1103/physrevb.87.035414 [ * * , ( ) ] http://journals.aps.org/prb/abstract/10.1103/physrevb.1.2362 [ * * , ( ) ]

<S> graphene supports strongly confined transverse - magnetic sheet plasmons whose spectral characteristics depend on the energetic distribution of dirac particles . </S><S> the question arises whether plasmons can become amplified when graphene is pumped into a state of inversion . in establishing a theory for the dynamic non - equilibrium polarizability </S><S> , we are able to determine the exact complex - frequency plasmon dispersion of photo - inverted graphene and study the impact of doping , collision loss , and temperature on the plasmon gain spectrum . we calculate the spontaneous emission spectra and carrier recombination rates self - consistently and compare the results with approximations based on fermi s golden rule . </S><S> our results show that amplification of plasmons is possible under realistic conditions but inevitably competes with ultrafast spontaneous emission , which , for intrinsic graphene , is a factor 5 faster than previously estimated . </S><S> this work casts new light on the nature of non - equilibrium plasmons and may aid the experimental realization of active plasmonic devices based on graphene . </S>

1105.4761

cyg ob2 # 5 ( v729 cyg , bd + 40 4220 ) is a radio - bright early - type multiple system comprising an eclipsing contact binary with a 6.6-day orbital period .in addition to the primary radio component coincident with the contact binary , abbott et al .( 1981 ) and miralles et al .( 1994 ) reported on the existence of a radio `` companion '' located @xmath1 to the north - east of the contact binary .contreras et al .( 1997 ) determined that this secondary radio component has an elongated shape and lies in - between the binary system and a b - type star , first reported by herbig ( 1967 ) .the emission from the secondary component is non - thermal ( i. e. synchrotron ; white 1985 ) and has been ascribed to the interaction region formed by the wind from the contact binary and that of the b - type star .these interaction regions consist of a contact discontinuity with two shocks , one shock on either side of the contact discontinuity .the study of objects such as cyg ob2 # 5 allows to investigate the physics of particle acceleration in massive stars environments ( de becker 2007 ; benaglia 2010 ) .the study of sources of cosmic ray acceleration ( in addition to supernova remnants ) is very important since there is observational evidence that they are produced in a variety of galactic sources ( adriani et al .2011 ) .observations of cyg ob2 # 5 taken over a period of about 20 years show that the radio emission varies in time between a low- and a high - flux state ( persi et al .1990 ; contreras et al .1997 ) . in the low - flux state, the emission has a flux density of about 2 mjy at 4.8 ghz , and is due to thermal free - free emission from the ionized material in the stellar wind driven by the contact binary .the spectral index of the emission is @xmath2 ( kennedy et al .2010 ) , as expected for a wind with a density decreasing as @xmath3 ( panagia & felli 1975 ; olnon 1975 ; wright & barlow 1975 ) . during the high - flux state ,the total flux rises to about 8 mjy at 4.8 ghz , as an additional non - thermal component appears . because of the non - thermal nature of the additional component , the spectral index of the system is flatter in the high - flux state than during the low - flux state .recently , kennedy et al . ( 2010 ) reanalysed all existing vla observations of cyg ob2 # 5 and showed that the radio flux variations arise directly in the primary component with a periodicity of @xmath4 yr , while the flux from the secondary component to the north - east remains constant in time .these authors proposed that the variations can be represented by a simple model in which a source ( with constant non - thermal emission ) moves around the contact binary in a @xmath5-yr orbit and the varying radio emission results from the variable free - free opacity of the wind between the source and the observer ( kennedy et al . 2010 ) .the non - thermal emission requires the presence of a fourth star in the system , but two different scenarios are possible . in the first one ,the non - thermal emission would arise from a wind - collision region between the contact binary and the fourth star in the system .the fourth star would evidently have to be sufficiently massive to drive a wind powerful enough to produce an interaction region with the wind from the contact binary .the presence of a wind - collision region is supported by the detection of x - rays ( waldron et al .1999 ; linder et al .interestingly , kennedy et al . (2010 ) found evidence in the radial velocities of the contact binary for a reflex motion resulting from a putative star with a mass of @xmath6 , consistent with a late o or early b type star .alternatively , the fourth star itself could generate the non - thermal emission as a consequence of coronal activity .this would favor a scenario where the fourth star in the system would be a low - mass object since convection in the stellar interior is required for coronal activity to occur .two diagnostics can be used to distinguish between these two scenarios .the first one is related to variability .coronal radio emission from low ( feigelson & montmerle 1999 ) and some intermediate - mass ( dzib et al .2010 ) stars tends to be very variable , and to exhibit short - duration flares on timescale of hours ( osten & wolk 2009 ) .if the non - thermal source in cyg ob2 # 5 is from a star with coronal activity , variability on timescales of hours would be expected to occur during the high - flux state , when the star travels in front of binary system . in the low - flux state, the free - free opacity of the wind from the contact binary would hide the orbiting non - thermal source and we would solely detect the constant , thermal emission from the wind . in the wind interaction scenario , on the other hand ,no rapid variations are expected in the high - flux state since wind collision regions are relatively extended and generate fairly constant radio emission , provided timescales significantly shorter than the orbital period are considered .the second diagnostic is related to the size of the non - thermal emitting region . for cygob2 # 5 at 1.7 kpc ( torres - dodgen et al .1991 ) , the emission due to a wind interaction is expected to have an extent of order 10 mas ( pittard & dougherty 2006 ) , whereas emission from the active magnetosphere of a low - mass star would be more compact that 0.1 mas .rodrguez et al .( 2010 ) estimated an angular size of @xmath7 for the non - thermal source by subtracting the persistent thermal component ( low - flux state ) from the high - flux state ( thermal plus non - thermal emission present ) .both scenarios remain consistent with this estimate , so higher angular resolutions are needed . in this paper ,we first re - analyze ( in section [ sec:1 ] ) a total of 19 archival observations obtained during the high - flux state of cyg ob2 # 5 between 1983 and 2003 , to search for short timescale flux variability .detection of such variability would support the scenario where the non - thermal emission comes from a low - mass companion , whereas a non - detection would be in better agreement with the wind interaction scenario . in the following section ( [ sec:2 ] ) , we present the first very long baseline interferometry ( vlbi ) observation ever reported of cyg ob2 # 5 .this observation will enable us to put direct constrains on the morphology and size of the non - thermal emission in cyg ob2 # 5 .the data analyzed here comprise 19 vla observations of cyg ob2 # 5 at frequencies between 1.4 and 43.3 ghz during the high - flux state of the source , and were obtained from the nrao archive .the on - source integration time for these observing sessions varies between 11.5 to 157.7 minutes .the detailed parameters of the observations are given in ortiz - len ( 2010 ) .the archive data were edited and calibrated using the software package astronomical image processing system ( aips ) of nrao .the absolute amplitude calibrator for 16 of the 19 epochs was 1331@xmath8305 , while 0137@xmath8331 was used for the remaining three epochs .the phase calibrator was 2007@xmath8404 for all epochs .the adopted and bootstrapped flux densities of the amplitude and phase calibrators are given in ortiz - len ( 2010 ) .these observations have been analyzed by several authors ( e.g. persi et al .1990 ; kennedy et al . 2010 ) but using the entire on - source integration time of each session as a single data point . here, we are interested , instead , in the short timescale variability .our strategy to search for such variability was the following .first , images of cyg ob2 # 5 were made with the aips task imagr for each observation epoch . with the aips task jmfit ,the precise position of the primary component was determined .the @xmath9 data was then recentered at the position of the main component using the aips task uvfix . with these new @xmath9 data ,the real and imaginary parts of the interferometer data were plotted as a function of time , averaged over the @xmath9 plane .the real part gives us information on the flux density of the source and the imaginary part on its position and symmetry .we then averaged over time in bins from 5 to 20 minutes .a detailed description of this technique is given by neria et al .( 2010 ) . to avoid contamination from extended emission , in observations at 4.8 ghz with the d configuration, we used only @xmath9 data with baselines larger than 0.125 km ( suppressing the emission from structures larger than @xmath10 ) .furthermore , a self - calibration process was applied in four epochs to diminish remaining phase errors in the visibility .the imaginary component is consistent with zero , indicating that the source is symmetric about the phase center ( the origin of the visibility plane ) and has no significant structure on these spatial scales .[ fig_visibilities],scaledwidth=50.0% ] we searched for rapid variability in the light curves of the real part of the visibility , but found no statistical evidence for variations at typical levels of 10% .an example of this data analysis is shown in figure 1 .this lack of strong short - timescale variability clearly favors the scenario where the non - thermal radio emission comes from a wind interaction region .it should be pointed out , however , that while fast variability is common in non - thermal stellar sources ( i. e. feigelson & montmerle 1999 ; gdel 2002 ; bower et al .2003 ; forbrich et al .2008 ; gmez et al .2008 ) , there is at least one example known of a young stellar source generating steady non - thermal emission ( andr et al .1988 , 1991 ; loinard et al .thus , the lack of fast variability can not be considered as a definitive proof that the non - thermal emission comes from a wind interaction region .a direct , high - angular resolution image is needed to distinguish between a compact stellar source and a wind - collision region .we will present such an image in the coming section .cyg ob2 # 5 was observed at @xmath11 = 3.6 cm ( @xmath12 = 8.4 ghz ) with the very long baseline array ( vlba ) on december 13 , 2010 .the position of the peak of the source was at @xmath13 = @xmath14 , @xmath15 = + @xmath16 .the observation was 5 hours , and consisted of series of cycles with two minutes spent on source , and one minute spent on the main phase - referencing quasar j0218 + 3851 , located @xmath17 away .every 15 minutes , the radio bright x - ray binary cyg x-3 , which is @xmath18 from cyg ob2 # 5 was also observed .although cyg x-3 is very close to cyg ob2 # 5 it was not used as a primary calibrator due to its extreme variability in both flux density and morphology and the fact it is heavily scattered at the observing frequency ( desai & fey 2001 ) .the data were edited and calibrated using the aips software following standard procedure for phase - referenced vlba observations .the calibration determined from the observations of j0218 + 3851 was applied to both cyg ob2 # 5 and cyg x-3 .at this point most of the phase errors left are caused by the @xmath17 separation between the source and the phase calibrator . to remove most of this ,cyg x-3 was self - calibrated and the phase gains determined from that self - calibration were then interpolated and applied to cyg ob2 # 5 .this resulted in a significant improvement in the quality of the image of cyg ob2 # 5 compared to a direct phase transfer between j0218 + 3851 and cyg ob2 # 5 .cyg ob2 # 5 is quite resolved so the best image was made by limiting the maximum uv length to 60000 k@xmath11 and using pure natural weighting .the resulting image is shown in figure 2 .the total flux density of the source is @xmath195 mjy , in agreement with the value expected for the non - thermal component as derived from lower angular resolution observations .this result suggests that at these high angular resolutions we are observing synchrotron emission .the peak flux density of the source is 1.1 mjy beam@xmath20 , implying a brightness temperature , @xmath21 k. this brightness temperature is not large enough to clearly favor a non - thermal origin and vlbi observations at several wavelengths are needed to establish the nature of the emission beyond doubt .the source is clearly resolved with a total extent , mostly along the north - south direction , of about 10 mas ( consistent with the upper limit reported by rodrguez et al .the arc - like morphology of the emission is the structure expected for a wind interaction region . since the apex of the arc shape is located to the west , one would expect the source with the strongest wind ( the contact binary system in the present case ) to be located west of the arc . to examine this issue ,the position of the contact binary at the epoch of the vlba observations must be determined . , scaledwidth=50.0% ] from the vla observations mentioned in section [ sec:1 ] , we selected the 3.6 and 6 cm data obtained in the a configuration of the array . using these observations ,the position of the radio emission associated with the contact binary was determined at 5 epochs between 1983 and 2004 ( figure 3 ) and its proper motion was measured to be : @xmath22 this is in good agreement with the proper motions of several wolf rayet stars located in the same region ( dzib & rodrguez 2009 ) .the position of the source at the epoch ( 2010.95 ) of the vlba observation and its error bar is shown as a cross on figure 2 . within the errors , this position lies along the direction of the apex of the arc seen in the vlba image , as expected .the star responsible for the second colliding wind must be located east of the arc along its bisector , and presumably fairly close to the arc itself since its wind is likely to be significantly less powerful than that of the contact binary .assuming that the wind momentum rate of the star responsible for the second colliding wind is 0.1 that of the contact binary and following cant et al .( 1996 ) , we obtain a location at the position of the small white square shown in figure 2 .the detection of an arc - like structure in the vlba observations demonstrates that the non - thermal emission detected during the high - flux state of cyg ob2 # 5 most probably originates in a wind collision region , but raises an interesting additional issue .the separation between the wind collision region and the position of the contact binary ( figure 2 ) is only about 12 mas .the diameter of the region surrounding the contact binary where the free - free radiation is emitted was measured to be about 60 mas ( rodrguez et al .thus , the wind - collision arc appears to be located within the nominal region occupied by the free - free emission .the free - free opacity in that region should prevent the non - thermal emission from escaping and reaching the observer .a possible solution to this apparent conundrum is that the wind driven by the contact binary is very inhomogeneous ( blomme et al . 2010 ;muijres et al . 2011 ) , and consequently partially transparent at centimeter wavelengths .this is a problem that has to be studied in the future because the non - thermal excess emission is clearly seen even at frequencies as low as 1.4 ghz ( persi et al .1985 ; ortiz - len et al .2011 ) , where the free - free opacity is expected to be quite large . assuming that our estimate of the position of the fourth star in the system is reasonable , the current separation between that star and the contact binary is about 15 mas , or 25.5 au ( at a distance of 1.7 kpc ; torres - dodgen et al .if this provided a good estimate of the orbital semi - major axis , then using kepler s third law the total mass of the system would have to be about 400 m@xmath23 .this seems unreasonably large , and suggests that the orbit is somewhat eccentric , and the system currently near apastron .for instance , taking an eccentricity of 0.5 could reduce the mass of the system by up to a factor 8 , down to about 50 m@xmath23 , and consistent with a system of 3 o - type stars .a high eccentricity is in agreement with that deduced by kennedy et al .( 2010 ) from their best fit model for the variable non - thermal flux from cyg ob2 # 5 .interestingly , if for these estimates we adopt the shorter distance of 925 pc proposed by linder et al .( 2009 ) , we obtain a total mass of the system of only about 60 m@xmath23 , assuming that the current separation between the contact binary and the putative position of its companion is representative of the orbital semi - major axis .we are currrently monitoring the emission of cyg ob2 # 5 with vlbi observations and may be able to provide the geometric parallax and an accurate distance in the future .in this paper , we have first re - analyzed archival vla observations of the radio emission from cyg ob2 # 5 taken when the system was in its high - flux state .this re - analysis shows that the flux of the non - thermal emission in the system does not suffer strong short - term variability .we then presented the first vlbi image of this system , and demonstrated that the non - thermal component most probably originates in an arc - shaped wind - collision region resulting from the interaction between the wind driven by the known contact binary in the system , and that powered by a companion star in a 6.7-yr , somewhat eccentric orbit .vlbi observations at several wavelenghts are needed to establish unambiguously the nature of the emission from the arc - shaped region . to our knowledge, cyg ob2 # 5 is the only multiple system known so far to harbor two radio - imaged wind - collision regions , one described in this paper , and the other reported previously by contreras et al . ( 1997 ) .we thank our refere , michal de becker , for his valuable comments that significantly improved this paper .g.o.l , l.l ., l.f.r . , ands.d acknowledge the financial support of dgapa , unam and conacyt , mxico .is grateful to the guggenheim foundation for financial support .the national radio astronomy observatory is a facility of the national science foundation operated under cooperative agreement by associated universities , inc .abbott , d. c. , bieging , j. h. , & churchwell , e. 1981 , , 250 , 645 adriani , o. , et al .2011 , science , 332 , 69 andre , p. , montmerle , t. , feigelson , e. d. , stine , p. c. , & klein , k .-1988 , , 335 , 940 andre , p. , phillips , r. b. , lestrade , j .- f . , &klein , k .-1991 , , 376 , 630 benaglia , p. 2010, high energy phenomena in massive stars , edited by josep mart , pedro l. luque - escamilla , and jorge a. combi , asp conference series , 422 , 111 blomme , r. , de becker , m. , volpi , d. , & rauw , g. 2010 , , 519 , a111 bower , g. c. , plambeck , r. l. , bolatto , a. , mccrady , n. , graham , j. r. , de pater , i. , liu , m. c. , & baganoff , f. k. 2003 , , 598 , 1140 briggs , d. 1995 , ph.d .thesis , new mexico inst . of mining and technology cant , j. , raga , a. c. , & wilkin , f. p. 1996, , 469 , 729 contreras , m. e. , rodrguez , l. f. , tapia , m. , cardini , d. , emanuele , a. , badiali , m. , & persi , p. 1997, , 488 , l153 de becker , m. 2007 , , 14 , 171 desai , k. m. , & fey , a. l. 2001 , , 133 , 395 dzib , s. , & rodrguez , l. f. 2009 , , 45 , 3 dzib , s. , loinard , l. , mioduszewski , a. j. , boden , a. f. , rodrguez , l. f. , & torres , r. m. 2010 , , 718 , 610 feigelson , e. d. , & montmerle , t. 1999 , , 37 , 363 forbrich , j. , menten , k. m. , & reid , m. j. 2008 , , 477 , 267 gmez , l. , rodrguez , l. f. , loinard , l. , lizano , s. , allen , c. , poveda , a. , & menten , k. m. 2008 , , 685 , 333 gdel , m. 2002 , , 40 , 217 herbig , g. h. 1967 , , 79 , 502 kennedy , m. , dougherty , s. m. , fink , a. , & williams , p. m. 2010 , , 709 , 632 linder , n. , rauw , g. , manfroid , j. , damerdji , y. , de becker , m. , eenens , p. , royer , p. , & vreux , j .-2009 , , 495 , 231 loinard , l. , torres , r. m. , mioduszewski , a. j. , & rodrguez , l. f. 2008 , , 675 , l29 miralles , m. p. , rodrguez , l. f. , tapia , m. , roth , m. , persi , p. , ferrari - toniolo , m. , & curiel , s. 1994 , , 282 , 547 muijres , l. e. , de koter , a. , vink , j. s. , krtika , j. , kubt , j. , & langer , n. 2011 , , 526 , a32 neria , c. , gmez , y. , & rodrguez , l. f. 2010 , , 46 , 253 olnon , f. m. 1975 , , 39 , 217 ortiz - len , g. n. 2010 , b.s .thesis , universidad nacional autnoma de mxico ortiz - len , g. n. , rodrguez , l. f. , & tapia , m. 2011 , in preparation osten , r. a. , & wolk , s. j. 2009 , , 691 , 1128 panagia , n. , & felli , m. 1975 , , 39 , 1 persi , p. , ferrari - toniolo , m. , tapia , m. , roth , m. , & rodrguez , l. f. 1985 , , 142 , 263 persi , p. , tapia , m. , rodrguez , l. f. , ferrari - toniolo , m. , & roth , m. 1990 , , 240 , 93 pittard , j. m. , & dougherty , s. m. 2006 , , 372 , 801 rodrguez , l. f. , gmez , y , loinard , l. , & mioduszewski , a. j. 2010 , rmxaa , 46 , 215 torres - dodgen , a. v. , tapia , m. , & carroll , m. 1991 , , 249 , 1 waldron , w. l. , corcoran , m. f. , drake , s. a. , & smale , a. p. 1998, , 118 , 217 white , r. l. 1985 , , 289 , 698 wright , a. e. , & barlow , m. j. 1975 , , 170 , 41

<S> the radio emission from the well - studied massive stellar system cyg ob2 # 5 is known to fluctuate with a period of 6.7 years between a low - flux state when the emission is entirely of free - free origin , and a high - flux state when an additional non - thermal component ( of hitherto unknown nature ) appears . in this paper , we demonstrate that the radio flux of that non - thermal component is steady on timescales of hours , and that its morphology is arc - like . </S><S> this shows that the non - thermal emission results from the collision between the strong wind driven by the known contact binary in the system , and that of an unseen companion on a somewhat eccentric orbit with a 6.7-yr period and a 5 to 10 mas semi - major axis . </S><S> together with the previously reported wind - collision region located about @xmath0 to the north - east of the contact binary , cyg ob2 # 5 appears to be the only multiple system known so far to harbor two radio - imaged wind - collision regions . </S>

hep-ph0008112

the famous hagedorn hypothesis @xcite , dating back to pre - chromodynamic times of the sixties , states that at asymptotically large masses , @xmath0 , the density of hadronic resonance states , @xmath1 , grows exponentially : @xmath2 the hagedorn temperature , @xmath3 , is a scale controlling the exponential growth of the spectrum .although the hagedorn hypothesis has sound thermodynamical consequences ( one can not heat - up a hadronic system above this temperature ) , @xmath3 should not be immediately associated with thermodynamics . in thistalk we are concerned with the spectrum of particles _ per se _ , as read off form the particle data tables @xcite . in this contextthe `` temperature '' @xmath4 is just a parameter in eq .( [ hag ] ) .ever since hypothesis ( [ hag ] ) was posed , it has been believed that there is one universal hagedorn temperature for all hadrons ._ presently available experimental data show that this is not the case _ , as has been pointed out by w. florkowski and wb in refs .@xcite .[ cumul ] this talk has two parts : experimental and theoretical . in the experimental part ( sec .[ experiment ] ) we show how well the hagedorn hypothesis works even for very low masses , and point out the key observation that _ the mesonic temperature is significantly larger from the baryonic temperature_. in the theoretical part ( sec .[ theory ] ) we argue that the only framework ( known to us ) which is capable of producing the observed behavior in a natural way are the dual string models @xcite . in sec .[ otherapp ] we discuss other approaches and more speculative ideas .in fig . [ cumul ] we compare the _ cumulants _ of the spectrum @xcite , defined as the number of states with mass lower than @xmath0 .the experimental curve is @xmath5 where @xmath6 is the spin - isospin degeneracy of the @xmath7th state , and @xmath8 is its mass .the theoretical curve corresponds to @xmath9 where @xmath10 with @xmath11 denoting a slowly - varying function .a typical choice @xcite , used in the plot of fig .[ cumul ] , is @xmath12 parameters @xmath3 and @xmath13 are obtained with the least - square fit to @xmath14 , made over the range up to @xmath15 , and skipping the lightest particle in the set .other choices of @xmath11 give fits of similar quality ( see fig .[ varhag ] ) .a striking feature of fig . [ cumul ] is the linearity of @xmath16 starting at very low @xmath0 , and extending till @xmath17 .clearly , this shows that ( [ hag ] ) is valid in the range of available data .however , the slopes in fig .[ cumul ] are _ different _ for mesons and baryons . for the assumed @xmath11 of eq .( [ hr ] ) we get @xmath18 this means that @xmath19 , and the inequality is substantial ! although it has been known to researchers in the field of hadron spectroscopy that the baryons multiply more rapidly than mesons @xcite , to our knowledge this fact has not been presented as vividly as in fig .[ cumul ] . to emphasize the strength of the effect we note that in order to make the meson line parallel to the baryon line , we would have to aggregate @xmath20 additional meson states up to @xmath21mev as compared to the present number of @xmath22 .an important question is whether the presently available range of masses is asymptotic in view of eq .( [ hag ] ) .the answer is _ no _ !this is how we can look at this question quantitatively ..various hagedorn - like fits .rows 1 - 4 use formulas of ref ., row 5 uses a simple exponent , and row 6 uses the scalar string model of ref . .the last two column display the mean suared deviation for the meson and baryon case , respectively . [ cols="^,^,^,^,^,^ " , ] consider the generic form of the spectrum of eq .( [ eee ] ) . we can rewrite it as @xmath23+(\overline{m}+\delta m)/t}= \nonumber \\ & & { \rm const}\;e^{\left ( \frac{1}{t}+\frac{f^{\prime } ( \overline{m})}{f(% \overline{m})}\right ) \deltam}={\rm const}\;e^{\frac{\delta m}{t_{{\rm eff}}% } } , \nonumber \label{asi}\end{aligned}\ ] ] where @xmath24 , and in the range of data @xmath25gev .we have defined @xmath26 as the _ effective _ hagedorn temperature in the ( non - asymptotic ) region around @xmath27 .the value of @xmath26 follows directly from the data .we have , according to eq .( [ asi ] ) , @xmath28 the following statements are obvious : * since @xmath29 , @xmath30 , * only at @xmath31 we have @xmath32 . in the region of datawe find significant differences between @xmath33 and @xmath26 .here is a numerical example .consider @xmath34 which leads to @xmath35 now we take @xmath36gev and @xmath37gev and find * for mesons : @xmath38mev , @xmath39mev ( exact fit : 195mev ) * for baryons : @xmath40mev , @xmath41mev ( exact fit : 141mev ) we conclude that only in the asymptotic region , @xmath42 , the choice of @xmath43 is not important . in the region of presently - available data @xmath11 matters very much for the extracted values of the hagedorn temperature .this simply means that we need a _ theory _ in order to make quantitative statements !the numerical parameters obtained from various choices of the function @xmath11 are collected in table [ tabhag ] .figure [ varhag ] shows the fits corresponding to the rows 1 , 4 , 5 and 6 of table [ tabhag ] .note the fits are very close to each other and the theoretical curves are virtually indistinguishable in the region of data . in view of the above discussionit makes little sense to treat the hagedorn temperature as an absolute parameter and to quote its value without specifying the model that yields the function @xmath11 . in fig .[ sns ] we show the cumulants of particle spectra of a given value of strangeness .we can clearly see that the slopes in the figure do not depend on strangeness .the meson plot includes various hagedorn fits of fig .[ varhag ] .the two sets of lines are displaced in the @xmath0 variable by roughly @xmath44mev , which is the difference of the masses on the strange and non - strange quarks .the conclusion here is that the addition of the strange quark mass has no effect on the rate of growth of the number of states with @xmath45 .certainly , we are rediscovering the @xmath46 flavor symmetry here !we end the experimental part of this talk by showing the same information as in fig .[ cumul ] , but instead of using logarithmic units on the vertical axes , we take exponential units on the horizontal axis .more precisely , we take the fit to the spectrum with of the form with the simple exponent ( row 5 in table [ tabhag ] ) , which leads to the cumulant @xmath47 , where the values of @xmath13 and @xmath33 result from the least - square fit . next , we define the variable @xmath48 and plot the cumulants as functions of @xmath49 .note that the @xmath13 and @xmath33 parameters are different for mesons and baryons .again , the linearity of data in the figure is striking .it starts at basically @xmath50 , and extends to @xmath51gev .the advantage of the plot in fig .[ expo ] to that of fig .[ cumul ] is that now the steps in the experimental cumulant are of a similar size independently of @xmath0 .we conclude this section by stating that the exponential growth of hadronic spectra in the region of @xmath0 up to about 1.8gev , with @xmath52 , is an _ experimental fact_.we are faced with two basic theoretical questions : 1 . _why is the spectrum of resonances exponential ? _ 2 ._ why do mesons and baryons behave so differently ? _concerning the first question , let us stress that it is not at all easy to get an exponentially rising spectrum of resonances .take the simplistic harmonic - oscillator model , whose density of states grows as @xmath53 , with @xmath54 denoting the number of dimensions . for mesonsthere is one relative coordinate , hence @xmath55 , whereas the two relative coordinates in the baryon give @xmath56 @xmath57 .weaker - growing potentials lead to a faster growth of the number of states , but fall short of the behavior ( [ hag ] ) .we know of three approaches yielding behavior ( [ hag ] ) , both involving combinatorics of infinitely - many degrees of freedom .these are the _ statistical bootstrap model _@xcite , bag models @xcite , and dual string models @xcite .the first two , however , lead to the same rate of growth for the mesons and baryons .statistical bootstrap models are discussed in sec .[ smmsec ] . in bag models@xcite the exponential growth of the spectrum is associated with the melting out of the vacuum around the bag when the hadron is being excited . since the scales in the bag model are practically the same for the meson and the baryon ( the size scales as the number of constituents to the power @xmath58 ) , the bag models are not capable of answering question 2 . on the other hand , the _ dual string models _@xcite is offer a natural explanation of questions 1 and 2 .this has already been pointed out in ref .@xcite . _ statistical bootstrap_ models @xcite form particles from clusters of particles , and employ the principle of self - similarity .the simplest , `` generic '' , bootstrap equation has the form @xmath59 where @xmath1 is the particle spectrum ( here , for a moment , mesons and baryons are not distinguished ) .equation ( [ boot ] ) can be nicely solved with help of laplace transforms @xcite , yielding the asymptotic solution @xmath60 , with @xmath61 .more complicated bootstrap equations involve integration over momenta , more degrees of freedom , different combinatorial factors @xcite , however , irrespectively of these details , they always lead to an exponentially growing spectrum .it can be shown , following _e.g. _ the steps of ref .@xcite , that the model leads to equal hagedorn temperatures for mesons and for baryons .this is quite obvious .since baryons are formed by attaching mesons to the `` input '' baryon , the baryon spectrum grows at exactly the same rate as the meson spectrum .specific calculations confirm this simple observation .thus the bootstrap idea _ is not capable _ of explaining the different behavior of mesons and baryons in fig .[ cumul ] .the _ dual string models _ @xcite also date back to pre - qcd times .their greatest success is a natural explanation of the regge trajectories a basic experimental fact which remain a serious problem for other approaches .similarly to the bootstrap models , the dual string models lead to exponentially - growing spectra , but they do give the demanded effect of @xmath62 , at least at asymptotic masses @xcite .let us analyze mesons first .the particle spectrum is generated by the harmonic - oscillator operator describing vibrations of the string , @xmath63 where @xmath64 labels the modes and @xmath65 labels additional degeneracy , related to the number of dimensions @xcite .eigenvalues of @xmath66 are composed in order to get the square of mass of the meson , according to the regge formula @xmath67 where @xmath68 is the regge slope , and @xmath69 is the intercept .here is an example : take @xmath70 .the value 5 can be formed by taking the @xmath71 eigenvalue of @xmath66 ( this is the leading regge trajectory , with a maximum angular momentum ) , but we can also obtain the same @xmath72 by exciting one @xmath73 and one @xmath74 mode , alternatively @xmath75 and @xmath76 modes , and so on .the number of possibilities corresponds to partitioning the number @xmath77 into natural components : 5 , 4 + 1 , 3 + 2 , 3 + 1 + 1 , 2 + 2 + 1 , 2 + 1 + 1 + 1 , 1 + 1 + 1 + 1 + 1 . herewe have 7 possibilities , but the number of partitions grows very fast with @xmath78 .partitions with more than one component describe the sub - leading regge trajectories . with @xmath79 degrees of freedomeach component can come in @xmath79 different species .let us denote the number of partitions in our problem as @xmath80 .for large @xmath78 the asymptotic formula for _ partitio numerorum _ leads to the exponential spectrum according to the formula @xcite .@xmath81 where @xmath82 .we can now read - off the mesonic hagedorn temperature : @xmath83 now the baryons : the `` mercedes - benz '' string configuration for the baryon is shown in fig .[ strcon ] .the three strings vibrate _ independently _ , and the corresponding vibration operators , @xmath66 , add up .consequently , their eigenvalues @xmath84 , @xmath85 , and @xmath86 add up .thus we simply have a partition problem with @xmath87 times more degrees of freedom than in the meson .the replacement @xmath88 in ( [ parti ] ) leads immediately to @xmath89 such that @xmath90 we stress that the presented picture is fully consistent with the regge phenomenology .the leading regge trajectory for baryons is generated by the excitation of a single string , _i.e. _ two out of three numbers @xmath91 vanish ( this is the quark - diquark configuration ) .the subleading trajectories for baryons come in a much larger degeneracy than for mesons , due to more combinatorial possibilities .the slopes of the meson and baryon trajectories are universal , and given by @xmath92 .we stress that the `` number - of - strings '' mechanism described above is asymptotic .thus , there is a problem in applying string models to the experimentally accessible range of @xmath0 .this range is not asymptotic enough to use eq .( [ parti ] ) . from the regge formula ( [ regge ] ) we find immediately that for @xmath45 in the range @xmath93gev the values of @xmath78 lie between @xmath94 and @xmath95 , hence @xmath78 is not large enough to justify the form ( [ parti ] ) .one can do better by using an improved asymptotic formula , derived in ref .the results obtained in the scalar string model @xcite are displayed in fig .[ ss ] . herethe formula for the meson spectrum is @xmath96 where @xmath97 is a modified bessel function , @xmath98 is the meson hagedorn temperature ( the _ only _ adjustable parameter here ) , and @xmath99 , with @xmath79 denoting the number of transverse dimensions . the factor of @xmath100 is just the spin - flavor degeneracy of the @xmath101 configuration @xcite . for the baryons we fold the three scalar - string densities , @xmath102 .we use @xmath103 ( rather than @xmath104 ) copies of the string , which is the degeneracy of the baryon multiplet in the ground state .we notice good agreement with data in fig .[ ss ] , for @xmath105 .note that both curves are fitted with only one parameter , @xmath98 .for lower values of @xmath79 one can fit the mesons equally well , but too many baryon states are predicted . during this workshopwe have heard many talks on hadron exotics . if an exotic is a multi - string configuration , _e.g. _ as in fig .[ exo ] , then the corresponding spectrum will grow exponentially with the hagedorn temperature inversely proportional to the square root of the number of strings .for instance , @xmath106 .this is reminiscent of the effect described in ref .for the glueballs , described by the closed string in fig .( [ exo ] ) , we get @xmath107 . thus , according to the string model ,the @xmath108 grow more rapidly than non - exotic mesons and baryons , and glueballs grow at the same rate as mesons .in the remaining part of this talk we will , in a sense , work against our results presented in previous sections , where have we argued that the plots of fig . [ cumul ] are linear , and offered an explanation of the difference between the mesonic and baryonic hagedorn temperatures within the dual string models .what if the experimental plots of fig .[ cumul ] are not really linear , and the effect of bending down of the curves at higher masses is physical , rather than due to incomplete experiments ? belowwe will show alternative descriptions which do not comply to eq .( [ hag ] ) , but nevertheless reproduce the present data at least as good as the hagedorn - like fits .in the statistical model of nuclear reactions one uses the _ compound - nucleus model _ @xcite . in this modelthe density of states grows at large excitation energies , @xmath109 , according to the formula @xmath110 where @xmath111 is a constant .formula ( [ comn ] ) can be derived within the fermi gas model @xcite .more generally , it can be derived in a model where the single - particle orbits are _ equally spaced_. one then considers @xmath112 , @xmath113 , @xmath114 , _ etc ._ , excitations and counts the number of states at a given excitation energy , @xmath109 . amusingly , this leads @xcite to the partitio numerorum formula ( [ parti ] ) , but now the number @xmath78 has the interpretation @xmath115 , with @xmath116 denoting the level spacing .we now use the following compound - hadron - model formula for the mass spectra : @xmath117 where @xmath13 is a constant , @xmath118 is the ground - state mass , and @xmath116 is the average level spacing .the constant @xmath119gev in the denominator has been introduced _ad hoc _ , similarly as in eq .( [ hr ] ) , in order for the formula to make sense at @xmath120 .asymptotically , the power of @xmath45 multiplying the exponent is @xmath121 , as in eq .( [ comn ] ) .the underlying physical picture behind compound hadrons is as follows : hadrons are bound objects of constituents ( quarks , gluons , pions ) .the fock space contains a ground state , and excitations on top of it . in the case of the compound nucleusthese elementary excitations are @xmath112 , @xmath113 , @xmath122 _ etc ._ states . in the case of hadronsthey are formed of @xmath123 and gluon excitations , _e.g. _ for mesons we have @xmath123 , @xmath124 , @xmath125 , @xmath126 , _ etc ._ we can form the excitation energy ( hadron mass ) by differently composing elementary excitations .this bring us to the above - described combinatorial problem @xcite .it seems reasonable to take zero ground - state energy for mesons , @xmath127 , since they are excitations on top of the vacuum . for baryonswe take @xmath128mev , which is the mass of the nucleon .the quantity @xmath116 is treated as a model parameter and is fitted to data . the results of the compound - hadron - model fit , eq .( [ chmrho ] ) , are shown in fig .the curves are slightly bent down , compared to the hagedorn - like fits of figs .[ cumul],[varhag ] , which is caused by the square root in the exponent of eq .( [ chmrho ] ) .but the fits are at least as good , or even better when the fit region is extended to @xmath129gev .numerically , the least - square fit for @xmath0 up to 1.8gev gives @xmath130mev for mesons , and @xmath131mev for baryons .the proximity of these numbers shows that the scales for mesons and baryons are similar , as should be the case .the obtained values for @xmath132 mean that the corresponding @xmath133 at @xmath21gev is around @xmath134 for mesons and @xmath135 for baryons .such values of @xmath78 are sufficiently large to justify the use of the asymptotic formulas .there is a possibility of an interesting effect we wish to point out .it is natural to expect that a bound hadronic system has an upper limit for the excitation energy .it is helpful to think here of bags of finite depth .thus , in constructing the single - particle fock space for bound objects we should have a limited number of quanta to our disposal . if such a limit is put into the compound hadron model , it will result in a maximum number of states that can possibly be formed out of light quarks @xcite .we can call it the _ `` light - flavor - desert hypothesis '' _ : above a certain mass there are no more light - flavor resonances . certainly , this is tangential to the conventional wisdom that the regge trajectories should continue indefinitely .note , however , that infinite regge trajectories have recently been challenged by brisudov , burakovsky and goldman , who claim that they should stop around @xmath136gev .amusingly , this is consistent with the presently - available data .the cumulants if fig .[ cumul ] flatten - out in that region .many talk in this workshop were devoted to variants of the quark model .here we present the result of counting of states in the model of refs .@xcite , as made by freund and rosner @xcite .when we look at fig .[ qm ] , we again see good agreement in the predicted and experimental number of states .this is not at all surprising , since the quark model is designed to fit the data `` state by state '' in the low - mass regime . as for other approaches , spectra at higher @xmath0would be needed to verify the predictions .there are many fundamental questions which should be cleared when more experimental data on hadron resonances are available : is the hagedorn hypothesis of exponentially - growing spectra indeed correct , or is the growth weaker at higher masses ?do the regge trajectories continue for ever , or stop ?consequently , is there a light - flavor desert above a certain mass ? are there exotic states , if so , at what rate do they grow ?... certainly , the spectrum above 2gev may reveal many answers and help us to verify various models and approaches .however , even the presently - available spectrum allows for interesting speculations .recall the remarks made here by leonid glozman , concerning the parity doublets in the @xmath66 and @xmath137 spectra above @xmath138gev @xcite .almost all states in that region can be paired , and such a regularity suggests that the data in that region may be complete ! this , in turn , indicates that the bending down of the cumulants in fig .[ cumul ] may be a physical , rather than experimental effect .another important aspect , not touched in this talk , are the thermodynamical implications of the presence of two distinct hagedorn temperatures for the phenomenology of heavy - ion collisions , transition to quark - gluon plasma , _ etc ._ this will be discussed in @xcite .the author thanks keith r. dienes for many profitable e - mail discussions on the issues of hadron spectra in string models , as well as to andrzej biaas , andrzej horzela , jan kwieciski , and kacper zalewski for numerous useful comments and encouragement .

<S> we analyze experimental particle spectra and show that the hagedorn temperature is significantly larger for mesons than for baryons . the effect can be explained within dual string models : excitations of three strings in the baryon produce `` faster '' combinatorics than a single string in the meson , hence lead to a more rapid growth of baryons than mesons . </S><S> predictions of other approaches for the gross features of particle spectra are also discussed . </S><S> psfig _ this research is being carried out in collaboration with wojciech florkowski and piotr enczykowski from inp , cracow . _ </S>

cond-mat0606300

the notion of stripes has become a standard concept in the field of high - temperature superconductivity about ten years ago , after tranquada _et al_.@xcite have observed the four - fold splitting of the antiferromagnetic ( af ) @xmath3 peak in _ elastic _neutron scattering experiments in the lanthanum family of cuprates .the above authors have interpreted the resulting four peaks , as coming from two different kinds of stripe domains , each exhibiting a unidirectional spin modulation ( referred to below as 1d ) running along one of the two principal lattice directions .the 1d spin modulation was to be accompanied by the 1d charge modulation peaked around the antiphase magnetic boundaries , hence the name stripes " .there exists , however , an alternative interpretation sometimes referred to as grid or checkerboard , which stipulates that the four - fold splitting of the @xmath3 neutron peak is caused by a two - dimensional ( 2d ) modulation of the af order shown in fig .[ fig - grid ] .such a modulation is accompanied by a 2d charge density wave localized around the antiphase boundaries , which have the appearance of grid the term to be used in the rest of this paper .( 75 , 75 ) ( 1 , 0 ) = 2.7 in grid superstructure has been mentioned in the literature since the early numerical works on the subject@xcite , however , it was only quite recently@xcite , that grid was advocated as a preferred interpretation of the @xmath3 peak splitting . it has also been suggested in ref.@xcite that similar 2d inhomogeneous structures are universally present in other families of cuprates in the normal and superconducting states .the insight of that work was not completely trivial , since its appearance predated the scanning tunneling spectroscopy ( sts ) studies of vershinin _et al_.@xcite , hanaguri _ et al_.@xcite , and the neutron study of hinkov _et al_.@xcite all three are the works , that brought the dimensionality of generic spin modulations in cuprates back into the spotlight ( see , e.g.,@xcite ) .the connection between the inhomogeneous structures in lanthanum cuprates and the generic properties of other high - temperature cuprate superconductors is a delicate issue .spin and charge modulations observed in this cuprate family may turn out to be a composition - specific feature , or , in the opposite extreme , similar modulations may be universally present in all families of cuprates but observable only in the lanthanum family due to anomalous slowing down . in this context , the difference in the dimensionality of the modulation is not inconsequential .the apparent message contained in the symmetry of the superconducting order parameter@xcite , the symmetry of the pseudogap@xcite , the sts observations on the surface of bi@xmath4sr@xmath4cacu@xmath4o@xmath5 ( bi-2212)@xcite and ca@xmath6na@xmath7cuo@xmath4cl@xmath4@xcite is two - dimensional .therefore , if the 1d nature of the spin modulations in lanthanum cuprates is established conclusively , then , in the opinion of this author , it will likely imply the composition - specific nature of the phenomenon .( the authors of refs.@xcite seem to have the opposite expectation . ) in contrast , if the spin modulations in this cuprate family are two - dimensional , then this would be yet another indication that the tendency to form 2d modulations is a general property of cuprates and , therefore , may be essential for the mechanism of superconductivity. the differences between different geometries of 2d modulations should be much less important . in lanthanum cuprates, magnetic modulations may develop into a diagonal grid pattern , while , in other cuprates , the system may phase separate into more - af - correlated and less - af - correlated nano - clusters@xcite without forming quasi - static magnetic order and , therefore , not accompanied by antiphase boundaries . since the argument for the non - universality of the modulations in lanthanum cuprates has merit , it is important to establish the dimensionality of these modulations without invoking the knowledge about other families of cuprates or nickelates .the discussion in the literature specific to lanthanum cuprates has been limited so far mostly to neutron@xcite and hard x - ray@xcite experiments .the primary focus of this paper is on extracting the dimensionality of spin modulations in la@xmath0ba@xmath1cuo@xmath2 from the nuclear quadrupole resonance ( nqr ) results of hunt _et al_.@xcite in combination with the muon spin rotation ( @xmath8sr ) results of nachumi _et al_.@xcite .the paper also elaborates on some of the dimensionality - related arguments made earlier in ref.@xcite .most of the paper will contrast 1d stripes with 2d grid , but 1d spin spirals@xcite will also be discussed in section [ discussion ] .the plan of the rest of the paper is as follows . in section [ review ] , the present non - nqr knowledge about the dimensionality of spin and charge modulations in lanthanum cuprates is discussed .section [ previous ] discusses relevant nqr findings of hunt _et al_. the nqr - related findings of the present work are presented in section [ computational ] and discussed in section [ discussion ] .sections [ review ] and [ previous ] amount [ mainly ] to an extended introductiry review .the readers familiar with the broader issues involved can skip the above sections and proceed directly to section [ computational ] .magnetic peaks corresponding to higher order modulation harmonics have not been observed so far and thus can not be used to discriminate between different modulation patterns . a direct resonant x - ray observation of charge carrier density modulations reported recently by abbamonte _et al_.@xcite has not addressed the dimensionality issue either .the direct attempts to discriminate between 1d stripe and 2d grid patterns have been made so far only on the basis of the orientation of charge satellites@xcite , which represent the response of the lattice to the periodic charge and/or spin modulations . in refs.@xcite , the analysis of experiments is based on landau expansion@xcite , which requires small spin and charge modulations .the resulting charge patterns for the 1d and the 2d cases are shown in fig .[ fig - patterns](a , b ) . in the framework of landau expansion , leading charge peaks corresponding to 1d stripe and 2d grid pictures do not coincide , and , therefore , the observation of charge satellites expected for the 1d case together with the non - observation of the satellites expected for the 2d case ( notwithstanding the weakness of the charge signal ) do present a clear argument .( 75 , 210 ) ( -5 , -5 ) = 2.6 in there are , however , good reasons to expect that the amplitude of the spin modulations is not small . for example , as indicated by @xmath8sr@xcite , the amplitude of the modulated staggered magnetization in la@xmath0ba@xmath1cuo@xmath2 is large about half of the staggered moment of the undoped parent compound. strong modulation of staggered magnetization would induce a deeper local potential for charge carriers thus forcing them to be stronger localized around the antiphase boundaries .larger amplitude of charge modulation implies stronger interaction between the charge carriers , which would undermine the theoretical basis for the landau expansion and introduce new charge peaks .one particular problem with the landau expansion is that it gives the maximum of charge density around the nodes of the antiphase grid [ see fig . [ fig - patterns](b ) ] .if the charge modulations are large , then the coulomb repulsion between the charge carriers would rather lead to a smaller charge density around the nodes . in the case of large charge modulations , the leading charge peaks corresponding to the 1d and 2d interpretations of the magnetic @xmath3 peaksplitting can coincide .if charge density is reasonably localized around the antiphase grid but avoids grid nodes , then a useful insight can be gained from analysing the density profile shown in fig .[ fig - patterns](c ) .it is given by function @xmath9 ^ 8 \\left [ \hbox{sin } \frac{\pi(-x+y)}{8 a } \right]^2 \\ & & + \ \left [ \hbox{cos } \frac{\pi(-x+y)}{8 a } \right]^8 \\left [ \hbox{sin } \frac{\pi(x+y)}{8 a}\right]^2 , \label{rho}\end{aligned}\ ] ] where @xmath10 and @xmath11 are the coordinates along two principal lattice directions and @xmath12 is the lattice period .this profile was mentioned in ref.@xcite but not elaborated upon .it corresponds to the same splitting of magnetic @xmath13 peak as the one arising from the 1d stripes spaced by @xmath14 .the cos@xmath15 factors in eq.([rho ] ) set the grid pattern , while the sin@xmath16 factors suppress the intensity near the grid nodes .the positions of the resulting charge peaks and their relative intensities are shown in fig .[ fig - peaks ] .( see the appendix for the calculation . )leading charge peaks corresponding to 1d stripes spaced by @xmath14 and to modulation ( [ rho ] ) coincide with each other and are located at @xmath17 and @xmath18 , where charge peaks were detected in refs.@xcite .modulation ( [ rho ] ) also generates peaks at @xmath19 predicted by landau expansion , but these peaks have only third largest intensity ( about 7 times smaller than the intensities of the leading peaks ) .the peaks of the second largest intensity ( 4 times smaller than that of the leading peaks ) are located at @xmath20 .it may appear that the leading peak position for the density profile ( [ rho ] ) is the result of a fine tuning of the powers of sine and cosine .this is , however , not the case .the only important aspect of this profile is that the factors sin@xmath16 suppress the charge density near the grid nodes , as required by the coulomb repulsion between charges . as illustrated in the appendix ,if the sin@xmath16 factors are removed , then the leading peaks are located at the position predicted by the landau expansion , but , if they are left , then the position of the leading peaks at @xmath17 and @xmath18 does not depend on the power of the cosine factors .the unrealistic aspect of profile ( [ rho ] ) is that the sin@xmath16 factors suppress the charge density near the grid nodes _completely_. however , even an incomplete coulomb suppression of the charge density near the grid nodes can easily shift the intensity maximum from the landau expansion result to @xmath17 and @xmath18 .( 75 , 75 ) ( 1 , 0 ) = 2.7 in from the above perspective , it is the non - observation of the weaker @xmath19 and @xmath20 peaks rather than the observation of the @xmath18 peaks that would be the real argument in favor of the 1d interpretation .however , to the best of this author s knowledge , the open literature contains only one experimental scan across the @xmath21 and @xmath22 , which has been presented in ref .@xcite for la@xmath23nd@xmath24sr@xmath25cuo@xmath2 . in that case , the experimental resolution was just not sufficient to claim observation or non - observation of the peaks , which would be 4 or 7 times weaker than the leading peak observed at @xmath26 .in fact , quite remarkably the scan around @xmath27 shows a promising sign of a peak . repeating that scan with a better resolution would certainly have merit .another argument frequently cited in support of the 1d stripe picture is that the observability of static spin and charge modulations is correlated with the onset of the `` non - generic '' low temperature tetragonal ( ltt ) crystal phase .et al_.@xcite have proposed , that the 1d stripes follow the lattice anisotropy of the ltt phase .this anisotropy alternates between adjacent cuo@xmath4 planes , which would imply , that stripes in two adjacent planes run perpendicular to each other .empirically , however , the onset of the ltt phase is not a necessary condition for the observability of static modulations , because quasielastic signatures of spin modulations have also been reported for the `` generic '' low temperature orthorhombic ( lto ) phase of la@xmath6sr@xmath7cuo@xmath2 with and without zn codoping@xcite .but one can still hope that the presence of the ltt phase is at least a sufficient condition for the observability of static modulations . although the above proposition of tranquada _et al_. has visual appeal , the onset of the ltt phase as such does not readily discriminate between the stripe and the grid interpretations .the general expectation is that the magnetic @xmath3 peak splittings observed by elastic neutron scattering in the ltt phase and by inelastic neutron scattering in the lto phase have the same origin .if true , this implies that the system has chosen the dimensionality of the modulation pattern ( 1d or 2d ) without the ltt phase being involved , and then the ltt phase only plays a role in slowing the modulations down , to make them observable in elastic experiments@xcite .therefore , to claim that the symmetry of the ltt phase discriminates between the grid and the stripe pictures is equivalent to claiming that the ltt phase would slow down fluctuating stripes , but would _ not _ slow down fluctuating grid .this author is not aware of any experimental or theoretical result , which would support the latter part of this claim .the picture of stripes running in perpendicular directions in the adjacent planes has further difficulty that such a configuration is not most favorable energetically from the viewpoint of interplane coulomb interaction .the lowest coulomb energy corresponds to stripes running in the same direction in two adjacent planes with pattern in one plane shifted with respect to the other by half a period of charge modulation . since the stripes are unscreenedlines of charge , the interplane coulomb energy should be large .can the coupling of stripes with the lattice anisotropy outweigh the coulomb coupling between stripes from adjacent planes ?it does not seem very likely .recently , robertson et al.@xcite have also argued that , due to the anisotropy of the ltt phase , the incommensurate grid modulation should have slightly different periodicity in the adjacent cuo@xmath4 planes , which should lead to 8-fold splitting of the @xmath3 peak instead of the 4-fold splitting observed experimentally . since the crystal lattice periods can not be detectably different in the adjacent planes ( because of the prohibitive elastic energy cost ) , the argument of roberson et al . is not very direct , as it relies on the additional theoretical assumption , that the mechanism similar to the fermi - surface nesting is behind the formation of the grid superstructure .this argument has further difficulty , that the mismatch of the charge superstructure periods would lead to the regions , where the maxima of charge modulation density in two adjacent planes lie on the top of each other , which , in turn , leads to a large coulomb energy cost .the critical discussion so far does not change the fact that the observed positions of the charge peaks , and the involvement of the lto - ltt transition can not be turned into an argument favoring the grid over the stripe pictures , and to this extent , these two observations do support the 1d stripe interpretation . in the present case , however , the the presumption of innocence can be rephrased as follows : a two - dimensional experimental peak pattern should be considered as an evidence for two - dimensional superstructure unless conclusively proven otherwise . in the opinion of this author ,the above cited facts fall far short of delivering a conclusive proof .one can , however , start from the presumption of guilt " by considering the 2d grid interpretation of the 2d neutron data as exotic " , and requiring compelling reasons to choose it over the 1d stripe picture .it is the purpose of the present paper to analyse the nqr experiments from this perspective , and to point out that , within the limits of experimental uncertainties , they do constitute a clear indication in favor of the 2d grid interpretation .a slowly fluctuating spin modulation , 1d or 2d , is expected to freeze at sufficiently low temperatures thus leading to a broad nqr spectrum having shape , which may discriminate between different dimensionalities of the modulation . in ref.[1 ] , hunt _ et al_. have presented , among other findings , the results of their nqr study of la@xmath0ba@xmath1cuo@xmath2 aimed at testing the stripe interpretation of neutron scattering experiments . at temperatures below 8k , they have , indeed observed a broad nqr spectrum of @xmath28cu [ shown in fig . [ fig - hunt](f ) ] , which signified the expected freezing of local hyperfine fields affecting copper nuclear spins .this study was done in the context of a broader effort to elucidate the dynamics of electronic spins in lanthanum cuprates by the means of nuclear magnetic and quadrupole resonances@xcite ( 75 , 140 ) ( -5 , 0 ) = 3.3 in the above nqr spectrum has exhibited one clear feature , the 31 mhz peak , but an important theoretical finding of hunt _et al_. was that this particular peak should be present for any continuous distribution of hyperfine fields passing the interval 0.5 - 1.5 t , where the minima of two nqr frequency branches are located - both corresponding to frequency 31 mhz ( see fig .[ fig - freq ] ) .therefore , the presence of the 31 mhz peak could not discriminate between different kinds of modulations .apart from this peak , the spectrum has exhibited no other structure .( 75 , 60 ) (-8 , 0 ) = 3.4 in hunt _ et al_. have attempted to fit the nqr spectrum to the predictions of many possible 1d _ static _ stripe pictures , but have found that all such pictures lead to very structured lineshapes having several peaks not observed experimentally [ see figs . [ fig - hunt](a - d ) ] .the origin of the structure in theoretical nqr lineshapes is two - fold .( i ) for commensurate superstructures , there is only a finite number of distinct nuclear sites .the traces of this discreetness then appear in the nqr lineshape .( ii ) for uniform incommensurate 1d modulations , which contain the infinite number of distinct sites , the peaks in the nqr lineshapes result from a van hove singularity in the distribution of 1d - modulated hyperfine fields .this singularity is located at the maximum in the absolute value of the hyperfine fields [ see the middle panel of fig [ fig - hunt](d ) ] . a priori , a 2d grid modulation represents a promising alternative to interpret the rather featureless nqr spectrum of hunt _et al_. , because , when the grid is commensurate , it contains a larger number of inequivalent nuclear sites , and thus tends to produce more smeared structure in nqr lineshape , and , when the grid is incommensurate , it would _ not _ lead to a van hove singularity at the maximum value of 2d modulated hyperfine fields .( it would lead to a singularity around the zero value of hyperfine field , but that singularity is less important .see the next section . )hunt _ et al_. have considered the possibility of grid modulation , but only the commensurate version with the orientation rotated by 45 degrees in comparison to the one expected from the positions of neutron peaks . in this way, they have , nevertheless , obtained a much better agreement with experiment in comparison with any of the 1d interpretations [ see fig . [ fig - hunt](e ) ] , which is not very surprising in view of the preceeding discussion . yet , the grid pattern used by hunt _et al_. generated too much structure , which led them to abandon the grid interpretation all together and to test another one in terms of the random distribution of hypefine fields .the latter interpretation gave nearly perfect agreement with the experimental nqr spectrum [ fig .[ fig - hunt](f ) ] .the problem with that interpretation is that the random configuration of staggered spin polarizations is inconsistent with the modulations of af spin structure observed by neutron scattering@xcite .random picture should also lead to a monotonic decay of the @xmath8sr asymmetry signal rather than the oscillations observed in experiment@xcite .the frequency of the @xmath8sr asymmetry oscillations implies that the modulation amplitude of the staggered polarization is about @xmath29@xcite , where @xmath30 is bohr magneton .this number imposes a difficult constraint on the interpretations of nqr spectrum .et al_. , have concluded that any kind of static broadening of the hyperfine field distribution corresponding to the modulation amplitude @xmath29 would fail to reproduce their nqr spectrum , because it would introduce too much spectral weight in the higher frequency interval , where much smaller weight is observed experimentally .they have found , however , that with the modulation amplitude @xmath31 and a very significant degree of broadening ( 50 - 100 per cent of the modulation amplitude ) , the observed nqr spectra can be reproduced .hunt _ et al_. have suggested that the above discrepancy may be due to spin fluctuations on the nqr timescale of 20 @xmath8s , which are not seen on the faster @xmath8sr timescale of 0.5@xmath8s .this suggestion , however , runs into the difficulty , that , in order to change the shape of the effective distribution of hyperfine fields , spins should fluctuate on the time scale of inverse nqr frequency ( @xmath32s ) or faster .( this , presumably , happens at temperatures higher than 30 k. ) the fluctuations on the scale of 20 @xmath8s can lead to the loss of resonant spins , which , in turn , can reduce the integrated signal intensity but should not change the shape of the spectral line .this is consistent with the fact , that the nqr spectra of hunt _ et al_. have identical shape but different intensity at 1.7 k and 350 mk .even if , one insists that the spin fluctuations may affect the observed nqr lineshape despite the above argument , one would have to make a further assumption that these fluctuations remain temperature - independent as the temperature is reduced by a factor of 5 from 1.7 k to 350 k , which is quite unlikely .the present work has been motivated by the observation that hunt _ et al_. did not consider incommensurate grid modulation , which has promise of describing their experiment without difficult assumptions .as mentioned in the previous section , hunt _ et al_. have considered grid modulation , which had different orientation , in comparison to the one that follows from the neutron experiments .their grid was oriented along the principal lattice directions , whereas the geometry of the neutron @xmath3 peak splitting implies the diagonal grid orientation@xcite given by the modulation function [ leading harmonic ] @xmath33 \hbox{sin}\left [ { \pi \delta \over a } ( x - y ) \right ] \label{f}\ ] ] and shown in fig.[fig - grid ] . here@xmath34 is the dimensionless measure of the splitting of the @xmath3 peak .the four peaks are located at @xmath35 $ ] and @xmath36 $ ] .the experiments indicate@xcite that @xmath37 , which corresponds to the distance @xmath38 between two nearest grid nodes . in this section ,three cases are computed theoretically and compared with experiment : ( a ) commensurate grid corresponding to @xmath39 ; ( b ) uniform incommensurate grid corresponding to @xmath40 , where @xmath41 is an irrational number much smaller than @xmath42 ; and ( c ) disordered incommensurate grid with @xmath37 . below , unless explicitly indicated otherwise , the same computational routine and the same values of parameters as those of hunt _et al_.@xcite have been used .when af spin structure is modulated by some function @xmath43 , the hyperfine field at position @xmath44 is given by @xmath45 \right\ } , \nonumber\end{aligned}\ ] ] where@xcite @xmath46 and @xmath47 are the hyperfine coupling coefficients and @xmath48 is the maximum value of staggered polarization . substituting @xmath43 given by eq.([f ] ) , one can obtain explicitly that @xmath49 \hbox{sin}\left [ { \pi \delta \over a } ( x - y ) \right ] , \label{hxy1}\ ] ] where @xmath50 . \label{hm}\ ] ] it is assumed that the hyperfine magnetic fields are parallel to the cuo@xmath4 planes@xcite , and that the quadrupolar part of the hamiltonian is axially symmetric with respect to the crystal axis perpendicular to the cuo@xmath4 planes@xcite ( @xmath51-axis ) . with these assumptions , the full hamiltonian for one @xmath28cu nuclear spin has form @xmath52 where @xmath53 is the plank constant , @xmath54 is the spin 3/2 operator , @xmath55 is the hyperfine field , @xmath56 is the giromagnetic ratio of @xmath28cu nuclei ( equal to 11.285 mhz / tesla ) , and @xmath57 is the quadrupolar parameter having average values@xcite 34.5 mhz and 41.8 mhz for for the a- and b - lines of the @xmath28cu nqr spectra .the relative spectral weight of the b - line with respect to the a - line was extracted from the high - temperature nqr spectra of hunt _et al_.@xcite to be 0.07 . in the calculation , the gaussian distribution of @xmath57 around the above mentioned average values was assumed with root - mean - squared deviation @xmath58mhz@xcite . for a given value of quadrupolar parameter @xmath57 and the hyperfine field @xmath59 , hamiltonian ( [ h ] ) generates four levels and six transition frequencies between those levels ( shown in fig .[ fig - freq ] ) .the contribution of each transition to the nqr spectrum was weighed by the square of the transition frequency , and then by the isotropic average of the transition probability @xmath60 where @xmath61 is the radio frequency field ( @xmath62100 g)(see though the discussion of this factor in the next section . ) when the grid structure is commensurate , it generates only a finite discrete set of different hyperfine fields @xmath63 .this discrete distribution was broadened to a set of gaussians exp@xmath64 with @xmath65@xcite .the commensurate diagonal grid modulation was important to test , because it implies a larger number of inequivalent nuclear sites , in comparison to the commensurate grid considered by hunt _et al_. however , as shown in fig .[ fig - main](a ) , theoretical nqr spectrum corresponding to the diagonal commensurate grid modulation still has too much structure in disagreement with experiment .( 75 , 105 ) ( -44 , 0 ) = 6.5 in in the case of _ incommensurate _ periodic grid modulation not considered by hunt _ et al_. , the distribution of local fields was obtained in the following way : grid superstructure is incommensurate , when the period of modulation along each of the two lattice directions is an irrational number [ in the units of lattice period ] . in this case , the two modulation phases are ergodic on the lattice , which means that , when taken modulo @xmath66 , both of them uniformly fill interval @xmath67 .this implies that the distribution of the hyperfine fields can be computed from the modulation function ( [ hxy1 ] ) by assuming that spins uniformly fill the space instead of being attached to a discrete lattice .the resulting distribution function can be found analytically .for @xmath68 , it has form @xmath69 where function @xmath70 is the elliptic integral of the first kind : @xmath71 here @xmath72 and @xmath73 are some function arguments , and @xmath74 is an integration variable . for @xmath75 , @xmath76 . in order to generate a specific distribution @xmath77 ,@xmath78 was substituted into eq.([hm ] ) , which gives @xmath79 t. strictly speaking @xmath39 corresponds to a commensurate modulation but the value of the small irrational number , which should be subtracted from @xmath42 makes very little difference , once the incommensurate modulation sets in and formula ( [ ph ] ) becomes applicable .the shape of @xmath77 is shown in fig .[ fig - main](b ) .it has a van hove singularity at @xmath80 , which is associated with the fact that @xmath81 given by eq.([hxy1 ] ) has a saddle point at @xmath82 , @xmath83 .this singularity , however , does not lead to a pronounced feature in the nqr spectrum , because , at small values of @xmath59 , the transition frequencies are dominated by the value of @xmath57 , and , therefore , the nqr spectrum is shaped by the non - singular distribution of @xmath57 .in contrast , the step feature of @xmath77 at @xmath84 falls in the range , where the value of h begins controlling the values of transition frequencies .this leads to the [ rounded ] steps in the nqr spectrum at frequencies corresponding to @xmath84 ( see fig .[ fig - freq ] ) .overall , as shown in fig .[ fig - main](b ) , the incommensurate grid modulation gives a better agreement with experiment in comparison with the commensurate one , but still with features in the form of steps absent in experiment .( a closer examination of experimental spectrum , in fact , reveals the hints of steps in the positions expected theoretically . ) in cuprates , there is no compelling reason to expect a perfectly uniform incommensurate spin density wave ( 1d or 2d ) .instead , one should rather expect incommensurate periodicity only on average with local pattern distorted either by disorder or by the mcmillan type of discommensurations in the commensurate superstructure@xcite . a straightforward way to incorporate both effectsis to broaden the hyperfine field distribution obtained for the perfect incommensurate modulation . as shown in fig .[ fig - main](c ) , only a minor degree of broadening 10 per cent of the original distribution width is sufficient to eliminate the essential disagreement between the theory and the experiment .the remaining disagreement should be well within the uncertainty of the theoretical scheme and the experiment ( see the discussed in the next section ) . the distribution of hyperfine fields in fig .[ fig - main](c ) was obtained by simple convolution of the distribution shown in fig .[ fig - main](b ) with a gaussian exp@xmath85 , where @xmath86 .this uniform broadening was chosen over the one proportional to @xmath59 [ as done in ref.@xcite and for fig .[ fig - main](a ) ] just for the sake of simplicity . in general, one should expect @xmath87 to have both @xmath59-independent and @xmath59-dependent terms .however , for the present purpose , the number of significance is @xmath88 , while other differences in the choice of @xmath89 lead only to minute differences in the nqr spectra .ten per cent broadening is a very reasonable value theoretically .it is expected , for example , in the case , where a grid line fluctuates between two adjacent diagonals on the lattice , which leads to the fluctuations in the size of the grid supercell by @xmath90 . given the approximate size of the supercell @xmath38 , the above fluctuation amounts to 12.5 per cent fluctuation in the size of the supercell , which , in turn , can easily lead to a 10 per cent fluctuation in the value of @xmath91 within the supercell .importantly , the amplitude of spin modulation used for the present calculation ( @xmath92 ) is consistent with the frequency of the @xmath8sr asymmetry oscillations@xcite .one can further observe that the shape of the distribution of hyperfine fields seen by nuclear spins should not be too different from the shape of the distribution of magnetic fields experienced by muons .( simulations in ref.@xcite indeed suggest so away from the close vicinity to the zero value of the field . ) therefore , by fourier transforming the distribution of the hyperfine fields , one can obtain a rough estimate of the amplitude of the @xmath8sr asymmetry oscillations .such a fourier transform is shown in the inset of the right panel in fig .[ fig - main](c ) .it exhibits clear oscillations with the amplitude roughly consistent with but smaller than the @xmath8sr observations .( the frequency of @xmath8sr asymmetry oscillations should , of course be rescaled with the strength of muon - electron magnetic coupling . )it should be mentioned though , that , in the absense of nqr data , the the smallness of the oscillation amplitude was considered by kojima _@xcite as an argument against the grid interpretation . to summarize , none of the propositions considered by huntet al_.@xcite came close to the kind of simultaneous agreement with the nqr and @xmath8sr results obtained in this section for the disordered incommensurate grid modulation .the significance of the findings presented in the previous section should be viewed in the context of uncertainties of the theoretical framework , within which these findings were obtained .an important quantitative check of that framework is the fact , that it correctly predicts the experimental position of the 31 mhz peak .it is the check of the correctness of hamiltonian ( [ h ] ) .more specifically , since the value of @xmath57 is known from the high - temperature measurements , the above agreement amounts to a combined consistency check of the assumptions that the hamiltonian is axially symmetric and that local hyperfine fields are parallel to the cuo@xmath4 planes .one uncertainty of the treatment in section [ computational ] is associated with the assumption of completely random distribution of a- and b - type nuclei in the inhomogeneous background .one can object , that charged grid lines are attracted to dopant atoms , and hence the nuclei of b - type ( attributed to the vicinity of dopant atoms ) see quite different environment than a - type nuclei@xcite .it may happen , for example , that small amplitude fluctuations of grid lines wipe out the signal from b - nuclei , while leaving a significant signal coming from a - nuclei .it turns out , however , that , if only a - type ( majority ) nuclei are included in the calculation for the disorder broadened incommensurate grid then the agreement between the theory and the experiment becomes even better than the one shown in fig .[ fig - main](c ) .a related difficulty is that the charge modulation , as it becomes stabilized , creates extra electric field gradient , which can lead to an extra broadening of the distribution of @xmath57 .a priori , the relative value of this broadening in comparison with the one due to the intrinsic factors ( such as charged dopants or crystal imperfections ) is difficult to estimate .however , it is also difficult to see , why the modulation - induced broadening would be much larger than the `` intrinsic '' one , known from the width of the high temperature nqr spectra . if it is not much larger , then the related uncertainty is not very significant .another important element missing from the description in section [ computational ] is what happens in the system as a result of the @xmath93 and @xmath94 radio - frequency ( rf ) pulses of the spin echo sequence .the computational procedure of section [ computational ] assumes that the @xmath93 and @xmath94 pulse conditions are perfectly satisfied for every transition of every spin . in reality , this can not be done simultaneously for all spins and all transitons , because ( i ) in powder samples ( used in ref.@xcite ) , the polarization of the rf field has a random orientation with respect to the lattice directions ; and ( ii ) even , if it were a single crystal , the rf field at a given frequency affects simultaneously several nqr transitions and couples to those transitions through different values of matrix elements ( which typically vary by a factor of 2 ) .the difference in the values of matrix elements implies that the pulse conditions can not be fulfilled simultaneously for every transition . even , if one were able to manipulate each transition individually , the use of weighing factor ( [ pnn ] ) would be doubtful .such a factor is standard and appropriate , when the resonance is detected by the steady state absorption technique . in the pulsed experiments ,however , one power of the matrix element @xmath95 is absorbed into the duration of the rf pulses , which means that , if the @xmath93 and @xmath94 conditions are perfectly satisfied , one should expect a weighing factor equal to the above matrix element and not to its square as eq.([pnn ] ) suggests .when the pulse conditions are not perfectly satisfied , higher powers of that matrix element may contribute to the weighing factor , but even then , the use of the factor ( [ pnn ] ) would likely amount to a drastic oversimplification . expression ( [ pnn ] ) was used in section [ computational ] only to make the results directly comparable to the calculations of hunt _et al_. yet another uncertainty of this study is the lack of theory describing the wipeout of the nqr signals on the timescale of rf pulses . in experiment of ref .@xcite , the duration of the rf pulses was chosen to be the one that maximizes the spin echo response .the wipeout effect would suppress the intensity of the echo , if the @xmath93 and @xmath94 conditions required too long pulses , and , therefore , could factor into the experimental definition of the pulse conditions .all unknown factors mentioned above can add up to a quantitatively significant weighing factor on the top of the theoretical calculation described in section .[ computational ] .this extra factor , however , is not expected to have strong frequency dependence and , therefore , would not mask the presence of the multiple peaks associated either with discrete or with singular distributions arising from 1d modulations of hyperfine fields .it is , therefore , the absence of the extra peaks beyond the one at 31 mhz that is the best indication against 1d stripe scenarios .the non - observation of multiple nqr peaks also constitutes an indication against spin spirals@xcite in 1/8-doped lanthanum cuprates .if the cu spins are indeed parallel to the cuo@xmath4 plane , and if the nqr tensor is axially symmetric with respect to the @xmath51-axis , then the spiral arrangement of spins implies that all nuclear sites are equivalent , which should lead to sharp multiple nqr lines .if any of the above two assumptions is incorrect then nqr spectrum should still exhibit van hove singularities similar to those shown in fig.[fig - hunt](d ) .the main message of this paper is that the absence of multiple peaks in the nqr spectra of hunt et al . can be easily explained on the basis of slightly disordered incommensurate grid modulation .the degree of disorder required for this interpretation entails approximately 10 per cent broadening of the non - disordered nqr line , which would be consistent with the fact that the signatures of spin modulations are observable by @xmath8sr . at the same time , all alternative interpretations of the above spectra in terms of 1d spin modulations encounter serious difficulties .it is , therefore , concluded that the combined @xmath8sr / nqr data favor the the 2d grid modulation over the 1d stripe or spiral modulation static or dynamic ._ note added : _ after the present manuscript was submitted , this author became aware of the preprint by christensen _et al._@xcite , reporting significant new experimental results of magnetic neutron scattering in la@xmath23nd@xmath24sr@xmath25cuo@xmath2 .these results were interpreted by the above authors as strongly supporting to the 1d stipe interpretation .the present author has expressed his reservations on this matter in ref.@xcite .the author is grateful to c.p.slichter and j. haase for drawing his attention to the nqr experiment of hunt _et al_. , to t. imai for the detailed discussion of that experiment , and to t. egami for discussions of this work .\ \text{cos}\left[\frac{\pi y}{2}\right ] + 28 \ \text{cos}\left[\frac{\pi x}{2}+\frac{\pi y}{2}\right ] + 28 \ \text{cos}\left[\frac{\pi x}{2}-\frac{\pi y}{2}\right ] \nonumber \\ & & + 21 \ \text{cos}\left[\frac{\pi x}{4}+\frac{\pi y}{4}\right ] + 21\ \text{cos}\left[\frac{\pi x}{4}-\frac{\pi y}{4}\right ] \nonumber \\ & & -14 \ \text{cos}\left[\frac{3 \pi x}{4}+\frac{\pi y}{4}\right ] -14 \ \text{cos}\left[\frac{\pi x}{4}+\frac{3 \pi y}{4}\right ] -14 \ \text{cos}\left[\frac{\pi x}{4}-\frac{3 \pi y}{4}\right ] -14 \ \text{cos}\left[\frac{3 \pi x}{4}-\frac{\pi y}{4}\right ] \nonumber \\ & & + \left ( 8 \ \text{cos}\left[\frac{3 \pi x}{4}+\frac{3 \pi y}{4}\right ] -\text{0.5$\grave { } $ } \\text{cos}\left[\frac{3 \pi x}{4}-\frac{5\pi y}{4}\right ] -\text{0.5$\grave { } $ } \\text{cos}\left[\frac{5 \pi x}{4}-\frac{3 \pi y}{4}\right ] \right ) \nonumber \\ & & + \left ( 8 \ \text{cos}\left[\frac{3\pi x}{4}-\frac{3 \pi y}{4}\right ] -\text{0.5$\grave{}$ } \ \text{cos}\left[\frac{5 \pi x}{4}+\frac{3 \pi y}{4}\right ] -\text{0.5$\grave { } $ } \ \text{cos}\left[\frac{3 \pi x}{4}+\frac{5 \pi y}{4}\right ] \right ) \nonumber \\ & & -4 \\text{cos}\left[\frac{\pi x}{2}+\pi y\right ] -4 \ \text{cos}\left[\pi x + \frac{\pi y}{2}\right ] -4 \ \text{cos}\left[\frac{\pi x}{2}-\pi y\right ] -4 \ \text{cos}\left[\pi x - \frac{\pi y}{2}\right ] \nonumber \\ & & + \text{cos}\left[\pi x + \pi y\right ] + \text{cos}\left[\pi x - \pi y\right ] \hbox{{\huge \ } } } \label{expansion}\end{aligned}\ ] ] the terms grouped in parentheses ( ... ) give the same peak positions in the first brillouin zone .the relative peak intensities exhibited in fig .[ fig - peaks ] are the squares of the fourier amplitudes appearing in the above expansion .the following examples illustrate that the leading character of the peak at @xmath98 and @xmath99 depends only on the presence of the sin@xmath16 factors , which suppress the charge density near the grid nodes .@xmath100 ^ 2 + \left [ \hbox{cos } \frac{\pi(-x+y)}{8 } \right]^2 \ = { 1 \over 2 } \left\{2 + \text{cos}\left[\frac{\pi x}{4}+\frac{\pi y}{4}\right ] + \text{cos}\left[\frac{\pi x}{4}-\frac{\pi y}{4}\right ] \right\ } \label{exp1}\ ] ] @xmath101 ^ 2 \\left [ \hbox{sin } \frac{\pi(-x+y)}{8 } \right]^2 + \ \left [ \hbox{cos } \frac{\pi(-x+y)}{8 } \right]^2 \ \left [ \hbox{sin } \frac{\pi(x+y)}{8}\right]^2 \\ & = & { 1 \over 4 } \left\ { 2 - \text{cos}\left[\frac{\pi x}{2}\right ] - \text{cos}\left[\frac{\pi y}{2}\right ] \right\ } \label{exp2}\end{aligned}\ ] ] @xmath101 ^ 4 \ \left [ \hbox{sin } \frac{\pi(-x+y)}{8 } \right]^2 + \ \left [ \hbox{cos } \frac{\pi(-x+y)}{8 } \right]^4 \ \left [ \hbox{sin } \frac{\pi(x+y)}{8}\right]^2 \\ & = { 1 \over 32 } \text{{\huge \ { } } & 12 - 8 \ \text{cos}\left[\frac{\pi x}{2}\right ] - 8 \ \text{cos}\left[\frac{\pi y}{2}\right ] + 2 \ \text{cos}\left[\frac{\pi x}{2}+\frac{\pi y}{2}\right ] + 2 \ \text{cos}\left[\frac{\pi x}{2}-\frac{\pi y}{2}\right ] \nonumber \\ & & + 2 \ \text{cos}\left[\frac{\pi x}{4}+\frac{\pi y}{4}\right ] + 2\ \text{cos}\left[\frac{\pi x}{4}-\frac{\pi y}{4}\right ] \nonumber \\ & & - \ \text{cos}\left[\frac{3 \pi x}{4}+\frac{\pi y}{4}\right ] - \ \text{cos}\left[\frac{\pi x}{4}+\frac{3 \pi y}{4}\right ] - \ \text{cos}\left[\frac{\pi x}{4}-\frac{3 \pi y}{4}\right ] - \ \text{cos}\left[\frac{3 \pi x}{4}-\frac{\pi y}{4}\right ] \hbox{{\huge \ } } } \label{exp3}\end{aligned}\ ] ]

<S> this paper reanalyzes the low temperature nuclear quadrupole resonance ( nqr ) experiments in la@xmath0ba@xmath1cuo@xmath2 by hunt _ et al_. [ phys . </S><S> rev . </S><S> b 64 , 134565 ( 2001 ) ] in an attempt to determine the dimensionality of spin modulations in this and other compounds of the lanthanum family of high temperature cuprate superconductors . </S><S> it is concluded that the shape of the nqr spectra obtained by hunt _ </S><S> et al_. favors the two - dimensional pattern of spin modulations known as `` grid '' or `` checkerboard '' . </S><S> the paper also contains the discussion of charge patterns , which can accompany the above spin modulation . </S>

math0405320

many stimulating problems in mathematics owe their existence to variational formulations of physical phenomena . in differential geometry , harmonic maps , candidate minimisers of the dirichlet energy , can be described as constraining a rubber sheet to fit on a marble manifold in a position of elastica equilibrium , i.e. without tension @xcite .however , when this scheme falls through , and it can , as corroborated by the case of the two - torus and the two - sphere @xcite , a best map will minimise this failure , measured by the total tension , called bienergy . in the more geometrically meaningful context of immersions , the fact that the tension field is normal to the image submanifold , suggests that the most effective deformations must be sought in the normal direction .then , two approaches to this optimization are available : the first one ( the free state ) consists in finding ( normal ) extrema of the bienergy , with complete disregard for the behaviour of the energy . in the second , in order to avoid paying too great a price for a smaller tension , a constancy condition on the energy level is imposed . in more intuitive terms , and even though we never consider the associated flows , these points of view correspond , at least in the more favorable situations , to reducing the overall tension of a surface , with or without controlling the mean curvature .however different , both outlooks are unified into a single mathematical description , amounting to a lagrange multiplier interpretation. this considerations lead to the following definitions .a map @xmath0 between riemannian manifolds is called _ biharmonic _ if it is a critical point , for all variations , of the bienergy functional : @xmath1 where @xmath2 is the tension field , vanishing for critical points of the dirichlet energy ( i.e. harmonic maps ) : @xmath3 in case of non - compact domain , these two definitions should be understood as for all compact subsets .the euler - lagrange operator attached to biharmonicity , called the _ bitension field _ and computed by jiang in @xcite , is : @xmath4 and vanishes if and only if the map @xmath5 is biharmonic .we are now ready to define the main object of this paper .an immersion @xmath6 ( @xmath7 ) between riemannian manifolds , or its image , is called _ biminimal _ if it is a critical point of the bienergy functional @xmath8 for variations normal to the image @xmath9 , with fixed energy .equivalently , there exists a constant @xmath10 such that @xmath5 is a critical point of the @xmath11-bienergy @xmath12 for any smooth variation of the map @xmath13-\epsilon , + \epsilon [ \times m \to n$ ] , @xmath14 , such that @xmath15 is normal to @xmath16 .the functional @xmath17 has been on the mathematical scene since the early seventies ( see @xcite ) , where its critical points , for all possible variations , are studied . in particular , it is shown to satisfy condition ( c ) when the domain has dimension two or three and the target is non - positively curved , ensuring the existence of minimisers in each homotopy class. however , l. lemaire , in @xcite , constructs counter - examples when no condition is imposed on the curvature . using the euler - lagrange equations for harmonic and biharmonic maps , we see that an immersion is biminimal if @xmath18^{\perp}=[\tau_{2}]^{\perp}- \lambda [ \tau]^{\perp}=0,\ ] ] for some value of @xmath19 , where @xmath20^{\perp}$ ] denotes the normal component of @xmath20 $ ] .we call an immersion _ free _ biminimal if it is biminimal for @xmath21 . in the instance of an isometric immersion @xmath22, the biminimal condition is @xmath23^{\perp}+\lambda { \mathbf h } = 0.\ ] ] note that this variational principle is close to the willmore problem , the disparity being that we do not vary through isometric immersions .while it is obvious that biharmonic immersions are biminimal , we will see in the following sections that the two notions are well distinct .for example , we construct families of biminimal surfaces in three - dimensional space forms of non - positive constant sectional curvature where biharmonic surfaces do not exist @xcite . in the same vein, we construct families of biminimal surfaces in almost all three - dimensional geometries of thurston . _notation_. we shall place ourselves in the @xmath24 category , i.e. manifolds , metrics , connections , maps will be assumed to be smooth .by @xmath25 we shall mean a connected manifold , of dimension @xmath26 , without boundary , endowed with a riemannian metric @xmath27 .we shall denote by @xmath28 the levi - civita connection on @xmath29 . for vector fields@xmath30 on @xmath31 we define the riemann curvature operator by @xmath32}z-[\nabla_{x},\nabla_{y}]z$ ] . for the laplacian we shall use @xmath33 for functions @xmath34 and @xmath35 for sections along a map @xmath36 .our quest for examples of biminimal immersions starts with curves .let @xmath37 be a curve parametrised by arc - length in a riemannian manifold @xmath38 , that is @xmath39 is an isometric immersion . before computing the bitension field of @xmath39, we recall the definition of frenet frames .[ def2.1 ] the frenet frame @xmath40 associated to a curve @xmath37 is the orthonormalisation of the @xmath41-uple @xmath42 , described by : @xmath43 where the functions @xmath44 are called the curvatures of @xmath39 .note that @xmath45 is the unit tangent vector field to the curve . in the instance of a curve @xmath39 on a surface ( @xmath46 ), the frenet frame reduces to the couple @xmath47 , @xmath48 being the unit tangent vector field along @xmath39 and @xmath49 a normal vector field along @xmath39 such that @xmath47 is a positive basis , while @xmath50 is the signed curvature of @xmath39 .biminimal curves in a riemannian manifold are characterised by : [ prop1 ] let @xmath51 ( @xmath52 ) be an isometric curve from an open interval of @xmath53 into a riemannian manifold @xmath29. then @xmath39 is biminimal if and only if there exists a real number @xmath11 such that : @xmath54 where @xmath55 is the curvature tensor of @xmath29 and @xmath40 the frenet frame of @xmath39 . with respect to its frenet frame ,the tension field of @xmath39 is : @xmath56 and its bitension field : @xmath57 the vanishing of the normal components yields system .asking for a free biharmonic curve @xmath39 to be biharmonic requires the supplementary condition @xmath58^{b_{1 } } = 0 $ ] , equivalent to @xmath59 , that is , either @xmath60 is constant or @xmath39 is a geodesic ( @xmath61 ). if the target manifold is a surface or a three - dimensional riemannian manifold with constant sectional curvature , equations are more manageable as shown in the following : [ cor2.1 ] + * an isometric curve @xmath39 on a surface of gaussian curvature @xmath62 is biminimal if and only if its signed curvature @xmath63 satisfies the ordinary differential equation : @xmath64 for some @xmath10 . * an isometric curve @xmath39 on a riemannian three - manifold of constant sectional curvature @xmath65 is biminimal if and only if its curvature @xmath63 and torsion @xmath66 fulfill the system : @xmath67 for some @xmath10 . for i ) .the two - dimensional frenet frame of @xmath39 consists only of @xmath48 and @xmath49 , and the curve is biminimal , with respect to @xmath11 , if and only if : @xmath68 but since @xmath69 , we obtain . in dimension three ,the frenet frame of @xmath39 is @xmath70 , and the conditions of proposition [ prop1 ] become : @xmath71 the constant sectional curvature of the target means that @xmath72 and @xmath73 .> from corollary [ cor2.1 ] , if @xmath39 is an isometric curve in a riemannian manifold @xmath74 of constant sectional curvature @xmath65 and dimension @xmath75 or @xmath76 , then the curvature of @xmath39 ( the signed curvature when @xmath77 ) , satisfies the equation : @xmath78 where @xmath79 and @xmath80 .multiplying equation by @xmath81 and integrating , we obtain : @xmath82 and setting @xmath83 yields : @xmath84 since this equation is of the form @xmath85 , @xmath86 being a polynomial of degree three , it can be solved by standard techniques in terms of elliptic functions ( see , for example @xcite ) . in a forthcoming paperwe shall give an accurate description of the solutions of equation .here we just point out that if @xmath31 is the flat @xmath87 , then equation , for free biminimal curves , reduces to @xmath88 a solution of which can be expressed in terms of elementary functions , that is @xmath89 , where @xmath90 is the arc - length . now using the standard formula to integrate a curve ofknown signed curvature , we find that , up to isometries of @xmath87 , this free biminimal curve is given by @xmath91 this is the standard parametrisation by arc - length of the logarithmic spiral plotted in figure [ figure1 ] .= 1.7 in on a riemannian manifold @xmath29 , any representative of the conformal class @xmath92 $ ] can be expressed as @xmath93 , @xmath94 , and the levi - civita connections are related , for @xmath95 , by ( cf .@xcite ) : @xmath96 observe that a geodesic @xmath39 on @xmath29 will not remain geodesic after a conformal change of metric , unless the conformal factor is constant since : @xmath97 let @xmath38 be a riemannian manifold .fix a point @xmath98 and choose a function @xmath99 depending only on the geodesic distance from @xmath100 .then any geodesic on @xmath29 going through @xmath100 will be a free biminimal curve on @xmath101 .let @xmath39 be a geodesic of @xmath38 and let @xmath40 be the associated frenet frame ( cf .definition [ def2.1 ] ) . since the function @xmath99 depends only on the geodesic distance from @xmath100 , i.e. the @xmath102-direction , @xmath103 .since @xmath39 is a geodesic on @xmath29 , then : @xmath104 and the tension field of @xmath39 with respect to the metric @xmath105 is : @xmath106 where in this last equality we have used .besides : @xmath107 thus @xmath108 .still with respect to @xmath109 , the bitension field of @xmath39 is : @xmath110 b_{1}. \end{aligned}\ ] ] so @xmath111 has no normal component and @xmath39 is free biminimal on @xmath112 .let @xmath113 be the geodesic distance from a point @xmath114 , and @xmath115 .then a geodesic on @xmath29 through @xmath100 becomes a biharmonic map on @xmath116 .> from the proof of theorem [ thm1 ] , a geodesic on @xmath29 through @xmath100 becomes a biharmonic map on @xmath117 if @xmath99 is a solution of the ordinary differential equation : @xmath118 to solve this equation , put @xmath119 to obtain @xmath120 then , using the transformation @xmath121 , equation becomes @xmath122 which has the solution @xmath123 . finally , from @xmath124 , we find the desired @xmath99 . as an example, one can take @xmath125 , and @xmath126 , where @xmath127 is the distance from the origin .thus any straight line on the flat @xmath87 turns into a biharmonic curve on @xmath128 which is the metric , in local isothermal coordinates , of the enneper minimal surface .figure [ enneper ] is a plot of the enneper surface in polar coordinates , so that radial curves on the picture are biharmonic .let @xmath129 be an isometric immersion of codimension - one .we denote by @xmath130 the second fundamental form of @xmath5 , by @xmath131 a unit normal vector field to @xmath9 and by @xmath132 the mean curvature vector field of @xmath5 ( @xmath133 the mean curvature function ). then we have [ pro - b - ricci ] let @xmath129 be an isometric immersion of codimension one and @xmath134 its mean curvature vector. then @xmath5 is biminimal if and only if : @xmath135 for some value of @xmath11 in @xmath53 . in a local orthonormal frame @xmath136 on @xmath31 ,the tension field of @xmath5 is @xmath137 and its bitension field is : @xmath138\\ = & n \sum_{i=1}^{n } \left [ -\nabla^{\phi}_{e_{i}}\left ( e_{i}(h ) { \mathbf n } + h \nabla^{\phi}_{e_{i}}{\mathbf n } \right ) + ( \nabla_{e_{i}}e_{i})(h ) { \mathbf n } + h\nabla^{\phi}_{\nabla_{e_{i}}e_{i } } { \mathbf n } - \right . \\ & \left.- h r^{n^{n+1}}(d\phi(e_{i } ) , { \mathbf n})d\phi(e_{i } ) \right]\\ = & n \sum_{i=1}^{n } \left [ -e_{i}e_{i}(h ) { \mathbf n } - 2 e_{i}(h ) \nabla^{\phi}_{e_{i}}{\mathbf n } - h \nabla^{\phi}_{e_{i } } \nabla^{\phi}_{e_{i } } { \mathbf n } + ( \nabla_{e_{i}}e_{i})(h ) { \mathbf n } + \right.\\ & \left.+ h\nabla^{\phi}_{\nabla_{e_{i}}e_{i } } { \mathbf n}\right ] - nh \sum_{i=1}^{n } r^{n^{n+1 } } ( d\phi(e_{i } ) , { \mathbf n})d\phi(e_{i } ) \\ = & n ( \delta h ) { \mathbf n } - 2n \sum_{i=1}^{n } e_{i}(h ) \nabla^{\phi}_{e_{i}}{\mathbf n } + n h \delta^{\phi } { \mathbf n } - nh\sum_{i=1}^{n } r^{n^{n+1 } } ( d\phi(e_{i } ) , { \mathbf n})d\phi(e_{i } ) .\end{aligned}\ ] ] but * @xmath139 ; * @xmath140 . for @xmath141 ,first we have : @xmath142 then , if @xmath130 is the second fundamental form of @xmath5 , which , in an orthonormal frame @xmath143 , is defined by : @xmath144 we have @xmath145 which implies that @xmath146 in conclusion : @xmath147 an isometric immersion @xmath148 into a space form of constant curvature @xmath65 is biminimal if and only if there exists a real number @xmath11 such that : @xmath149 where @xmath133 is the mean curvature and @xmath90 the scalar curvature of @xmath74 .moreover , an isometric immersion @xmath150 from a surface to a three - dimensional space form is biminimal if and only if : @xmath151 for some @xmath11 in @xmath53 .let @xmath152 be an orthonormal frame of @xmath74 corresponding to the principal curvatures @xmath153 and @xmath130 its second fundamental form , then : @xmath154 where @xmath155 is the sectional curvature on @xmath74 of the plane spanned by @xmath156 and @xmath157 , and @xmath158 is the scalar curvature of @xmath74 .since @xmath159 , the map @xmath5 is biminimal if and only if : @xmath160 for some @xmath11 in @xmath53 .condition , for free biminimal immersions , is very similar to the equation of the willmore problem ( @xmath161 ) but the minus sign in rules out the existence of compact solutions when @xmath162 .we shall now describe some constructions to produce examples of biminimal immersions . recall that a submersion @xmath163 between two riemannian manifolds is _ horizontally homothetic _ if there exists a function @xmath164 , the _ dilation _ , such that : * at each point @xmath165 the differential @xmath166 is a conformal map with factor @xmath167 , i.e. @xmath168 for all @xmath169 ; * @xmath170 for all horizontal vector fields .[ lemma1 ] let @xmath171 be a horizontally homothetic submersion with @xmath172 and minimal fibres and let @xmath173 be a curve parametrised by arc - length , of signed curvature @xmath174 . then the codimension - one submanifold @xmath175 has mean curvature @xmath176 .let @xmath177 be the frenet frame of @xmath39 , i.e. @xmath178 .choose a local orthogonal frame @xmath179 on @xmath180 such that @xmath181 , @xmath182 and @xmath183 , for @xmath184 . since @xmath5 is a horizontally homothetic submersion, we have that @xmath185 and can choose @xmath186 of unit length . the restriction to @xmath175 of the vector fields @xmath187 and@xmath186 give a local frame of vector fields tangent to the submanifold @xmath188 , while the restriction of @xmath189 gives a unit vector field normal to @xmath190 .therefore the mean curvature of @xmath190 is : @xmath191 where @xmath192 is the mean curvature of the fibres .the fibres being minimal ( @xmath193 ) , we have : @xmath194 moreover : @xmath195 since @xmath196 for a horizontally homothetic submersion ( cf .@xcite ) . finally , @xmath197 and , taking into account , @xmath198 .[ teo - main1 ] let @xmath199 be a horizontally homothetic submersion with dilation @xmath172 , minimal fibres and integrable horizontal distribution , from a space form of constant sectional curvature @xmath65 to a surface .let @xmath200 be a curve parametrised by arc - length such that the surface @xmath201 has constant gaussian curvature @xmath65 .then @xmath201 is a biminimal surface ( w.r.t .@xmath202 ) if and only if @xmath39 is a free biminimal curve . let @xmath177 be the frenet frame of @xmath39 , i.e. @xmath203 . let @xmath204 be a horizontal lift of @xmath39 , so that @xmath205 is horizontal and @xmath206 .let @xmath207 be a local parametrisation of the surface @xmath201 , where , for a fixed @xmath208 , @xmath209 is a parametrisation by arc - length of the fibre of @xmath5 through @xmath210 .then @xmath211 induces on the surface @xmath190 the metric @xmath212 where @xmath172 is the dilation of @xmath5 which , when restricted to the surface @xmath190 , depends only on @xmath90 .the laplacian on @xmath190 is then given by : @xmath213 whilst the gaussian curvature of @xmath190 reduces to : @xmath214 now , assuming that @xmath190 has constant gaussian curvature @xmath215 , from , @xmath190 is biminimal ( w.r.t .@xmath202 ) in @xmath216 if and only if : @xmath217 by lemma [ lemma1 ] , @xmath218 , thus @xmath219\nonumber\\ = & \lambda^3 \left[k_{\gamma}''-k_{\gamma}^3 + \frac{k_{\gamma}}{\lambda^2}\left(c+(\operatorname{grad}(\log\lambda))^2\right ) \right]\label{eq - teo - hh}. \end{aligned}\ ] ] finally , from the generalised oneil formula relating the sectional curvatures of the domain and target manifolds for a given horizontally homothetic submersion with integrable horizontal distribution ( see , for example ( * ? ? ?* corollary 11.2.3 ) ) we get : @xmath220 which , together with , gives : @xmath221 then the theorem follows from corollary [ cor2.1 ] .when the horizontal space is not integrable , we can reformulate theorem [ teo - main1 ] , for riemannian submersions .[ teo - main2 ] let @xmath222 be a riemannian submersion with minimal fibres from a space of constant sectional curvature @xmath65 to a surface of constant gaussian curvature @xmath223 .let @xmath224 be a curve parametrised by arc - length .then @xmath201 is a biminimal surface ( w.r.t .@xmath11 ) if and only if @xmath39 is a biminimal curve ( w.r.t .@xmath225 ) . first , from and , since @xmath226 , we have : @xmath227 thus , taking into account lemma [ lemma1 ] , @xmath190 is biminimal if and only if @xmath228 > from corollary [ cor2.1 ] , the latter equation is clearly biminimality ( w.r.t .@xmath225 ) for a curve @xmath229 .we apply theorem [ teo - main1 ] to construct examples of biminimal surfaces in @xmath230 with the flat metric . 1 .first we consider the orthogonal projection @xmath231 , given by @xmath232 .the projection @xmath233 is clearly a riemannian submersion with minimal fibres ( vertical straight lines in @xmath230 ) and integrable horizontal distribution .thus , from theorem [ teo - main1 ] a vertical cylinder with generatrix a free biminimal curve of @xmath87 is a free biminimal surface . for exampleone can consider the cylinder on the logarithmic spiral .2 . the space @xmath234 can be described as the warped product @xmath235 with the warped metric @xmath236 , @xmath237 being the canonical metric on @xmath238 .then projection onto the second factor @xmath239 is a horizontally homothetic submersion with dilation @xmath240 , integrable horizontal distribution and minimal fibres .geometrically @xmath241 is the radial projection @xmath242 . applying again theorem [ teo - main1 ], we see that the cone on a free biminimal curve on @xmath238 is a free biminimal surface of @xmath230 .for example if we take the parallel on @xmath238 of latitude @xmath243 , which is a biharmonic curve , and thus free biminimal , we get the standard cone of revolution in @xmath230 .3 . the following example does not seem to enter the picture of theorem [ teo - main1 ] .let @xmath244 be a space curve with curvature @xmath63 equal to its torsion @xmath66 and @xmath245 its frenet frame .it is easy to see that the envelope @xmath190 of @xmath39 , parametrised by : @xmath246 , has mean curvature @xmath247 .thus @xmath190 is free biminimal if and only if : @xmath248 geometrically the curve @xmath39 is a curve with constant slope , i.e. there exists a vector @xmath249 such that @xmath250 is constant . then @xmath39 can be described as an helix of the cylinder on a plane curve @xmath251 ( the orthogonal projection of @xmath39 onto a plane orthogonal to @xmath252 ) whose geodesic curvature is a solution of .for example we can take @xmath251 to be the logarithmic spiral of natural equation @xmath253 . 1 .let @xmath255 be the half - space model for the hyperbolic space endowed with the metric of constant sectional curvature @xmath256 given by @xmath257 .then the projection onto the plane at infinity defines a horizontally homothetic submersion @xmath258 with dilation @xmath259 , integrable horizontal distribution and minimal fibres ( vertical lines in @xmath254 ) .then , from theorem [ teo - main1 ] , a vertical cylinder with generatrix a free biminimal curve of @xmath87 is a biminimal surface ( w.r.t .@xmath260 ) in the hyperbolic space . for examplethe cylinder on the logarithmic spiral is free biminimal in @xmath230 whilst it is biminimal ( w.r.t .@xmath260 ) in @xmath254 .2 . let @xmath261 defined by @xmath262 .the fibre of @xmath233 over @xmath263 is the semicircle with centre @xmath263 , radius @xmath113 , and parallel to the coordinate @xmath264-plane .thus the map @xmath233 has minimal fibres .geometrically , this map is a projection along the geodesics of @xmath254 which are orthogonal to @xmath265 .this is again a horizontally homothetic submersion with dilation , along the fibres , @xmath266 , @xmath90 being the arc - length parameter of the fibre .an easy computation shows that for any curve @xmath39 parametrised by arc - length in @xmath265 the surface @xmath267 is of constant gaussian curvature @xmath256 .then , applying theorem [ teo - main1 ] , for any free biminimal curve of @xmath265 @xmath267 is a biminimal surface ( w.r.t .@xmath260 ) in the hyperbolic space . 1 . let @xmath269 be two antipodal points .then the space @xmath270 can be described as the warped product @xmath271 with the warped metric @xmath272 , @xmath237 being the canonical metric on @xmath238 .then the projection to the second factor @xmath239 is a horizontally homothetic submersion with dilation @xmath273 , integrable horizontal distribution and minimal fibres .geometrically , @xmath241 is the projection along the longitudes onto the equatorial sphere .theorem [ teo - main1 ] gives a correspondence between free biminimal curves on @xmath238 and biminimal surfaces ( w.r.t .@xmath75 ) of @xmath268 given by @xmath274 .this is the only example for which we use theorem [ teo - main2 ] .let @xmath275 be the hopf map defined by @xmath276 , where we have identified @xmath277 and @xmath278 .the hopf map is a riemannian submersion with minimal fibres ( great circles in @xmath268 ) .thus , from theorem [ teo - main2 ] , we see that a hopf cylinder @xmath279 is a biminimal surface ( w.r.t .@xmath11 ) of @xmath268 if and only if the curve @xmath39 is a biminimal curve ( w.r.t .@xmath280 ) of @xmath281 .of thurston s eight geometries ( cf .* section 10.2 ) ) , three have constant sectional curvature , @xmath230 , @xmath268 and @xmath254 , and contain biminimal surfaces as described in the previous section , two are riemannian products , @xmath282 and @xmath283 , and will be our first class of examples , two are line bundles , over @xmath87 for @xmath284 and over @xmath285 for @xmath286 , and one , @xmath287 , does not allow riemannian submersion or horizontally homothetic maps with minimal fibres to a surface , even locally , and therefore does not fit our framework . [ [ biminimal - surfaces - of - mathbb - s2-times - mathbb - r - and - mathbb - h2-times - mathbb - r ] ] biminimal surfaces of @xmath282 and @xmath283 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ in both cases , consider the riemannian submersion with totally geodesic fibres , given by the projection onto the first factor , @xmath288 . given a curve @xmath289 parametrised by arc length , take its frenet frame @xmath177 , and consider @xmath290 its horizontal lift .the unit vertical vector @xmath291 completes @xmath292 into an orthonormal frame of @xmath293 , such that @xmath294 is a basis of @xmath295 , for @xmath267 , with @xmath296 the normal to the surface .then , from lemma [ lemma1 ] the mean curvature of @xmath190 is @xmath297 , where @xmath63 is the signed curvature of @xmath39 , and , from proposition [ pro - b - ricci ] , @xmath190 is biminimal ( w.r.t .@xmath11 ) in @xmath298 if : @xmath299 with respect to the frame @xmath294 the matrix associated to the second fundamental form of @xmath190 is : @xmath300 besides , @xmath301 in both cases , using equation , @xmath302 , so @xmath190 is biminimal in @xmath298 if : @xmath303 and @xmath304 now comparing with , we have the following the three - dimensional heisenberg space @xmath284 is the two - step nilpotent lie group standardly represented in @xmath306 by @xmath307\ ] ] with @xmath308 .endowed with the left - invariant metric @xmath309 @xmath310 has a rich geometric structure , reflected by the fact that its group of isometries is of dimension @xmath311 , the maximal possible dimension for a metric of non - constant curvature on a three - manifold .also , from the algebraic point of view , this is a two - step nilpotent lie group , i.e. `` almost abelian '' .an orthonormal basis of left - invariant vector fields is given , with respect to the coordinates vector fields , by let now @xmath313 be the projection @xmath314 . at a point @xmath315 the vertical space of the submersion @xmath233 is @xmath316 and the horizontal space is @xmath317 .an easy computation shows that @xmath233 is a riemannian submersion with minimal fibres .take a curve @xmath318 in @xmath87 , parametrized by arc length , with signed curvature @xmath63 , and consider the flat cylinder @xmath267 in @xmath284 .since the left invariant vector fields are orthonormal , the vector fields @xmath319 give an orthonormal frame tangent to @xmath190 and @xmath320 is a unit normal vector field of @xmath190 in @xmath284 .the second fundamental form of @xmath190 is @xmath321 clearly @xmath322 , @xmath323 and a direct computation shows that @xmath324 .thus , from , @xmath190 is biminimal w.r.t .@xmath11 if and only if @xmath325 or equivalently @xmath326 finally , taking into account , we have : then the projection @xmath331 defined by @xmath332 is a submersion and if we denote , as usual , by @xmath265 the space @xmath285 with the hyperbolic metric @xmath333 , the submersion @xmath334 becomes a riemannian submersion with minimal fibres .the vertical space at a point @xmath335 is @xmath336 and the horizontal space at @xmath100 is @xmath337 , where now , given a curve @xmath339 on @xmath265 , parametrized by arc length , and the flat cylinder @xmath267 in @xmath286 , as @xmath340 and @xmath341 are orthonormal , the vector fields @xmath342 give an orthonormal frame tangent to @xmath190 and @xmath343 is a unit normal vector field of @xmath190 in @xmath286 .

<S> we study biminimal immersions , that is immersions which are critical points of the bienergy for normal variations with fixed energy . we give a geometrical description of the euler - lagrange equation associated to biminimal immersions for : i ) </S><S> biminimal curves in a riemannian manifold , with particular care to the case of curves in a space form ii ) isometric immersions of codimension one in a riemannian manifold , in particular for surfaces of a three - dimensional manifold . </S><S> we describe two methods to construct families of biminimal surfaces using both riemannian and horizontally homothetic submersions . </S>

nucl-th9804062

understanding the pion is a key problem in strong interaction physics . as the lowest mass excitation in the strong interaction spectrum it must provide the long - range attraction in potentials @xcite . in qcd, it is a quark - antiquark bound state whose low- and high - energy properties should be understandable in terms of its internal structure , and it is also that nearly - massless , collective excitation which is the realisation of the goldstone mode associated with dynamical chiral symmetry breaking ( dcsb ) .an explanation of these properties requires a melding of the study of the many body aspects of the qcd vacuum with the analysis of two body bound states .the dyson - schwinger equations ( dses ) provide a single , poincar invariant framework that is well suited to this problem .the dses are a system of coupled integral equations and truncations are employed to define a tractable problem . in truncating the system it is straightforward to preserve the global symmetries of a gauge field theory @xcite and , although preserving the local symmetry is more difficult , progress is being made @xcite .the approach has been applied extensively @xcite to the study of confinement , and to dcsb where the similarity between the ground state of qcd and that of a superconductor can be exploited , with the qcd gap equation realised as the quark dse .it has also been employed in studying meson - meson and meson - photon interactions @xcite , heavy meson decays @xcite , qcd at finite temperature and density @xcite , and strong interaction contributions to weak interaction phenomena @xcite .studying the pion as a bound state requires an understanding of its ( fully - amputated ) bethe - salpeter amplitude , which has the general form @xmath10\,,\end{aligned}\ ] ] where @xmath11 are the pauli matrices .@xmath12 satisfies the renormalised , homogeneous bethe - salpeter equation @xmath13_{tu } & = & \int^\lambda_q \ , [ \chi_\pi^j(q;p)]_{sr } \,k^{rs}_{tu}(q , k;p)\,,\end{aligned}\ ] ] where @xmath1 is the relative and @xmath2 the total momentum of the quark - antiquark pair , @xmath14 , @xmath15, ,@xmath16 represent colour , flavour and dirac indices , @xmath17 , and @xmath18 represents mnemonically a translationally - invariant regularisation of the integral , with @xmath19 the regularisation mass - scale : @xmath20 is the last step in any calculation . in eq .( [ genbse ] ) , @xmath21 is the fully - amputated , renormalised quark - antiquark scattering kernel and @xmath22 is the renormalised dressed - quark propagator , which is the solution of @xmath23 where @xmath24 is the renormalised dressed - gluon propagator , @xmath25 is the renormalised dressed - quark - gluon vertex and @xmath26 is the lagrangian current - quark bare mass . in eq .( [ gendse ] ) , @xmath27 and @xmath28 are the renormalisation constants for the quark - gluon vertex and quark wave function , and the chiral limit is obtained with @xmath29 .the solution of eq .( [ gendse ] ) has the general form @xmath30 also important in the study of the pion is the chiral - limit , axial - vector ward - takahashi identity @xmath31 this identity relates the renormalised , dressed - quark propagator to the renormalised axial - vector vertex , which satisfies @xmath32_{tu } & = & z_2 \ , \left[\gamma_5\gamma_\mu \frac{\tau^j}{2}\right]_{tu } \,+ \int^\lambda_q \ , [ \chi_{5\mu}^j(q;p)]_{sr } \,k^{rs}_{tu}(q , k;p)\,,\end{aligned}\ ] ] where @xmath33 . in the chiral limit the general solution of eq .( [ genave ] ) is @xcite @xmath34\\ & + & \nonumber \tilde\gamma_{5\mu}^{j}(k;p ) + \,f_\pi^0\ , \frac{p_\mu}{p^2 } \,\gamma_\pi(k;p)\,,\end{aligned}\ ] ] where : @xmath35 , @xmath36 , @xmath37 and @xmath38 are regular as @xmath39 ; @xmath40 ; @xmath41 is the amplitude in eq .( [ genpibsa ] ) ; and the residue of the pole in the axial - vector vertex is @xmath42 , the chiral - limit leptonic decay constant , which is obtained from : @xmath43\,.\end{aligned}\ ] ] [ this expression is valid for arbitrary values of the quark mass . ] now , independent of assumptions about the the form of @xmath21 , it follows @xcite from eqs .( [ sp ] ) , ( [ avwti ] ) and ( [ truavv ] ) that @xmath44 where @xmath45 and @xmath46 are the chiral limit solutions of eq . ( [ gendse ] ) . a necessary consequence of eqs .( [ bwti])-([gwti ] ) is that the pseudovector components @xmath47 and @xmath48 , and the pseudotensor component @xmath49 , are nonzero in eq .( [ genpibsa ] ) .this corrects a misapprehension @xcite that only @xmath50 and has important phenomenological consequences .to highlight one such consequence we note that eq .( [ genbse ] ) , the homogeneous bethe - salpeter equation , does not determine the normalisation of the bethe - salpeter amplitude .the canonical normalisation is fixed by requiring that the pion pole in the quark - antiquark scattering amplitude : @xmath51 $ ] , have unit residue . as an alternative , one can normalise the solution of eq .( [ genbse ] ) by requiring that @xmath52 in the chiral limit . in terms of the amplitude @xmath53 defined in this way ,the canonical normalisation condition is @xmath54 \right . } \\ & & \nonumber \left . \;\;\;\;\;\;\;\;\;\;\;\;\;\;\;\;\;\;\ ; + { \rm tr } \left [ \bar{\cal g}_\pi^i(q ;- p ) s(q_+ ) { \cal g}_\pi^j(q;p ) \frac{\partial s(q_-)}{\!\!\!\!\!\!\partial p_\mu}\right ] \rule{0mm}{5mm}\right\ } \\ & & \nonumber \;\;\;\;\;\;\;\;\;\ ; + \int^\lambda_{q}\int^\lambda_{k } \,[\bar\chi_{{\cal g}_\pi}^i(q ;- p)]_{sr } \frac{\partial k^{rs}_{tu}(q , k;p ) } { \!\!\!\!\!\!\!\!\!\!\!\!\partial p_\mu}\ , [ \chi_{{\cal g}_\pi}^j(k;p)]_{ut}\,,\end{aligned}\ ] ] where @xmath55 with @xmath56 , the charge conjugation matrix , and @xmath57 denoting the matrix transpose of @xmath58 .equation ( [ pinorm ] ) defines the pion normalisation constant , @xmath59 , which has mass - dimension one .physical observables are expressed in terms of @xmath60 . in the chiral limit ,when all the amplitudes in eq .( [ genpibsa ] ) are retained , one obtains @xcite @xmath61 this result verifies a core assumption in chiral perturbation theory ; i.e. , that the pion field is normalised by @xmath42 , which is implicit in the expression of the chiral field as @xmath62 however , eq .( [ fpinpi ] ) is violated in bound state treatments of the pion that neglect the pseudovector components of the bethe - salpeter amplitude @xcite . in eq .( [ pinorm ] ) provides the best numerical approximation to the pion s leptonic decay constant in analyses that neglect the pseudovector components and employ a @xmath2-independent form for @xmath21 . ]the pseudovector components of the pion also play a special role in the anomalous @xmath63 decay .consider the renormalised , impulse approximation to the axial - vector photon - photon ( avv ) amplitude : , @xmath64 , and a spacelike vector , @xmath65 , has @xmath66 . ]@xmath67\right\}\,,\end{aligned}\ ] ] where @xmath68 , @xmath69 are the photon momenta [ @xmath70 , @xmath71 , @xmath72 is the loop - momentum , and @xmath73 , @xmath74 , @xmath75 , @xmath76 .here @xmath77 is the renormalised , dressed - quark - photon vertex , and it is because this vertex satisfies the vector ward - takahashi identity : @xmath78 that no renormalisation constants appear explicitly in eq .( [ trho ] ) .@xmath77 has been much studied @xcite and , although its exact form remains unknown , its robust qualitative features have been elucidated so that a phenomenologically efficacious ansatz has emerged @xcite @xmath79\,;\\ % \sigma_f(p^2,q^2 ) & : = & \case{1}{2}\,[f(p^2)+f(q^2)]\,,\\ % \delta_f(p^2,q^2 ) & : = & \frac{f(p^2)-f(q^2)}{p^2-q^2}\,,\end{aligned}\ ] ] where @xmath80 . a feature of eq .( [ bcvtx ] ) is that the vertex is completely determined by the renormalised dressed - quark propagator . in landau gauge the quantitative effect of modifications , such as that canvassed in ref .@xcite , is small and can be compensated for by small changes in the parameters that characterise a given model study @xcite . in the chiral limit ( @xmath81 ) using eqs .( [ genpibsa ] ) and ( [ truavv ] ) , the divergence of the avv vertex is @xmath82 where the direct contribution from the axial - vector vertex is @xmath83\,,\end{aligned}\ ] ] and that from the pion bound state is @xmath84\,.\end{aligned}\ ] ] using eqs .( [ awti])-([gwti ] ) , eq .( [ sepa ] ) simplifies : @xmath85\,,\\ % \label{hatt } \hat t^3_{\mu\nu}(k_1,k_2 ) & : = & n_c\int^\lambda_q { \rm tr}_{df}\left[\rule{0mm}{\baselineskip } s(q_1)\,\gamma_5\tau^3 \ , ie_\pi(\hat q ; 0 ) \ , s(q_2)\,i\gamma^\gamma_\mu(q_2,q_{12 } )s(q_{12})\,i\gamma^\gamma_\nu(q_{12},q_1 ) \right]\,.\end{aligned}\ ] ] now using eqs .( [ sp ] ) and ( [ bcvtx ] ) in eq .( [ hatr ] ) yields @xmath86 where ( @xmath87 , @xmath88 , etc . ) , @xmath89 + \sigma_v^{0}\left[s a_0^\prime\,\sigma_v^{0 } + b_0^\prime\,\sigma_s^{0}\right ] \right)\\ % \label{hatrzero } & \equiv & 0\,.\end{aligned}\ ] ] the last line follows because , using eq .( [ sp ] ) to eliminate @xmath90 and @xmath91 in favour of @xmath45 and @xmath46 , the integrand is identically zero . hence the pseudovector components of the neutral - pion bethe - salpeter amplitude _ combine _ with the regular pieces of the axial - vector vertex to generate that part of the avv vertex which is _conserved_. to reveal the anomalous contribution to the divergence , consider eq .( [ hatt ] ) , in which using eqs .( [ sp ] ) and ( [ bcvtx ] ) yields @xmath92 \right)\,.\end{aligned}\ ] ] now , introducing @xmath93 $ ] , eq .( [ calto ] ) simplifies @xmath94 ^ 3}\,,\ ] ] which , using eq .( [ bwti ] ) , yields @xmath95 so that , in the chiral limit , @xmath96 hence the pseudoscalar piece of the neutral - pion bethe - salpeter amplitude provides the only nonzero contribution to the divergence of the avv amplitude .this contribution is just that identified with the `` triangle anomaly '' , and the result is _ independent _ of detailed information about @xmath97 and @xmath98 .it follows straightforwardly from eqs .( [ hattval ] ) and ( [ hatthalf ] ) that @xmath99 we emphasise that in obtaining the results in this section dcsb was crucial , since it originates and is manifest in a nonzero value of @xmath46 , in the identity between @xmath46 and @xmath100 , and in the other identities : eqs .( [ awti])-([gwti ] ) .our derivation is a generalisation of that in ref .@xcite and , to make it simple , particular care was taken in choosing the momentum routing in eq .( [ trho ] ) .this was necessary because it is impossible to simultaneously preserve the vector and axial vector ward - takahashi identities for triangle diagrams in field theories with axial currents that are bilinear in fermion fields .this choice of variables ensures the preservation of the vector ward - takahashi identity , which is tied to electromagnetic current conservation . with another choice of variables ,surface terms arise that modify the value of @xmath101 .however , these are always eliminated by subtraction in any regularisation of the theory that ensures electromagnetic current conservation.@xciteas another example of the importance of @xmath97 s pseudovector components , we consider the electromagnetic pion form factor , calculated in the renormalised impulse approximation : @xmath102,\end{aligned}\ ] ] @xmath103 and @xmath104 .again , no renormalisation constants appear explicitly in eq .( [ pipia ] ) because the renormalised dressed - quark - photon vertex , @xmath105 , satisfies the vector ward - takahashi identity , eq .( [ vwti ] ) .this also ensures current conservation : @xmath106 we note that from the normalisation condition for @xmath107 , eq . ( [ pinorm ] ) , and eqs .( [ vwti ] ) and ( [ pipia ] ) @xmath108 if , and only if , one employs a truncation in which @xmath21 is independent of @xmath2 .one such scheme is the rainbow - ladder truncation of ref .@xcite . to calculate @xmath109 we employ an algebraic parametrisation of the renormalised dressed - quark propagator that efficiently characterises many essential and robust elements of the solutions obtained in studies of the quark dse .this defines eq .( [ pipia ] ) directly @xmath110 ; in particular at the pion mass shell .we introduce the dimensionless functions : @xmath111 , @xmath112 , where @xmath113 , @xmath114 is a mass - scale , with @xmath115 and @xmath116/y$ ] .this five - parameter algebraic form , where @xmath117 is the @xmath118 current - quark mass , combines the effects of confinement as an entire function is motivated by the algebraic solutions of eq .( [ gendse ] ) in refs .the concomitant absence of a lehmann representation is a sufficient condition for confinement .@xcite ] and dcsb with free - particle behaviour at large , spacelike @xmath119 . : @xmath120 and @xmath121 .the parametrisation therefore does not incorporate the additional @xmath122-suppression characteristic of qcd .it is a useful but not necessary simplification , which introduces model artefacts that are easily identified and accounted for .@xmath123 is introduced only to decouple the large- and intermediate-@xmath119 domains . ] the chiral limit vacuum quark condensate in qcd is @xcite : @xmath124 where at one - loop order @xmath125^{\gamma_m ( 1 + \xi/3)}$ ] , with @xmath126 the covariant - gauge fixing parameter ( @xmath127 specifies landau gauge ) and @xmath128 the gauge - independent anomalous mass dimension .the @xmath126-dependence of @xmath129 is just that required to ensure that @xmath130 is gauge independent .the parametrisation of eq .( [ ssm ] ) is a model that corresponds to the replacement @xmath131 in landau gauge , in which case eq .( [ condensate ] ) yields @xmath132 this is the signature of dcsb in the model parametrisation and we calculate the pion mass from @xmath133 when all the components of @xmath97 are retained , eq .( [ gmor ] ) yields an approximation to the pion mass found in a solution of the bethe - salpeter equation that is accurate to 2% @xcite .the model parameters are fixed by requiring a good description of a range of pion observables .this procedure explores our hypothesis that the bulk of pion observables can be understood as the result of nonperturbative dressing of the quark and gluon propagators .the chebyshev moments of the scalar functions in @xmath41 are , for example , @xmath134 with @xmath135 , where @xmath136 is a chebyshev polynomial of the second kind . at large-@xmath137 , independent of assumptions about the form of @xmath21 , one has @xcite @xmath138 @xmath139 , @xmath140 and @xmath141 have precisely the same behaviour ; i.e. , the asymptotic momentum - dependence of all these functions is identical to that of @xmath142 .this makes manifest the `` hard - gluon '' contribution to @xmath109 in eq .( [ pipia ] ) .further , in an asymptotically free theory , where a well constructed rainbow - ladder truncation yields model - independent results at large-@xmath137 @xcite , @xmath143 with @xmath144 . in a model exemplar used in ref .@xcite the zeroth chebyshev moments provided results for @xmath145 and @xmath146 that were indistinguishable from those obtained with the full solution . also @xmath147 and hence it was quantitatively unimportant in the calculation of @xmath145 and @xmath146 .we expect that these results are not specific to that particular model . in the latter case because the right - hand - side of eq .( [ gwti ] ) is zero and hence in general there is no `` seed '' term for @xmath49 . these observations , combined with eqs .( [ bwti])([gwti ] ) , motivate a model for @xmath97 : @xmath148 with @xmath149 , @xmath150 $ ] and @xmath151 .the relative magnitude of these functions at large @xmath137 is chosen to reproduce the numerical results of ref .@xcite .we determined the model parameters by optimising a least - squares fit to @xmath146 , @xmath152 and @xmath153 , and a selection of pion form factor data on the domain @xmath154\,$]gev@xmath8 .the procedure does not yield a unique parameter set with , for example , the two sets : @xmath155 providing equally good fits , as illustrated in table [ tabres ] .. here @xmath156gev whereas @xmath157gev in ref .it is also qualitatively similar to the numerical solution obtained in ref .@xcite , where @xmath158gev . indeed , our results are not sensitive to details of the fitting function : fitting with different confining , algebraic forms yields @xmath98 that is pointwise little changed , and the same results for observables . ] there is a domain of parameter sets that satisfy our fitting criterion and they are distinguished only by the calculated magnitude of the pion form factor at large-@xmath4 .the two sets in eq .( [ params ] ) delimit reasonable boundaries and illustrate the model dependence in our result . with all parameter sets in the acceptable domain , eq .( [ fpinpi ] ) is satisfied exactly in the chiral limit , in which case we obtain @xmath159gev , while at the fitted value of @xmath160 , @xmath161 . in our calculation @xmath162is 20% too small .this discrepancy can not be reduced in impulse approximation because the nonanalytic contributions to the dressed - quark - photon vertex associated with @xmath163-@xmath163 rescattering and the tail of the @xmath164-meson resonance are ignored @xcite .it can only be eliminated if these contributions are included .we have thus identified a constraint on realistic , impulse approximation calculations : they should not reproduce the experimental value of @xmath162 to better than @xmath165% , otherwise the model employed has unphysical degrees - of - freedom .our calculated pion form factor is compared with available data in figs .[ smallf ] and [ midf ] .it is also compared with the result obtained in ref .@xcite , wherein the calculation is identical _ except _ that the pseudovector components of the pion were neglected . figure [ smallf ] shows a small , systematic discrepancy between both calculations and the data at low @xmath4 , which is due to the underestimate of @xmath152 in impulse approximation . in ref. @xcite . however , the mass - scale is fixed so that @xmath166 , which is why this result appears to agree better with the data at small-@xmath4 : @xmath152 is larger . ]the results obtained with or without the pseudovector components of the pion bethe - salpeter amplitude are quantitatively the same , which indicates that the pseudoscalar component , @xmath100 , is dominant in this domain .the increasing uncertainty in the experimental data at intermediate @xmath4 is apparent in fig .[ midf ] , as is the difference between the results calculated with or without the pseudovector components of the pion bethe - salpeter amplitude .these components provide the dominant contribution to @xmath109 at large pion energy because of the multiplicative factors : @xmath167 and @xmath168 , which contribute an additional power of @xmath4 in the numerator of those terms involving @xmath169 , @xmath170 and @xmath171 relative to those proportional to @xmath172 . using the method of ref .@xcite and the model - independent asymptotic behaviour indicated by eq .( [ uvbeh ] ) we find @xmath173 i.e. , @xmath174 , up to calculable @xmath175-corrections . if the pseudovector components of @xmath97 are neglected , the additional numerator factor of @xmath4 is missing and one obtains @xcite @xmath176 in our model the behaviour identified in eq .( [ fuv ] ) becomes apparent at @xmath177 .this is the domain on which the asymptotic behaviour indicated by eq .( [ uvbeh ] ) is manifest .our calculated results , obtained with the two sets of parameters in eq .( [ params ] ) , illustrate the model dependent uncertainty : @xmath178 this uncertainty arises primarily because the model allows for a change in one parameter to be compensated by a change in another . in our example : @xmath179 but @xmath180 ; and @xmath181 .this allows an equally good fit to low - energy properties but alters the intermediate-@xmath4 behaviour of @xmath109 . in a solution of eq .( [ gendse ] ) these coefficients of the @xmath182 and @xmath183 terms are correlated and such compensations can not occur . as a comparison , evaluating the leading - order perturbative - qcd result with the asymptotic quark distribution amplitude : @xmath184 , yields @xmath185 , assuming a value of @xmath186 .however , the perturbative analysis neglects the anomalous dimension accompanying condensate formation.)-([gwti ] ) are not satisfied in ref .using the dyson - schwinger equations it is straightforward to show that , as a consequence of the dynamical chiral symmetry breaking mechanism , the pion is a nearly - massless , pseudoscalar , quark - antiquark bound state @xcite . as a corollary, the complete pion bethe - salpeter amplitude necessarily contains pseudovector and pseudotensor components , which are always qualitatively important . in model studies , the quantitative effect of these componentscan be obscured in the calculation of many pion observables ; i.e. , within a judiciously constructed framework , applied at low- to intermediate - energy , their effect can be absorbed into the values of the model parameters @xcite .however , they are crucial to a proper realisation of anomalous current divergences , crucial to obtaining a uniformly accurate connection between the low- and high - energy domains , and they provide the dominant contribution to the electromagnetic pion form factor at @xmath187gev@xmath8 .this work was supported by the us department of energy , nuclear physics division , under contract number w-31 - 109-eng-38 and benefited from the resources of the national energy research scientific computing center .r. b. wiringa , v. g. j. stoks and r. schiavilla , phys .c * 51 * , 38 ( 1995 ) ; and references therein .a. bender , c. d. roberts and l. v. smekal , phys .b * 380 * ( 1996 ) 7 ; c. d. roberts , in _ quark confinement and the hadron spectrum ii _ , edited by n. brambilla and g. m. prosperi ( world scientific , singapore , 1997 ) , pp .224 - 230 .a. bashir , a. kizilersu and m.r .pennington , phys .d * 57 * , 1242 ( 1998 ) ; and references therein . c. d. roberts and a. g. williams , prog .phys . * 33 * , 477 ( 1994 ) .p. c. tandy , prog .. phys . * 39 * , 117 ( 1997 ) ; m. a. pichowsky and t .- s .h. lee , phys .d * 56 * , 1644 ( 1997 ) .m. a. ivanov , yu . l. kalinovsky , p. maris and c. d. roberts , phys .rev . c. * 57 * , 1991 ( 1998 ) ; and references therein .d. blaschke , c.d .roberts and s. schmidt , `` thermodynamic properties of a simple , confining model '' , nucl - th/9706070 , phys .b , in press ; and references therein . yu . kalinovsky , k. l. mitchell and c. d. roberts , phys .b * 399 * , 22 ( 1997 ) .m. b. hecht , b. h. j. mckellar , `` dipole moments of the @xmath164-meson '' , hep - ph/9704326 .p. maris , c. d. roberts and p. c. tandy , phys .b * 420 * , 267 ( 1998 ) .p. maris and c. d. roberts , phys .c * 56 * , 3369 ( 1997 ) .r. delbourgo and m. d. scadron , j. phys .g * 5 * , 1621 ( 1979 ) . j. s. ball and t .- w .chiu , phys .d * 22 * , 2542 ( 1980 ) . c. d. roberts , nucl .a * 605 * , 475 ( 1996 ) .r. jackiw , `` field theoretic investigations in current algebra '' , in _ current algebra and anomalies _ ( world scientific , singapore , 1985 ) pp .108 - 141 .m. burkardt , m. r. frank and k. l. mitchell , phys .lett . * 78 * , 3059 ( 1997 ) .h. munczek , phys .b * 175 * , 215 ( 1986 ) ; c. j. burden , c. d. roberts and a. g. williams , _ ibid _ * 285 * , 347 ( 1992 ) . c. d. roberts , a. g. williams and g. krein , int .phys . a * 4 * , 1681 ( 1992 ) .d. c. curtis and m. r. pennington , phys .d * 46 * , 2663 ( 1992 ) .f. t. hawes , c. d. roberts and a. g. williams , phys .d * 49 * , 4683 ( 1994 ) .d. b. leinweber , ann . phys . * 254 * , 328 ( 1997 ) .particle data group( r. m. barnett _et al_. ) , phys .d * 54 * , 1 ( 1996 ) .s. r. amendolia , _ et al_. , nucl .b * 277 * , 168 ( 1986 ) . c. j. bebek , _ et al_. , phys .d * 13 * , 25 ( 1976 ) . c. j. bebek , _ et al_. , phys .d * 17 * , 1693 ( 1978 ) .g. r. farrar and d. r. jackson , phys .lett . * 43 * , 246 ( 1979 ) ..a comparison between our calculated values of low - energy pion observables and experiment or , in the case of @xmath188 and @xmath189 , the values estimated using other theoretical tools .each of the parameter sets in eq .( [ params ] ) yields the same calculated values . for consistency with ref .@xcite , we use @xmath190gev throughout .[ tabres ] [ cols="^,<,<",options="header " , ]

<S> as a consequence of dynamical chiral symmetry breaking the pion bethe - salpeter amplitude necessarily contains terms proportional to @xmath0 and , where @xmath1 is the relative and @xmath2 the total momentum of the constituents . </S><S> these terms are essential for the preservation of low energy theorems , such as the gell - mann oakes - renner relation and those describing anomalous decays of the pion , and to obtaining an electromagnetic pion form factor that falls as @xmath3 for large @xmath4 , up to calculable @xmath5-corrections . in a simple model , which correlates </S><S> low- and high - energy pion observables , we find @xmath6 - @xmath7gev@xmath8 for @xmath9 . </S>

1501.07412

in this paper in the context of ads / cft @xcite we study numerically the ground state of a pair of open strings stretching between two coincident @xmath2-branes with opposite orientations in @xmath3 of @xmath4 .the main motivation for the study is that according to string - theory the ground state of such a configuration is expected to be tachyonic for large values of the t hooft coupling @xcite . in our workwe rely on the perturbatively discovered and later `` all loop conjectured '' integrability @xcite of both the @xmath4 super - string and the dual large @xmath5 gauge theory . for string configurations with d - branes integrability enabled one to describe string configurations ending on different types of d - branes as 1-dimensional integrable scattering theories with boundaries @xcite .this formulation of the problem makes it possible to go beyond the approaches of perturbative gauge and string theories being valid for small and large values of the t hooft coupling respectively , and to determine the exact spectrum of the model at any value of the coupling constant .however , even with the help of the powerful techniques offered by integrability , the exact analytical solution of the problem is not possible .remarkeble analytical results are available in the small @xcite and large @xcite coupling regimes , but the determination of the spectrum at any value of the coupling constant can only be carried out by high precision numerical solution @xcite of the corresponding nonlinear integral equations .in our paper we consider the case , when the two @xmath2-branes are giant gravitons @xcite , namely they carry @xmath5 units of angular momenta in @xmath3 .if the @xmath3 of @xmath4 is parametrized by three complex coordinates @xmath6 satisfying the constraint : @xmath7 , then our @xmath2-brane and anti-@xmath2-brane are given by the conditions @xmath8 and @xmath9 respectively .they wrap the same @xmath10 , but with opposite orientation . as a consequence of gauss law sucha system can support only even number of open strings .for this reason we study the minimal number of allowed open strings , a single pair , ending on our d - brane anti - d - brane ( @xmath11 ) system with open string angular momenta @xmath12 and @xmath13 . on the large @xmath5 gauge theory side a @xmath8 braneis represented by a determinant operator @xcite composed of @xmath5 copies of the field @xmath14 : @xmath15 where @xmath16 and @xmath17 are color indices and @xmath18 is a product of two regular epsilon tensors @xmath19 . the local operator corresponding to an open string ending on a @xmath8 giant graviton can be obtained from ( [ detyy ] ) by replacing one @xmath14 field with an adjoint valued operator @xmath20 @xcite : @xmath21 the gauge theory description of a pair of open strings stretching between two @xmath22-branes is given by a double determinant operator , such that the string insertions @xmath20 and @xmath23 connect the two determinants of the @xmath14 fields : @xmath24 unfortunately , the precise gauge theory dual of the @xmath11-system of our interest is not known .in @xcite it was approximated by a double determinant operator similar to ( [ detyywv ] ) , but in one of the determinants the @xmath14 fields are replaced with @xmath25 fields and @xmath25 , but should be similar to the double determinant form ( [ yybar ] ) and the mixing with other fields seem to be suppressed at large @xmath5 ] : @xmath27 for the ground state the insertions are @xmath28 and @xmath29 respectively . based on one - loop resultsthe planar dilatation operator is expected to act independently on the two words @xmath30 corresponding to the open string states @xcite : @xmath31 = \delta_{\rm bare } [ { \cal o}_{y\bar y}^{{\cal w } , { \cal v } } ] + \delta \delta [ { \cal w}_{y\bar y } ] + \delta \delta [ { \cal v}_{\bar y y } ] .\label{dim split}\ ] ] this observation allows us to apply the boundary thermodynamic bethe ansatz technique ( btba ) @xcite to each open string separately .the necessary ingredients of this technique are the boundary reflection factors @xcite and the asymptotic bethe equations of the problem @xcite .unfortunately , apart from some very special cases @xcite , it is still unknown how to derive btba equations for a general non - diagonal scattering theory in the context of the thermodynamical considerations of @xcite .this is why in @xcite the @xmath14-system @xcite and the related discontinuity @xcite equations supplemented by analyticity assumptions compatible with the asymptotic solution @xcite were used to derive btba equations for the nonperturbative study of the ground state of the @xmath11-system @xcite .the btba description of the system is an infinite set of nonlinear integral equations .the numerical solution of the equations @xcite showed that the ground state energy is a monotonously decreasing function of the coupling constant and the t hooft coupling @xmath32 is given by : @xmath33 . ]the analytical investigation of the large rapidity and large index behavior of the @xmath14-functions of the btba revealed that the usual btba description of the system breaks down when the energy of an open string state with angular momentum @xmath12 gets close to the critical value : @xmath35 .this point was interpreted in @xcite as a transition point where the ground state becomes tachyonic . approaching the critical point the contribution of infinitely many @xmath14-functions must be taken into account to get accurate numerical result for the energy .this fact suggests reformulating the finite size problem in terms of finite number of unknown functions. the possible candidates could be the finlie @xcite , the quantum spectral curve ( qsc ) @xcite or the hybrid - nlie ( hnlie ) @xcite formulation of the problem .since at present it is not known ( not even for the konishi problem ) how to use the analytically very efficient @xcite qsc method for numerical purposes , we choose the hnlie method to reformulate the finite size problem of the @xmath11-system . in this paperwe transformed the infinite set of boundary tba equations @xcite into a finite set of hybrid - nlie type of nonlinear integral equations .we perform the extensive numerical study of these type of equations in order to get as close to the special @xmath36 critical point as it is possible .our numerical results reproduce the numerical evaluation of the boundary lscher formula @xcite in the linear approximation , and the numerical btba results of @xcite as well .these numerical comparisons give further numerical checks on the hybrid - nlie technique of @xcite .unfortunately , as @xmath34 increases new local singularities enter the hnlie formulation of the problem .thus we could not approach very close to the critical point .nevertheless , in the range of @xmath34 where physically acceptable numerical results were obtained , the hnlie results could give higher numerical precision than that of the btba and also some interesting facts could be read off from our numerical data . during the numerical solution of the hnlie equations straightforward numerical iterative methods failed to converge ,thus new numerical methods were worked out to solve the equations .the ground state of the @xmath0 state is a very special case , since there the critical point is right at @xmath37 and so far neither perturbative field theory computations nor the boundary lscher formula could provide a finite quantitative answer to the anomalous dimension of this state . on the integrability side the hnlie approach allows us to get some numerical insight into this problem .the outline of the paper is as follows : section 2 . contains the hnlie equations . in section 3 .the numerical method is described . in section 4 .the numerical results and their interpretation is presented .section 5 .contains some comments on the mysterious @xmath0 case and finally our conclusion is given in section 6 .various notations , kernels of the integral equations together with the necessary asymptotic solutions are placed in the appendices of the paper .in this section we transform the previously proposed btba equations of @xcite for the ground state of our d - brane anti - d - brane system to finite component hybrid - nlie equations . for presentational purposes we group the equations into 3 types .there are tba - type equations , horizontal @xmath38 hybrid - nlie type equations , and vertical @xmath39 hybrid - nlie type equations .they together form a closed set of nonlinear integral - equations , which are solved numerically in this paper . as it is usual ,structurally the equations consist of source terms plus convolutions containing coupling dependent kernels and nonlinear combinations of the unknown functions . the objects appearing in the arguments of the source functionsare subjected to quantization conditions , but similarly to the boundary tba description @xcite , due to the @xmath40 symmetry of the problem they are tied to the origin of the complex plane , thus extra quantization conditions are unnecessary to be imposed , since they are automatically satisfied by symmetry . since these source term objects have fixed positions their positions are exactly the same as that of their asymptotic counterparts .this fact saves us from the tedious computation of the source terms , since if we take the difference of the exact equations and their asymptotic counterparts the source terms cancel from the equations . to be pragmatic and save time and space ,the equations will be presented in such a difference form .thus for any combination @xmath41 of the unknown functions , we introduce the notation @xmath42 , where @xmath43 is the asymptotic counterpart of @xmath41 . having introduced this notation , we start the presentation of the equations by the tba - type part . for the labeling of the @xmath14-functions we use the string - hypothesis @xcite based notations of @xcite . for the presentation of the equations a few more notations need to be introduced : @xmath44 , \qquad { \mathscr l}_{m}=\log\left[\tau^2 \left ( 1+\frac{1}{y_{m|vw}}\right ) \right ] , \qquad \tau(u)=\tanh(\frac{\pi g u}{4}).\ ] ] for later numerical purposes we re - parametrize @xmath45 by the formula : @xmath46}(u)}{x^{[-q]}(u)}+\log \bar{y}_q(u ) + c_q+\varepsilon \ , \log\left(u^2+\frac{(q+1)^2}{g^2}\right ) , \qquad q=1,2 , ... \label{separ}\ ] ] such that @xmath47 is the constant value of @xmath45 at infinity and @xmath48 is minus twice the energy multiplier of @xmath48 in ( [ separ ] ) is chosen not to modify the constant term in the large @xmath49 behavior and to satisfy @xmath50 , which is the lhs of an important @xmath14-system equation divided by its asymptotic counterpart . ] : @xmath51 . from the tba equations of the problem @xcite, it follows that @xmath52 is @xmath53-independent , and for small @xmath34 , @xmath54 is a smooth deformation of its asymptotic counterpart , such that @xmath55 tends to zero at infinity .can not be considered as smooth deformation of @xmath56 , because @xmath57 diverges for large @xmath49 at any @xmath34 . on the other hand @xmath58is small for any @xmath49 at small @xmath34 and tends to zero at infinity . ] using this decomposition the following notations are need to be introduced : @xmath59 then the tba - type equations take the form : @xmath60\star s- \log ( 1+y_{m+1})\star s,\quad 2\leq m \leq p_0 - 2 , \label{tbamvw}\ ] ] @xmath61\ \hat\star\ s,\label{tba1vw}\ ] ] @xmath62\star \ !s+\ ! \ ! \sum_{q=1}^{p_0 - 2 } \log ( 1+y_q)\star\left [ -k_q+2k^{q1}_{xv}\star \ !s\right ] \nonumber \\ & -&\omega(k_q)+2 \ , \omega(k^{q1}_{xv}\star \ !s ) \label{ypym}\end{aligned}\ ] ] for @xmath63 the modified hybrid form @xcite of the btba equations is used , @xmath64\ \hat\star\ s\star k^{11}_{vwx } + 2\delta{\mathscr l}_-\ \hat\star\ k^{y1}_-+2\delta{\mathscr l}_+\ \hat\star\ k^{y1}_+\\ & + \sum_{q=1}^{p_0 - 2 } \log(1+y_{q})\star\left[k^{q1}_{{\mathfrak{sl}(2 ) } } + 2s\star k^{q-1,\,1}_{vwx } \right ] \\ & + \omega(k^{q1}_{{\mathfrak{sl}(2)}})+2 \omega(s\star k^{q-1,\,1}_{vwx } ) , \end{split } \label{hybrid}\ ] ] where @xmath65 is the index limit starting from which the upper part of the tba equations is replaced by an @xmath39 nlie of @xcite ( see figure [ yhnlie ] . ) . for any kernel vector appearing in the tba equations@xmath66 denotes the residual sum @xmath67 , and following the method of @xcite for @xmath68 it can be expressed by next to nearest neighbor @xmath14-functions as follows : @xmath69 where @xmath70 , the kernels @xmath71 are hyperbolic functions @xcite , @xmath72 while the other tba kernels can be found in appendix a. as a consequence of the re - parametrization ( [ separ ] ) the two constants @xmath73 and @xmath48 also become part of the set of equations notation means simply integration from @xmath74 to @xmath75 . ] : @xmath76 @xmath77\ \hat\star\ s\star ck^{11}_{vwx } + 2\delta{\mathscr l}_-\ \hat\star\ ck^{y1}_-+2\delta{\mathscr l}_+\ \hat\star\ ck^{y1}_+\\ & + \sum_{q=1}^{p_0 - 2 } \log(1+y_{q})\star\left[ck^{q1}_{{\mathfrak{sl}(2 ) } } + 2s\star ck^{q-1,\,1}_{vwx } \right ] \\ & + \omega(ck^{q1}_{{\mathfrak{sl}(2)}})+2 \omega(s\star ck^{q-1,\,1}_{vwx } ) , \end{split } \label{chybrid}\ ] ] where for any kernel @xmath78 : @xmath79 denotes the constant term in the large @xmath80 expansion of @xmath81 .expansion , all the other kernels has either constant term or they simply vanish at infinity .] as we mentioned @xmath82 is the tba energy , thus ( [ eps ] ) gives the energy formula in our formulation of the finite size problem .the asymptotic forms of the @xmath14-functions necessary for the formulation of ( [ tbamvw]-[chybrid ] ) are listed in appendix d. to close the discussion of the tba - type equations we note that equations ( [ ymperyp ] ) and ( [ ypym ] ) determine @xmath83 up to an overall sign factor .the sign factor can be fixed from the asymptotic solution and its value is @xmath84 .thus the fermionic y - functions can be expressed in terms of the lhs of ( [ ymperyp ] ) and ( [ ypym ] ) by the formula : @xmath85 the horizontal @xmath38 wing of the tba is resumed by an @xmath38-type nlie @xcite , which in our case takes the form : @xmath86}\star\delta\log(1+y_{1|w})+g\star\delta\log(-1-b)- g^{[-2\gamma ] } \star \delta\log(-1-\bar{b}),\ ] ] @xmath87}\star\delta\log(1+y_{1|w})+g\star\delta\log(-1-\bar{b})- g^{[2\gamma]}\star\delta\log(-1-b),\ ] ] @xmath88 } \star \delta\!\log(-1-b ) + s^{[1-\gamma ] } \star \delta\!\log(-1-\bar{b})+ \delta\!\log\left[\frac{1-\frac{1}{y_-}}{1-\frac{1}{y_+}}\right ] \\hat\star\ s \,,\label{tba1w}\ ] ] where @xmath89 is a contour shift parameter , the kernel @xmath90 is given by ( [ g ] ) and the asymptotic solution for @xmath91 and @xmath92 is given in appendix d and @xmath92 are complex conjugate of each other . ] .the upper @xmath39 nlie of @xcite is attached to the tba equations at the @xmath65-th node .the upper nlie is for 12 complex unknown functions : @xmath93 and @xmath94 .they are combinations of the @xmath95-functions of the upper wing @xmath39 bcklund - hierarchy @xcite .their relations to the unknowns introduced in @xcite are given by ( [ buj],[duj ] ) in appendix b and their asymptotic forms are given in appendix c. using the notation @xmath96 and @xmath97 , the equations they satisfy take the form : @xmath98 @xmath99 where the kernels are given in ( [ 93a]-[kdd ] ) .the shifts in the kernels which is equivalent to fixing the lines on which the nlie variables live , are chosen in a symmetrical way and fixed as follows : @xmath100 @xmath101 this choice satisfies the constraint inequalities of @xcite and satisfy the relation @xmath102 with @xmath103 given by ( [ m ] ) .its advantage is that choosing the @xmath104 asymptotic solution from appendix c to formulate the equations , the @xmath91- and @xmath105-type variables are related in a simple manner : @xmath106 in practice this reduces to half the number of @xmath39 nlie variables .the vectors @xmath107 and @xmath108 are conjugate to each other and they give the tba input into the upper nlie . to give their form we introduce the notations : @xmath109 } , \qquad { \bar{\eta}_1 } = y_{p0 - 1|vw}^{[\epsilon_3 ] } , \qquad \epsilon_1=-\epsilon_3=-\frac{7}{12},\ ] ] @xmath110 } \star \delta \! \log ( 1+\eta_1),\qquad \qquad e_3=s^{[\frac56 ] } \star \delta \ ! \log ( 1+\bar{\eta}_1 ) , \qquad \qquad e_5=e_6=0,\ ] ] @xmath111 } \star \delta \ ! \log ( 1+\eta_1)-\frac12 s^{[\frac23 ] } \star \delta \ !\log ( 1+\bar{\eta}_1)+\varepsilon_2 + i \ , \varphi_2,\ ] ] @xmath112 } \star \delta \ !\log ( 1+\eta_1)-\frac12 s^{[\frac56 ] } \star \delta \ ! \log ( 1+\bar{\eta}_1)+\varepsilon_4 + i \ , \varphi_4,\ ] ] where @xmath113 @xmath114 @xmath115 @xmath116 with @xmath117 the last set of equations gives , how the upper nlie variables couple to the tba part of the equations .\star \delta\!\log(1+\eta_1)+ s \star \delta\!\log ( 1+y_{p_0 - 3|vw})-s \star l_{p_0 - 1 } , \label{yvwp0m2}\ ] ] @xmath119 } \star \delta\!\log b_1 + s^{[1+\epsilon_1-\eta_1 ] } \star \delta\!\log d_1-s^{[\epsilon_1-\gamma_2 ] } \star \delta\!\log b_2 \\ & + s^{[\epsilon_1 ] } \star \delta\!\log ( 1+y_{p_0 - 2|vw } ) -s^{[\epsilon_1 ] } \star l_{p_0 } , \end{split}\ ] ] @xmath120 } \star \delta\!\log b_3 + s^{[1+\epsilon_3-\eta_3 ] } \star \delta\!\log d_3-s^{[\epsilon_3-\eta_2 ] } \star \delta\!\log d_2 \\ & + s^{[\epsilon_3 ] } \star \delta\!\log ( 1+y_{p_0 - 2|vw } ) -s^{[\epsilon_3 ] } \star l_{p_0 } , \end{split}\ ] ] @xmath121 } \star \left [ \delta\!\log\bar{b}_2-\delta\!\log b_2 \right]+ s^{[1-\eta_2 ] } \star \left [ \delta\!\log\bar{d}_2-\delta\!\log d_2 \right ]\\ + s^{[-\gamma_3 ] } \star \delta\!\log \frac{b_3}{b_3}+ s^{[-\eta_1 ] } \star \delta\!\log \frac{d_1}{d_1 } + s^{[-\epsilon_1 ] } \star \delta\!\log \frac{1+\eta_1}{\eta_1 } \\ + s^{[-\epsilon_3 ] } \star \delta\!\log \frac{1+\bar{\eta}_1}{\bar{\eta}_1}- s \star \delta\ ! r_{p_0 - 1 } , \label{lasteq } \end{split}\ ] ] where @xmath122 and @xmath123 are from the re - parametrization of @xmath124 and @xmath125 : @xmath126}(u)}\right)^{2l } \ ,\exp\left\ { \varepsilon \left [ \log\left(u+i\,\frac{\gamma_2-p_0 - 1}{g}\right ) + i \,\frac{\pi}{2 } \right]+\frac{\delta c}{2 } \right\ } \ , \bar{b}_2(u ) , \label{b2bar}\ ] ] @xmath127}(u)}\right)^{2l } \ , \exp\left\ { \varepsilon \left [ \log\left(u+i \ , \frac{\eta_2+p_0 + 1}{g}\right)-i \,\frac{\pi}{2 } \right]+\frac{\delta c}{2 } \right\ } \ , \bar{d}_2(u ) , \label{d2bar}\ ] ] with @xmath128 being a global sign factor .similarly to the definition of @xmath129 , also here the benefit of using @xmath122 and @xmath123 is that , for small @xmath34 , @xmath130 and @xmath131 are smooth deformations of their asymptotic counterparts , and in addition @xmath132 and @xmath133 vanishes at infinity , which is necessary for the convergence of certain integrals .the decompositions ( [ b2bar]),([d2bar ] ) are chosen to be compatible with the functional relation @xmath134 } \ , d_2^{[-\eta_2]}=y_{p_0}$ ] in ref .equations ( [ tbamvw])-([lasteq ] ) constitute our complete set of nonlinear integral equations , which governs the finite size dependence of the vacuum of our d - brane anti - d - brane system .here we describe our numerical method for solving the hybrid - nlie equations presented in the previous section . during the iterative numerical solution of the equations we faced with very serious convergence problems , which forced us to work out such a method that overcomes all the difficulties emerged .our numerical method can be applied to solve other type of nonlinear integral equations as well .the power of the method is shown by the fact that numerical convergence was reached even in such cases , when the solution was physically unacceptable .the numerical method consist of two main steps , namely : * discretization of the equations * iterative solution .the first step involves the discretization of the unknown functions and kernels , furthermore the discrete approximate representation of the convolutions . having carried out the appropriate discretization of the problem ,the equations are considered as large nonlinear algebraic set of equations .thus eventually instead of integral equations we solve discrete algebraic equations . in this paperwe will present two methods to solve them numerically .the discretization serves two goals .first it allows us to reduce the numerical problem from solving integral equations to solving algebraic equations .second choosing the discretization points appropriately it reduces the number of degrees of freedom as much as it is possible to reach the desired numerical accuracy . in our actual numerical computation instead of @xmath49 of section 2 .we used the new rapidity @xmath135 , because with such a scaling almost all the rapidity difference dependent kernels become @xmath34 independent .thus for example @xmath136 will be defined in @xmath137 $ ] . to decrease the number of discretization points the @xmath138 symmetry of the problemis exploited .this means that the @xmath14-functions are to be discretized only on @xmath139 $ ] or @xmath140 $ ] and as for the nlie variables it is enough to discretize the @xmath91- and @xmath105- type variables on @xmath139 $ ] .since we do not want to introduce any cutoff in the rapidity space first we transform the @xmath141 $ ] interval to a finite interval @xmath142 $ ] through the transformation formula : @xmath143 this formula is chosen such that the branch point @xmath144 corresponds to @xmath145 for any choice of the parameter @xmath146 , where @xmath146 is a global scaling factor which changes from unknown to unknown .we chose the values as follows : for @xmath147 @xmath148 , for @xmath91 and @xmath92 @xmath149 , for @xmath150 and @xmath151 @xmath152 , for @xmath153 and @xmath154 @xmath155 , and finally for the @xmath91- and @xmath105-type nlie functions @xmath155 .these values are chosen to preserve the smoothness of the transformed functions in the finite interval .after this transformation all of our unknown functions live on a finite interval .to discretize them we used piecewise chebyshev approximation .this means that we divide the finite interval into subintervals and on each subinterval the functions are approximated by a given order chebyshev series .the choice of subintervals is not equidistant .the subintervals are placed more densely around the branch points , since the function @xmath156 , which governs the decay of the massive @xmath157-functions , has the largest change around this point .the advantage of the chebyshev approximation is that if the function is smooth enough on the subinterval , the coefficients of the chebyshev series decay rapidly and the order of magnitude of the last coefficient allows us to estimate the numerical errors of the procedure .now we describe the discretization method in more detail .our functions are defined on either @xmath158 $ ] or on @xmath159 $ ] .this is why two type of subinterval vectors are defined @xmath160 and @xmath161 , such that the endpoints of the subintervals of @xmath158 $ ] are put into the vector @xmath160 and the endpoints of the subintervals of @xmath159 $ ] define @xmath161 .let @xmath162 be the order of the chebyshev approximation , then using the general rules of the chebyshev approximation , a given function @xmath163 is approximated in the @xmath164th subinterval @xmath165 $ ] as : @xmath166 , \label{tkoz}\ ] ] where now the vector @xmath167 stand for either @xmath160 or @xmath161 , furthermore @xmath168 are a slightly modified chebyshev polynomials ) in a more compact way . ]@xmath169 with @xmath170 being the @xmath171th chebyshev polynomial ] .the coefficients @xmath172 are the chebyshev coefficients of the function @xmath41 , which can be computed from the sampling points of the chebyshev approximation : @xmath173 by the simple formula : @xmath174 where @xmath175 are the zeros of the @xmath162 order chebyshev polynomial : @xmath176 , \qquad \hat{t}_{l_k}(c^{(i)}(l_k))=0 , \qquad i=1, ... ,l_k\ ] ] and @xmath177 is given by : @xmath178 , \qquad \qquad i , k\in\{1, ... ,l_k\}.\ ] ] in our method the next step is to formulate the convolutions and the equations themselves in terms of the discrete values of our functions . herewill sketch the basic idea in some typical scenarios appearing in our equations .then its application to the concrete unknowns and kernels of the problem is straightforward .if one takes the equations at the required discretized points @xmath179 the following typical pattern arises : @xmath180 where @xmath181 is the symmetrized kernel to exploit left - right symmetry of the problem for reducing to half the number of variables .@xmath182 is intended to modelize the variables in the left - hand side of the equations taken at the discretized points of the transformed variable @xmath183 and @xmath184 stands for some nonlinear combination of some unknown function of the equations can be thought of as @xmath185 and @xmath184 can be @xmath186 for any type of @xmath14 . ] .if @xmath187 is discretized by a subinterval vector @xmath167 of @xmath188 $ ] , then the numerical approximation of the right hand side goes as follows ; * first the integration variable is changed from @xmath189 to @xmath183 , * then on each subinterval @xmath187 is approximated by its chebyshev series , * finally the integration is carried out and the convolution is expressed in terms of the discretized values of @xmath187 .the final approximation formula takes the form : @xmath190 where @xmath191 , @xmath192 denotes the dimension of @xmath167 and @xmath193 is the discretized convolution matrix given by the formula : @xmath194 in this manner a convolution is reduced to a discrete matrix - vector multiplication .the other type of typical convolution is when the integration is taken from zero to @xmath144 . in certain casesthe function @xmath184 has square root behavior close to the branch points and @xmath195 . ] . for such functions the truncated chebyshev series does not give accurate approximation . in these casesnot the function @xmath184 is approximated , but that part of it which remains after the elimination of the square root behavior .namely , we write @xmath196 , then @xmath197 is approximated by a truncated chebyshev series and finally the approximate discretized form of the corresponding convolution is very similar to ( [ intapproxt ] ) : @xmath198 where @xmath199 and @xmath200 is the square root factor modified version of ( [ tksmatrix ] ) ; @xmath201 here depending on the left hand side of the equation @xmath202 can stand for @xmath203 , @xmath142 $ ] for some @xmath146 , or it can denote the sampling points on @xmath159 $ ] . applying our discretization technique to all unknowns and convolutions of our equations, we can reduce the integral equations to a discrete set of nonlinear algebraic equations .however , the transformation from integral equations to algebraic equations is obviously not exact .the typical error comes from the fact that on each subinterval the chebyshev series is truncated , so the magnitude of the typical errors in our numerical method is governed by the neglected terms of the chebyshev series , which can be approximated by the magnitude of the last chebysev coefficient . in our casethis is typically somewhere between @xmath204 and @xmath205 .the last step of our numerical method is the iterative solution starting from the asymptotic solution . herewe will describe two methods to solve our integral equations iteratively .since our actual equations have very complicated form , we will describe our methods using a model example , which has similar structure to our equations .let the model equations take the form : @xmath206 where @xmath207 are some kernel matrices , @xmath208 are some source terms and @xmath209s are the unknown functions of the problem .the solution of ( [ modelleq ] ) is expanded around the asymptotic solution and the equations are formulated in terms of the corrections .to fix the conventions , the correction functions @xmath210 are defined by : @xmath211 as a consequence : @xmath212 the source term is also expanded around its asymptotic counterpart : @xmath213 . then equations ( [ modelleq ] ) can be reformulated in terms of the @xmath210 functions as follows : @xmath214 to define the iterative method , ( [ dmodelleq ] ) are reformulated so that only @xmath215 terms remain on the right hand side of the equations .thus the equations are rewritten in the form : @xmath216-\left [ \log(1+\delta y_a)-\delta y_a \right ] .\label{iteq1}\ ] ] it can be seen that on the left hand side of ( [ iteq1 ] ) all the quantities are @xmath217 , while on the right hand side all the quantities are @xmath215 .this separation allows us to define an iterative solution .if @xmath210s are small then the rhs is a small correction with respect to the lhs , this is why in an iterative solution the rhs can be simply taken at the value of the previous iteration .let @xmath218 the value of @xmath210 after the @xmath219th iteration , then @xmath220 can be determined from @xmath218 by solving a set of linear integral equations : @xmath221-\left [ \log(1+\delta y^{(n)}_a)-\delta y^{(n)}_a \right ] .\label{iteq2}\ ] ] thus at each step of this iterative method a set of linear integral equations must be solved . using the discretization method of the previous subsection, the problem reduces to solving a set of linear algebraic equations , which is a straightforward task in numerical mathematics .the very first ( @xmath222th ) iteration starts from the asymptotic solution @xmath223 and it corresponds to the solution of the linearized equations , which in our case gives the lscher - formula for the energy .this ( first ) method in a certain range of the coupling constant defined a numerically convergent iteration to solve the equations for the ground state of our d - brane anti - d - brane problem , but beyond a certain value of @xmath34 the method failed to converge anymore .this is why we worked out a second method , which proved to be much more efficient than the first one .this efficiency is manifested in two facts .first it converges much faster than the previous iterative method , second it gives convergent solutions to our equations even when the solution can not be accepted as physical one .this second method can be described simply in words . instead of defining an iteration as above , we simply take the discretized version of ( [ dmodelleq ] ) .we consider it as a set of nonlinear algebraic equations . as a first step we solve the linearized discrete equations ( i.e. ( [ iteq2 ] ) with @xmath224 ) and starting from the solution of the linearized equations we solve the discrete nonlinear system by newton - method``newton '' ] . ] .in this section we summarize our numerical results .we solved numerically the equations for several integer values of the length parameter @xmath12 . in this sectionwe concentrate on the states with @xmath225 .the @xmath0 special case is discussed in the next section . for the explanation of the numerical data we will mostly use the @xmath226 case as an example , because the critical point of this state is the closest one to zero , so it is enough to work with relatively small values of the coupling constant .this is important from the numerical point of view , since by increasing @xmath34 the numerical method becomes more and more time consuming .first the parameters of the numerical method is discussed .there are three parameters in the nonlinear integral equations ( [ tbamvw])-([lasteq ] ) .the most important one is the coupling constant @xmath34 , then there are two other parameters which allow us to formulate the equations according to our purposes .the two parameters are @xmath65 and @xmath227 , where @xmath65 is a kind of `` truncation index '' , which tells us the node number starting from which the upper tba equations are replaced by @xmath39 nlie variables ( see figure [ yhnlie ] . ) .the parameter @xmath227 is a free parameter in the asymptotic solution for the upper @xmath39 nlie variables ( [ bfirst]-[dvect ] ) and it enters the equations such that the asymptotic solution around which the equations are formulated contain this parameter . from this discussionit is obvious that @xmath34 is a physical parameter which means that the energy depends on it , while the other two parameters @xmath65 and @xmath227 correspond to different formulations of the same mathematical problem , so the energy does not depend on them .thus the choice of these parameters is in our hand and we tried to choose such values for them which allows us numerical convergence in the widest range in @xmath34 .for example the @xmath104 choice is the best for numerical purposes since due to ( [ bmmd ] ) a @xmath40 symmetry arises in the @xmath39 hnlie variables minimizing the number of unknowns in the problem .tuning @xmath65 might have two advantages .first , numerical experience shows that for large @xmath65 the chebyshev coefficients of the unknowns entering the formula ( [ omegakq ] ) for @xmath228 , decay faster , which allows for higher numerical precision .second also from numerics we learn that with @xmath65 fixed at certain values of @xmath34 non physical results are obtained from the numerical solution of the problem .this is a consequence of new local singularities entering the problem , but we still did not take them into account in the equations .these new singularities show up mostly in the @xmath39 nlie variables , thus by increasing the value of @xmath65 the appearance of such singularities in the equations can be postponed to higher values of @xmath34 .we solved numerically our equations for different values of @xmath12 and with various values of @xmath65 and @xmath227 , and in case the numerical result was physically acceptable for all @xmath65 and @xmath227 we tried , it was also independent of these parameters within the numerical errors of the method .so far we discussed the parameters of the continuous integral equations and their role in the numerical solution .now we turn to discuss the numerical parameters of the equations .the numerical parameters are artifacts of the numerical method , and they arise mostly from the discretization method described in section 3 .we note that there is no cutoff parameter in our numerical method , neither in the integration range nor in the index of @xmath14-functions .everything is treated in an exact manner , the only source of numerical errors is the discretization of the unknowns and the convolutions . herewe give the most used subinterval vectors of our numerical computations . on each subintervalwe used an @xmath229 order chebyshev approximation .the subinterval vector @xmath161 of @xmath230 $ ] is given by the empirical formula : @xmath231 where the vectors in components take the form : @xmath232 + 1 } , \qquad k=1, ... ,[2g]+1,\ ] ] @xmath233 with @xmath234 having vector components : @xmath235 } , \qquad j=1, ... ,\left[2 g-\frac32\right].\ ] ] here @xmath236 $ ] stands for integer part .the set of subinterval vectors @xmath160 of @xmath158 $ ] could also be given by an appropriate empirical formula , but it would take such a complicated form , that it is better to write down the requirements from which it can be constructed . ] .the requirements can be formulated in the language of the variable @xmath237 $ ] .the elements of the vector @xmath160 divide the interval @xmath158 $ ] into subintervals .our requirements constrain the allowed length of the subintervals with respect their location within the whole interval @xmath158 $ ] .the requirements are as follows : * the first element of @xmath160 is @xmath238 . *the length of subintervals @xmath239 in the range @xmath240 is approximately @xmath241 : @xmath242 . *the length of subintervals in the range @xmath243 is approximately @xmath238 : @xmath244 . *the length of subintervals in the range @xmath245 is approximately @xmath246 : @xmath247 . in practicethe length of the subintervals are slightly `` squeezed '' with respect to the conditions above to fill the full @xmath158 $ ] properly . ] .finally , we note that for checking the numerical precision , we also did numerical computations with @xmath248 keeping the subintervals fixed and also with keeping @xmath229 , but doubling the number of subinterval points . before turning to present the numerical results we would like to say a few words about the possible tests of the numerical results .namely , how one can recognize a wrong result .this is also a very important point of the numerical method , since there are a lot of equations with very complicated kernels and it is easy to make mistakes during writing the code of the numerical solution .there are three basic things that we can check from the numerical results .the first check is dictated by the energy equation ( [ eps ] ) .it is known that the energy starts at the first wrapping order ( i.e. @xmath249 ) and this first order correction is exactly given by the lscher formula @xcite : @xmath250 with @xmath251 given explicitly in ( [ yqo ] ) .this quantity can be computed numerically with any digits of precision , so its value is known exactly at any values of @xmath34 and @xmath12 . the lscher - formula ( [ luscher0 ] )corresponds to the linearized version of our equations ( [ tbamvw])-([lasteq ] ) , this is why solving the linearized set of equations ( which is the first step for the iterative solution ) we should reproduce the numerical evaluation of ( [ luscher0 ] ) .this is a nontrivial check on the kernels , on the discretization method and on the equations themselves as well .in addition since this test is quantitative it can tell some information also on the numerical precision of the method ) within certain digits of precision , than the deviation from the lscher result can be a good starting estimate to the numerical error .one can not expect better accuracy , but the precision will not become much worse either . ] .this test can signal problems on solving the linearized problem .the remaining two tests can signal some discrepancies during the solution of the nonlinear problem .the second testing condition is that from the numerical solution @xmath252 must be real .this sound trivial , but it is not trivial at all .if one takes a look at the equation ( [ yvwp0m2 ] ) of @xmath252 , one can recognize that there are complex quantities on the rhs which do not form conjugate pairs .so , the reality of the lhs is not guaranteed by the form of the equations , but it is guaranteed by the form of the solution .thus the second testing condition is expressed by the inequality : @xmath253 here we wrote the typical numerical errors we had during the computations .the third test is based on the approximation scheme we use .one must check whether the chebyshev coefficients of the unknowns decay as it is expected .from such a check the numerical precision of the method can be read off and it can shed light on some anomalous divergent behavior of the numerical solution . thus it can indicate possible errors in the elimination of the divergent @xmath254 terms in ( [ separ ] ) and ( [ b2bar],[d2bar ] ) . the numerical results for the @xmath226 case can be seen in figure [ al2 ] and table [ tl2 ] . in the tablewe show not only the energy @xmath255 at different values of the coupling @xmath34 , but the constant @xmath73 , as well .the other columns of the table are related to the solution of the linearized equations ; @xmath256 and @xmath257 are the energy and the global constant from the numerical solution of the linearized equations .@xmath258 stands for the deviation of @xmath256 from the exact lscher result .this quantity gives some information on the numerical accuracy of the method .finally the column `` number of nodes '' tells us the cutoff index of the lscher formula , which is necessary to get the lscher energy with the precision given by @xmath258 .this number is not equal to @xmath65 in our equations . for the @xmath226 state , in case of @xmath259 we used @xmath260 , for @xmath261 we used @xmath262 , and in the range @xmath263 we took @xmath264 .finally at @xmath265 we used @xmath266 to get acceptable numerical results . then beyond this pointwe could not save our equations from the entrance of new singularities by increasing the value of @xmath65 with a reasonable @xmath267 amount . because of this reason we could not get really close to the supposed critical point . there @xmath268 , but we could reach only @xmath269 at @xmath265 .apart from this very embarrassing fact , some important features can be read off from the numerical data .first of all it can be seen that in the range @xmath270 the energy is very slowly varying function of @xmath34 , so there is no sign of any divergent behavior .what is more interesting is the behavior of the global constant @xmath73 .it is negative and it decreases faster and faster as @xmath34 is increased . from the definition of @xmath73 ( [ separ ] )it follows that all @xmath157-functions are proportional to its exponent : @xmath271 .the fast decrease of @xmath73 indicates that though @xmath157 has worse and worse large @xmath49 asymptotic by the increase of @xmath34 , its global magnitude is actually decreasing .this remark can be understood from the tba formulation of the energy . @xmath272 close to the critical point @xmath255 is supposed to be finite @xcite @xmath273 , but naively the sum in the rhs of ( [ ebtba ] ) would diverge due to the large @xmath53 terms . since @xmath157 is small for large @xmath53 , in leading order . ]the @xmath274 replacement can be done : @xmath275 where @xmath276 is an arbitrary index cutoff scale and @xmath277 replacement was applied .since @xmath278 is @xmath53-independent all the dangerous @xmath53 dependence is still in @xmath279 . in ( [ ebtba1 ] ) approaching to the critical point the second sum starts to diverge , and the global multiplicative factor @xmath278 must tend to zero in order to ensure the finiteness of both sides of the equation . our numerical data seems to support this picture .namely @xmath280 as going closer and closer to the critical point .( 260,70 ) ( 0,15 ) ( 100,15 ) ( 0,0 ) .numerical data for the @xmath226 state . [ cols="^,^,^,^,^,^,^",options="header " , ]the @xmath0 ground state is mysterious , since so far the anomalous dimension of this state could not be determined even for small @xmath34 either from field theory or from integrability considerations @xcite .here we concentrate on the integrability side .there the boundary lscher formula @xcite diverges for this state @xcite . for generic @xmath12the lscher formula is simply the expansion of the tba energy formula around the asymptotic solution with the replacement : @xmath281 . for small couplingit takes the form @xcite : @xmath282 this small coupling expression diverges for @xmath0 , since this point sits exactly on the pole of the @xmath283 -function . as for the origin of this divergence; in ( [ luscher ] ) the individual integrals are convergent , but their sum for @xmath53 causes the divergence . in @xcite it was argued that also for any larger @xmath12 the tba energy formula would diverge beyond a certain critical value of the coupling : @xmath284 . assuming that the energy is a monotonously decreasing function of @xmath34 , which is supported by numerical results , this critical point can be expressed clearly in terms of the energy by the criterion : @xmath285 in @xcite this point was interpreted as a turning point where the energy becomes imaginary and as a physical consequence the ground state becomes tachyonic . for the @xmath0 statethe critical point is right at @xmath37 assuming that for small @xmath34 the energy is also small .now let us turn our attention to the hnlie description of the problem detailed in section 2 .here there are no infinite sums and even for @xmath0 all the convolutions of the integral equations seem to converge behavior of the unknown functions , which was used to derive the btba equations from discontinuity relations and y - system . ] .for the first sight there is no sign of any problem in the hnlie description and it seems that only the tba description is inappropriate to treat the @xmath0 case .but unfortunately this is not the case .we can write down the discretized integral equations for the @xmath0 case as well , and using the newton - method , we can solve them for small values of the coupling . ] .we always get some numerical solution for the discretized problem , but it turns out that the chebyshev coefficients of the unknowns , which correspond to the large @xmath49 subinterval do not form a decaying series .this phenomenon is a typical sign of some weak ( probably logarithmic ) large @xmath49 divergence of the unknowns .if one increases the number of subintervals and sampling points the situation remains the same .the conclusion is that we can solve the discretized problem , but the solution can not be interpreted as the discretely approximated version of the continuous solution of our integral equations . in other wordsthe continuous hnlie equations have no solution for @xmath0 . in order to get some analytical insight why the solutions become diverging at large @xmath49let us consider the tba formulation of the problem ( @xmath286 in hnlie ) .it is known @xcite that the tba energy comes from the coefficient of the most divergent @xmath287 term in the large @xmath49 expansion of @xmath45 : @xmath288 the @xmath255 term originates from the rhs of the tba equations for @xmath45 from the convolution term @xmath289 by exploiting the large @xmath49 expansion of the kernel : @xmath290 . @xmath291 has better large @xmath292 behavior than that of @xmath293 , since it behaves like @xmath294 . as a consequence contrary to the energy formula ,the sum of dressing convolutions is convergent indeed .thus one might think that for @xmath0 the problem emerges , because for the derivation of the energy formula we expanded the sum of dressing convolutions term by term for large @xmath49 .this is why instead of this usual procedure , we consider the sum of dressing convolutions itself , compute it and then at the end of the computation we take the large @xmath49 expansion .this procedure is carried out in the small coupling limit .we need the leading order small coupling expression of the dressing kernel in the mirror - mirror channel . ] : @xmath295 + .... \end{split}\ ] ] then the formula , the large @xmath49 expansion of which accounts for the small coupling expanded energy , is given by : @xmath296 where @xmath297 is the leading small coupling expression of ( [ yqo ] ) at @xmath0 .the second derivative of @xmath298 can be computed explicitly by simple fourier space technique .we take the fourier form of each functions under integration , the convolution is the product of the individual fourier transforms , the sum for @xmath292 can be easily done in fourier space and at the end of the process everything is transformed back to the @xmath49 space . in such a manner one gets a bulky , but explicit expression for @xmath299 , which we do not present here , only its large @xmath49 expansion : @xmath300 integrating twice the large @xmath49 expansion at small coupling becomes : @xmath301 from ( [ o ] ) it is obvious why the naive lscher energy formula diverged . because the leading order large @xmath49 term is not the expected @xmath302 , but @xmath303 .this is the key point of the problem , since in this case after this first iteration @xmath157 acquires an unwanted type of large @xmath49 term , which makes @xmath157 divergent for large @xmath49 : @xmath304 this large @xmath49 divergence contradicts to what was assumed about the large @xmath49 behavior of @xmath157 at the derivation of the integral equations , since it was supposed to decay . in this examplewe have shown in the small coupling limit , that during the iterative solution of the btba equations , @xmath45 acquires an extra @xmath305 behavior at infinity , which made @xmath157 an exploding function at infinity .this means that the iterative solution of the tba equations leaves the class of physically acceptable solutions .one might ask the question , whether it is possible to keep somehow the qualitative large @xmath49 behaviors that we assumed at the derivation of the equations ? here we sketch a possible idea for small coupling to the @xmath0 case . let us assume that we managed to modify the tba equations , such that all @xmath14-functions have the large @xmath49 behavior we want . since most of the tba equations reflect the structure of the @xmath14-system functional equations we expect to modify only those equations which are affected by also the discontinuity relations .it follows , that for large @xmath53 , the formula for the estimate for @xmath157 ( [ largeyq ] ) remains the same .now , we assume that for small @xmath34 the energy is also small and take the simultaneous small @xmath34 and small energy expansion of the rhs of the tba energy formula ( [ ebtba ] ) . inleading order the large @xmath53 terms will dominate : @xmath306 where @xmath307 denotes the large @xmath53 estimate ( [ largeyq ] ) of @xmath157 , @xmath308 as a consequence of the @xmath309 change of variables and the pole term in @xmath255 comes from the pole of the @xmath283-function . in our hnlie approachthe energy @xmath255 and the constant @xmath73 are parts of the equations which means that they are not simply expressed by explicit formulas based on the solution of the equations , but must me obtained by solving the set of non - trivially entangled equations . in this sense ( [ espec1 ] ) defines an equation for @xmath255 for small @xmath34 .its leading order solution is : @xmath310 if @xmath311 then @xmath255 becomes imaginary as it would be expected from string - theory expectations @xcite . to decide the sign of @xmath312 , the equation ( [ chybrid ] ) has to be analyzed in the context of the small @xmath34 and @xmath255 expansion .it turns out that @xmath312 is positive and @xmath313 for small @xmath34 , so according to ( [ eim1 ] ) @xmath255 is imaginary .another remarkable fact is that according to ( [ eim1 ] ) @xmath255 starts at @xmath314 instead of the @xmath315 prediction of the boundary lscher formula ( [ luscher ] ) .this might be another explanation why the coefficient of @xmath316 diverges in the lscher formula for the @xmath0 case .finally , we note that in the small @xmath34 and @xmath255 expansion of the @xmath0 state , the energy is pure imaginary only at leading order in @xmath34 , but in higher orders it acquires real part as well .for the first sight , it might seem that without modifying the equations one immediately gets imaginary energy when going through the critical point .but , the situation is a bit more subtle .there is a hidden tacit modification of the equations .this is realized in ( [ espec1 ] ) by the replacement : @xmath317 for the @xmath0 case it is an identity for @xmath318 , but for @xmath319 it is not an identity anymore , but a nontrivial analytical continuation in @xmath255 .such an analytical continuation would require the exact determination of complicated sums of convolutions of the tba equations as functions of the energy .since this does not seem to be feasible in practice , we give such an alternative modification of the tba equations which preserves the infinite sum structure of the equations , but the sums will converge everywhere for @xmath320 except at the critical value @xmath321 .the basic idea of the modification comes from the sum representations of the @xmath283-function .the usual one converges for @xmath322 : @xmath323 but there is another representation which converges for @xmath324 : @xmath325 then the original tba equations are modified through their infinite sums by the replacements : @xmath326 where @xmath327 . taking into account the large @xmath53 behavior of all @xmath157 functions and all the kernels of the infinite sums of the tba equations ,the new representation will converge for @xmath328 .this slight modification of the tba equations might make it possible to go beyond the critical point and get solution of the tba equations with large @xmath49 asymptotics being in accordance with the ones used for the derivation of the equations .the conclusion of this heuristic argument is that to keep the expected for large @xmath49 , while other @xmath14-functions tend to constant . ]qualitative large @xmath49 behavior a nontrivial modification of the tba equations must be carried out , which might lead to complex energies .in this paper we studied the ground state energy of a pair of open strings stretching between a coincident @xmath2-brane anti-@xmath2-brane pair in @xmath3 of @xmath4 .the main motivation for the study is that string - theory predicts that the ground state of such a configuration becomes tachyonic for large values of the t hooft coupling @xcite . in @xciteit was shown that the usual integrability based btba approach always give real energies for the ground state and it breaks down at latest when the energy gets close to the critical value : @xmath35 .this point was interpreted in @xcite as a transition point where the ground state becomes tachyonic .approaching this critical point the contribution of all the @xmath14-functions of the btba becomes quantitatively relevant , thus the numerical solution of the truncated btba equations can not give accurate results close to the critical point . to resolve this difficulty and get more accurate numerical results we transformed the previously proposed btba equations into finite component hnlie equations .the hnlie equations were solved at different values of @xmath34 and @xmath12 and the numerical results confirmed the earlier btba data . during the numerical solution of the hnlie equations the usual iterative methods failed to converge ,this is why we worked out two numerical methods to reach convergence .the most effective one is , if one transforms the integral equations into discrete nonlinear algebraic equations and solves them by newton - method .the power of this method is demonstrated by the fact that it gives convergent results even if the numerical solution is not physically acceptable .unfortunately , in our numerical studies we could not get very close to the critical point , because new singularities entered the hnlie equations taking into account of which would have required an enormous amount of additional work .nevertheless , in the range where we could get physically acceptable results , the precision of the hnlie data were higher than those of btba and the hnlie approach could give a deeper understanding of the problem .for the ground state of the @xmath0 state the critical point is right at @xmath37 and neither perturbative field theory computations nor the boundary lscher formula could provide a finite quantitative answer to the anomalous dimension . even in this special case the numerical solution of the hnlie equations was possible .the results showed that without an appropriate modification of the equations , they can not give physically acceptable results . in this case , it means that the solution of the dicretized problem can not be considered as a discretized solution of the continuous nonlinear integral equations .moreover the large rapidity behavior of the numerical solution is incompatible with the one assumed for the derivation of the equations .this phenomenon is analytically analyzed in the framework of btba and an idea is sketched to preserve the expected large rapidity behavior of the unknowns .this method is based on an appropriate modification of the tba equations which would lead to complex energies beyond the critical point .hopefully the @xmath0 case at @xmath37 could be treated analytically in the framework of the quantum spectral curve method @xcite , solving the mystery of this state in the context of integrability .the author thanks nadav drukker , lszl palla and jnos balog for useful discussions .this work was supported by the jnos bolyai research scholarship of the hungarian academy of sciences and otka k109312 .the author also would like to thank the support of an mta - lendlet grant , and the hungarian - french bilateral grant tt-12-fr-1 - 2013 - 0024 .finally , the author appreciates apctp for its hospitality where part of this work was done .throughout the paper we use the basic notations and tba kernels of ref .@xcite , which we summarize below . for any function @xmath41, we denote @xmath329 and in general @xmath330}(u)=f(u\pm \frac{i}{g } a)$ ] , where the relation between @xmath34 and the t hooft coupling @xmath32 is given by @xmath33 .most of the kernels and also the asymptotic solutions of the hnlie - system are expressed in terms of the function @xmath331 : @xmath332 which maps the @xmath49-plane with cuts @xmath333 \cup [ 2,\infty]$ ] onto the physical region of the mirror theory , and in terms of the function @xmath334 @xmath335 which maps the @xmath49-plane with the cut @xmath336 $ ] onto the physical region of the string theory .both functions satisfy the identity @xmath337 and they are related by the @xmath338 and @xmath339 relations on the lower and upper half planes of the complex plane respectively .the momentum @xmath340 and the energy @xmath341 of a mirror @xmath53-particle are expressed in terms of @xmath331 as follows : @xmath342 two different types of convolutions appear in the hnlie equations .these are : @xmath343 the kernels and kernel vectors entering the hnlie equations can be grouped into two sets .the kernels from the first group are functions of only the difference of the rapidities , thus actually they depend on a single variable .the other group of kernels composed of those , which are not of difference type .we start with listing the kernels depending on a single variable : @xmath344\ , , \nonumber \\k_q ( u ) & = \frac{1}{2\pi i } \ , \frac{d}{du } \ , \log s_q(u ) = \frac{1}{\pi } \ , \frac{g\ , q}{q^2 + g^2 u^2}\,,\quad s_q(u)= \frac{u - \frac{iq}{g}}{u + \frac{i q}{g } } \ , , \nonumber\\ k_{mn}(u ) & = \frac{1}{2\pi i } \ , \frac{d}{du } \ , \log s_{mn}(u)=k_{m+n}(u)+k_{n - m}(u)+2\sum_{j=1}^{m-1}k_{n - m+2j}(u)\,,\nonumber\\ s_{mn}(u ) & = s_{m+n}(u)s_{n - m}(u)\prod_{j=1}^{m-1}s_{n - m+2j}(u)^2 = s_{nm}(u)\ , .\label{skq}\end{aligned}\ ] ] the fundamental building block of kernels which are not of difference type is : @xmath345 using the kernels @xmath346 and @xmath347 it is possible to define a series of kernels which are connected to the fermionic @xmath83-functions .they are : @xmath348 and @xmath349 further important kernels entering the @xmath83 related tba - type equations are defined as follows : @xmath350 the kernels entering the right hand sides of the equation ( [ hybrid ] ) for @xmath63 are @xmath351 and the dressing - phase related kernel @xmath352 , which is built from the @xmath353 s - matrix of the model @xcite .it is of the form @xmath354 where @xmath355 is the improved dressing factor @xcite .the corresponding @xmath353 and dressing kernels are defined in the usual way @xmath356 explicit expressions for the improved dressing factors @xmath357 can be found in section 6 of ref .@xcite . here for our numerical computations we used the single integral representation given in @xcite .finally we mention that along the lines of @xcite in the derivation of the formula ( [ omegakq ] ) for @xmath66 , it was exploited that all the necessary kernels : @xmath358 satisfy the identity : @xmath359in this appendix the kernel matrices appearing in the upper hnlie part of our equations ( [ ba],[da ] ) are presented . herethe kernel matrices are different compared to those published in @xcite .the difference comes simply from a reformulation the equations in the language of new unknown functions . in @xcitethe unknowns are 6 @xmath91-type functions : @xmath360}_{1,s},b^{(3)[\gamma_2]}_{2,s},b^{(3)[\gamma_3]}_{3,s } , b^{(2)[-1+\gamma_4]}_{1,s},b^{(2)[-1+\gamma_5]}_{2,s},b^{(1)[-2+\gamma_6]}_{1,s } \},\ ] ] and 6 @xmath105-type functions : @xmath361}_{1,s},d^{(3)[\eta_2]}_{2,s},d^{(3)[\eta_3]}_{3,s } , d^{(2)[\eta_4]}_{1,s},d^{(2)[\eta_5]}_{2,s},d^{(1)[\eta_6]}_{1,s } \},\ ] ] with shift vectors @xmath362 and @xmath363 given by ( [ vgamma],[veta ] ) .we recognized that the kernels become simpler if we formulate the equations in terms of the unknowns : @xmath364}_{1,s},\eta / b^{(3)[\gamma_2]}_{2,s},b^{(3)[\gamma_3]}_{3,s } , \eta / b^{(2)[-1+\gamma_4]}_{1,s},b^{(2)[-1+\gamma_5]}_{2,s},b^{(1)[-2+\gamma_6]}_{1,s } \ } , \label{buj}\ ] ] and @xmath365}_{1,s},\eta / d^{(3)[\eta_2]}_{2,s},d^{(3)[\eta_3]}_{3,s } , d^{(2)[\eta_4]}_{1,s},\eta / d^{(2)[\eta_5]}_{2,s},d^{(1)[\eta_6]}_{1,s } \ } , \label{duj}\ ] ] where @xmath128 is a global sign factor and @xmath366 , if one adopts the notation of @xcite for the hnlie equations ( [ ba],[da ] ) .another advantage of using the variables ( [ buj],[duj ] ) is that they are either @xmath313 or exponentially small for large volumes . for the sake of simplicity, here we give the form of the kernels of ( [ ba],[da ] ) before the application of the contour shifts ( [ vgamma],[veta ] ) .the kernels of the equations can be obtained from these by simply shifting their arguments according to the formulas below : @xmath367 the kernel matrices can be expressed by the functions as follows . ] : @xmath368 and they take the form : @xmath369 @xmath370 @xmath371 @xmath372in this section along the lines of @xcite the asymptotic solutions of the upper @xmath39 nlie variables are presented . in the asymptotic limit the t - hook of ads / cft splits into two @xmath373 fat - hooks .the basic building blocks of the asymptotic solution are the nine @xmath53-functions corresponding to the left and right @xmath373 fat - hooks . due to the left - right symmetry of the @xmath14-systemit is enough to give the right @xmath53-functions .they can be derived from the asymptotic solution of the y - functions given in @xcite .they take the form : @xmath374 where @xmath375 and @xmath376 are arbitrary constants which cancel from the final form of the asymptotic nlie variables .furthermore @xmath377 to satisfy the recursion @xmath378 .the further building blocks of the asymptotic solution are as follows : @xmath379}},\ ] ] and @xmath380 the solution of the recursions @xmath381 are as follows : @xmath382 where @xmath383 and @xmath227 are arbitrary constants . using the building blocks listed above , the asymptotic form of the upper @xmath39 nlie functions can be determined @xcite and take the form : @xmath384 } } , \label{bfirst}\\ b^{(3)o}_{1,s}(u)&=&b^{(3)o}_{3,s}(u)=\frac{(s+1 ) \ , u}{u^{[-s]}},\\ b^{(3)o}_{2,s}(u)&=&b^{(3)o}_{2,s}(u)=-\phi^{[-s]}(u)\,\frac{1}{4\ , g\ , s\ , u},\\ b^{(2)o-}_{1,s}(u)&=&b^{(2)o-}_{1,s}(u)=\frac{\phi^{[-s]}(u)}{4\ , g\ , s\ , u+c},\\ b^{(2)o-}_{2,s}(u)&=&-\frac{c+4\ , g\ , s\ , u^-}{4\ , g\ , u^{[-s]}},\qquad\qquad b^{(2)o-}_{2,s}(u)=-\frac{c+4 \ , g\ , ( s-1 ) \ , u}{4\ , g\ , u^{[-s]}},\\ b^{(1)o--}_{1,s}(u)&=&\frac{c+4 \ ,g\ , ( s-1 ) \, u}{4\ , g\ , u^{[-s]}},\qquad\qquad b^{(1)o--}_{1,s}(u)=\frac{c+4\ , g\ , s\ , u^-}{4\ , g\ , u^{[-s]}},\end{aligned}\ ] ] @xmath385}},\\ d^{(3)o}_{1,s}(u)&=&d^{(3)o}_{3,s}(u)=\frac{(s+1 ) \ , u}{u^{[s]}},\\ d^{(3)o}_{2,s}(u)&=&d^{(3)o}_{2,s}(u)=-\frac{1}{\phi^{[s]}(u)}\,\frac{(g\ , u^{[s]})^2}{4\ , g\ , s\ , u},\\ d^{(2)o}_{1,s}(u)&=&-\frac{c+4\ , g\ , s\ , u^+}{4\ , g\ , u^{[s]}},\qquad\qquad d^{(2)o}_{1,s}(u)=-\frac{c+4 \ , g\ , ( s-1 ) \ , u}{4\ , g\ , u^{[s]}},\\ d^{(2)o}_{2,s}(u)&=&d^{(2)o}_{2,s}(u)=-\frac{\phi^{[s]}(u ) \ , ( g \,u^{[s]})^2}{4\ , g\ , s\ , u+c},\\ d^{(1)o}_{1,s}(u)&=&\frac{c+4 \ , g\ , ( s-1 ) \ ,u}{4\ , g\ , u^{[s]}},\qquad\qquad d^{(1)o}_{1,s}(u)=\frac{c+4\ , g\ , s\ , u^+}{4\ , g\ , u^{[s]}},\end{aligned}\ ] ] where @xmath386 is the `` cutoff index '' where the tba @xmath387 hnlie replacements starts , here the notation @xmath386 is kept to fit to formulas of @xcite . ] , furthermore for any index distribution @xmath388 and @xmath389 stand for @xmath390 and @xmath391 respectively . @xmath392 and @xmath227 is the arbitrary constant that does not cancel from the formula for the hnlie variables. the asymptotic solution for the six @xmath91- and @xmath105-type nlie - functions of the system can be obtained from the bcklund functions above by appropriately shifting their arguments : @xmath393}_{1,s},\eta / b^{(3)o[\gamma_2]}_{2,s},b^{(3)o[\gamma_3]}_{3,s } , \eta / b^{(2)o[-1+\gamma_4]}_{1,s } , b^{(2)o[-1+\gamma_5]}_{2,s } , b^{(1)o[-2+\gamma_6]}_{1,s}\ } , \label{bvect}\ ] ] @xmath394}_{1,s},\eta / d^{(3)o[\eta_2]}_{2,s},d^{(3)o[\eta_3]}_{3,s } , d^{(2)o[\eta_4]}_{1,s},d^{(2)o[\eta_5]}_{2,s},d^{(1)o[\eta_6]}_{1,s}\ } , \label{dvect}\ ] ] with the shifts given in ( [ vgamma],[veta ] ) .finally we note that the @xmath104 choice implies a symmetry relation between the @xmath91- and @xmath105-type variables .let @xmath103 the @xmath395 by @xmath395 matrix : @xmath396 then at @xmath104 the @xmath397 and @xmath398 vectors satisfy the relations as follows : @xmath399 @xmath400 where @xmath401 denotes complex conjugation . in our numerical studieswe mostly use the @xmath104 asymptotic solution to setup the equations to solve . in this casethe exact equations guarantee the fulfillment of ( [ bmmd ] ) , which reduces to @xmath395 the number of independent complex functions of the upper nlie part .this appendix is devoted to give the asymptotic solution for the @xmath14-functions and the variables of the horizontal @xmath38 nlie .the asymptotic form of the @xmath14-functions can be read off from the asymptotic @xmath95-functions in @xcite .they take the form : @xmath402 @xmath403}(u)\ , x_s^{[-q]}(u)}\right)^{2l } \frac{16 \ , q^2 \ , g^2 \ ,u^2}{g^2\ , u^2 + q^2 } , \qquad q=1,2, ... \label{yqo}\ ] ] @xmath404 @xmath405 following the lines of @xcite the asymptotic horizontal @xmath38 nlie variables can be determined from the asymptotic @xmath53-functions ( [ constantq ] ) .here we just list the final formulas : @xmath406 where @xmath407 @xmath408 and @xmath89 is the arbitrary contour shift parameter of the horizontal @xmath38 nlie .99 j. m. maldacena , _ `` the large n limit of superconformal field theories and supergravity , '' _ _ adv .* 2 * ( 1998 ) 231 , [ _ int .j. theor . phys . _* 38 * ( 1999 ) 1113 ] ; s. s. gubser , i. r. klebanov and a. m. polyakov , _ `` gauge theory correlators from non - critical string theory , '' _ _ phys .b _ * 428 * ( 1998 ) 105 ; e. witten , _ `` anti - de sitter space and holography , '' _ _ adv . theor . math . phys . _ * 2 * ( 1998 ) 253 ; z. bajnok , n. drukker , .hegeds , r. i. nepomechie , l. palla , c. sieg , r. suzuki , _ `` the spectrum of tachyons in ads / cft '' _ _ jhep _ * 1403 * ( 2014 ) 055 .n. beisert , c. ahn , l. f. alday , z. bajnok , j. m. drummond , l. freyhult , n. gromov and r. a. janik _ et al ._ , _ `` review of ads / cft integrability : an overview , '' _ _ lett ._ * 99 * ( 2012 ) 3 .o. dewolfe and n. mann , _ integrable open spin chains in defect conformal field theory _ , _ jhep _ * 0404 * ( 2004 ) 035 , d. berenstein and s. e. vazquez , _integrable open spin chains from giant gravitons _ ,_ jhep _ * 0506 * ( 2005 ) 059 , d. m. hofman and j. m. maldacena , _ reflecting magnons _ , _ jhep _ * 0711 * ( 2007 ) 063 , d. correa and c. young , _ reflecting magnons from d7 and d5 branes _ ,_ j.phys.a _ * a41 * ( 2008 ) 455401 , d. h. correa , v. regelskis , and c. a. young , _ integrable achiral d5-brane reflections and asymptotic bethe equations _ , _ j.phys.a _ * a44 * ( 2011 ) 325403 , z. bajnok and r. a. janik , _ `` six and seven loop konishi from luscher corrections , '' _ _ jhep _ * 1211 * ( 2012 ) 002 s. leurent , d. serban and d. volin , _`` six - loop konishi anomalous dimension from the y - system , '' _ phys .lett . * 109 * ( 2012 ) 241601 s. leurent and d. volin , _ `` multiple zeta functions and double wrapping in planar @xmath409 sym , '' _ nucl . phys .b * 875 * ( 2013 ) 757 z. bajnok , m. kim and l. palla , _`` spectral curve for open strings attached to the y=0 brane , '' _ _ jhep _ * 1404 * ( 2014 ) 035 n. gromov , f. levkovich - maslyuk , g. sizov and s. valatka , _`` quantum spectral curve at work : from small spin to strong coupling in @xmath411 = 4 sym , '' _ jhep * 1407 * ( 2014 ) 156 n. gromov , v. kazakov and p. vieira , _`` exact spectrum of planar @xmath412 supersymmetric yang - mills theory : konishi dimension at any coupling , '' _ phys .* 104 * ( 2010 ) 211601 j. mcgreevy , l. susskind , and n. toumbas , _ invasion of the giant gravitons from anti - de sitter space _ , _ jhep _ * 0006 * ( 2000 ) 008 , v. balasubramanian , m. berkooz , a. naqvi , and m. j. strassler , _ giant gravitons in conformal field theory _, _ jhep _ * 0204 * ( 2002 ) 034 , v. balasubramanian , m .- x . huang , t. s. levi , and a. naqvi , _ open strings from @xmath412 superyang - mills _, _ jhep _ * 0208 * ( 2002 ) 037 , a. leclair , g. mussardo , h. saleur , and s. skorik , _ boundary energy and boundary states in integrable quantum field theories _ , _ nucl.phys ._ * b453 * ( 1995 ) 581618 , d. correa and c. young , _ finite size corrections for open strings / open chains in planar @xmath413/cft _ , _ jhep _ * 0908 * ( 2009 ) 097 , l. palla , _ yangian symmetry of boundary scattering in ads / cft and the explicit form of bound state reflection matrices _ , _ jhep _ * 1103 * ( 2011 ) 110 , c. ahn and r. i. nepomechie , _ yangian symmetry and bound states in ads / cft boundary scattering _ , _ jhep _ * 1005 * ( 2010 ) 016 , n. drukker , _ integrable wilson loops _ , _ jhep _ * 1310 * ( 2013 ) 135 , d. correa , j. maldacena , and a. sever , _ the quark anti - quark potential and the cusp anomalous dimension from a tba equation _, _ jhep _ * 1208 * ( 2012 ) 134 , z. bajnok , j. balog , d. h. correa , .hegeds , f. i. schaposnik massolo and g. zsolt tth , _`` reformulating the tba equations for the quark anti - quark potential and their two loop expansion , '' _ _ jhep _ * 1403 * ( 2014 ) 056 , n. gromov , v. kazakov and p. vieira , _ `` integrability for the full spectrum of planar ads / cft , ''_ _ phys.rev.lett . _ * 103:131601*,2009 , d. bombardelli , d. fioravanti and r. tateo , _`` thermodynamic bethe ansatz for planar ads / cft : a proposal , '' _ j. phys. a * 42 * ( 2009 ) 375401 g. arutyunov and s. frolov , _`` thermodynamic bethe ansatz for the @xmath4 mirror model , '' _ jhep * 0905 * ( 2009 ) 068 a. cavaglia , d. fioravanti , and r. tateo , _ extended y - system for the @xmath1 correspondence __ nucl.phys ._ * b843 * ( 2011 ) 302343 , j. balog and a. hegedus , _ `` @xmath414 mirror tba equations from y - system and discontinuity relations , '' _ jhep * 1108 * ( 2011 ) 095 j. balog , .hegeds , _`` hybrid - nlie for the ads / cft spectral problem '' _ _ jhep _ * 1208 * ( 2012 ) 022 .z. bajnok and l. palla , _ boundary finite size corrections for multiparticle states and planar @xmath413/cft _ , _ jhep _ * 1101 * ( 2011 ) 011 , g. arutyunov , s. frolov , r. suzuki , _`` string hypothesis for the @xmath4 mirror '' _ _ jhep_*03 * ( 2009 ) 152 . c. destri and h. j. de vega , _ `` nonlinear integral equation and excited states scaling functions in the sine - gordon model , '' _ _ nucl . phys .b _ * 504 * ( 1997 ) 621 - 644 a. hegedus , _ `` finite size effects and 2-string deviations in the spin-1 xxz chains , '' _ _j. phys .a _ * 40 * ( 2007 ) 12007 n. gromov , v. kazakov , s. leurent and d. volin , _ `` solving the ads / cft y - system , '' _ jhep * 1207 * ( 2012 ) 023 n. gromov , v. kazakov , s. leurent and d. volin , _ `` quantum spectral curve for planar @xmath415 super - yang - mills theory , '' _ phys . rev .* 112 * ( 2014 ) 1 , 011602 n. gromov , v. kazakov , s. leurent , d. volin , _ quantum spectral curve for arbitrary state / operator in ads@xmath416/cft@xmath417 _ http://arxiv.org/abs/1405.4857 [ ] .

<S> in this paper the hybrid - nlie approach of @xcite is extended to the ground state of a d - brane anti - d - brane system in ads / cft . </S><S> the hybrid - nlie equations presented in the paper are finite component alternatives of the previously proposed tba equations and they admit an appropriate framework for the numerical investigation of the ground state of the problem . </S><S> straightforward numerical iterative methods fail to converge , thus new numerical methods are worked out to solve the equations . </S><S> our numerical data confirm the previous tba data . in view of the numerical results </S><S> the mysterious @xmath0 case is also commented in the paper . * extensive numerical study of a d - brane , anti - d - brane system in @xmath1 * rpd hegeds wigner research centre for physics , + h-1525 budapest 114 , p.o.b . </S><S> 49 , hungary + </S>

cond-mat9803234

driven interfaces in random media present several features of interest , both with regard to the morphology of the moving interface as well as the dynamics of the growth process .experiments have been performed on several sorts of systems , ranging from fluid flow in porous media @xcite to propagation of burning fronts @xcite .these indicate that the disorder in the medium affects the properties of the interface in a crucial way .in particular , the large - distance scaling properties differ considerably from those of interfaces in uniform media . in both theoretical and experimental investigations, it is customary to characterize the spatial structure of the interface by its _ roughness_. the main point of this paper is to show that there is sometimes an unusual sort of pattern formation @xcite in the system , which results in the interface acquiring a _ time - averaged shape_. in such situations , this pattern provides an alternative characterization of interface morphology . a customary measure of the roughness is provided by the exponent @xmath0 , defined by @xmath1 , where @xmath2 is the root mean squared width of the interface and @xmath3 is the size of the system .the experiments mentioned above , and others similar to these , report an anomalously large value of @xmath0 large compared to the predictions of existing theories for interface growth in nonrandom media @xcite .it is recognised that the quenched nature of the disorder in the medium is responsible for this difference in scaling properties of the interface .unlike thermal noise , which varies rapidly , a portion of the interface subject to a quenched - noise environment continues to experience the same forces until the growth process takes it forward to a new region .the pinning effect of quenched noise has a strong effect on the large scale properties of the interface .several theoretical models have been put forward to try to account for the effect of quenched disorder on the properties of the interface ; @xcite gives an account of some of the early work and the relationship to other problems involving pinning , while @xcite is a a recent review . amongst various proposals put forward to explain anomalous roughening are theories based on continuum equations with quenched disorder @xcite , the inclusion of noise with power - law amplitude @xcite or long - ranged correlation @xcite , power - law distributions of pinning - center strengths @xcite , as well as a class of models with microscopic rules based on directed percolation @xcite and models which relate the large - scale structure to the wetting properties of the invading fluid @xcite .a number of these models base the explanation of anomalous roughening on the phenomenon of critical depinning , which is relevant to an interface just at the threshold of motion . in the opposite limit of large velocity, the interface encounters the disorder at any site only for a short time , suggesting that the quenched nature of disorder is not important in this limit , and the interface behaves much as in a nonrandom medium @xcite .however this is not true at low velocities of the interface near the depinning threshold .in particular , the limit of zero velocity is thought to be a dynamical critical point @xcite where the scaling properties of the interface are strongly affected by the disorder . in a certain class of models ,the quenched disorder enters as barriers of random strengths which impede interface motion .the formation of infinitely long directed percolating paths of these barriers is of special significance , as such paths can block the entire interface effectively @xcite .the model proposed by sneppen @xcite involves `` extremal '' dynamics : at each time step , the interface advances only along the weakest of the barriers .extremal models were first introduced much earlier , in the context of invasion percolation for two - phase fluid flow in porous media , with a non - wetting fluid displacing a wetting one @xcite .the predictions of this invasion percolation model were borne out by experiments @xcite .dynamical correlation functions involving the center of activity in invasion percolation obey scaling @xcite , and the process defines a self - organised critical phenomenon @xcite ; the interface organises itself to align along critical paths bordered by large barriers , without the necessity of tuning any external parameters .the model proposed by sneppen @xcite is a modification of the invasion percolation model , incorporating surface tension effects which prevent very strong local convolutions of the interface , resulting in a self - affine , rather than self - similar , geometry .an extremal model close to the sneppen model was shown to result @xcite from a model of wetting - fluid invasion , by considering an interface advancing by merging meniscus arcs between adjacent pairs of pinning centers .an interesting feature of extremal dynamics is that it induces strong spatial correlation in sites at which growth occurs @xcite at successive time steps . in this paperwe introduce and study a variant of the sneppen model and show that the interface develops an interesting time - averaged structure as a result of correlations .the defining equation for the structure is @xmath4 \label{eq : pattern}\ ] ] where @xmath5 is the position of the active site at time @xmath6 , @xmath7 denotes the height at the site @xmath8 , @xmath9 is a time average in the steady state , and @xmath10 is the overall slope of the interface .the unusual point is that this structure @xmath11 is not fixed in space , but moves with its center always at the growth site , which itself follows an erratic path .the moving origin is crucial to the definition , as time averages performed at a fixed point in space reveal no structure at all .this average structure defines a pattern , and we study its formation numerically and analytically .figure 1a depicts schematically the activity - centered pattern in height gradients with respect to the moving origin , for the untilted ( @xmath12 ) interface .as can be seen , the interface develops an overall shape which is described by the height - gradient profile @xmath11 .the tail of the pattern falls slowly at large distances , as a power law .the corresponding height profile of the interface @xmath13 with respect to the active site is represented in fig .the nature of the pattern is sensitive to tilt , and figure 2 shows the height - gradient pattern and height profile for a tilted interface .we propose that this pattern is a simple way of characterising a new aspect of the morphology of the interface .traditional ways of characterising the morphology involve , as has already been mentioned , determining the roughness exponent @xmath0 .however such a definition does not hone in on the overall _ shape _ of the interface . in situations such asthe one considered in this paper when the interface does develop a nontrivial structure , the pattern is a useful quantitative characterization .an important point about the time - averaged pattern is that eq .1 defines a one - point correlation function . as such, it would be expected to strongly influence the properties of customary two - point correlation functions .we verify this by numerically studying two - point correlations .further we find that this sort of pattern formation is not restricted only to the sneppen model , but also occurs in other extremal models , such as the zaitsev model for low - temperature creep @xcite and the bak - sneppen model @xcite of biological evolution , albeit in other quantities .the plan of this paper is as follows . in sectionii we introduce our model and discuss the connection with the problem of directed percolation , well established from earlier studies . in section iii ,we discuss the correlations in the location of successive growth sites , a concept central to this paper because of its connection with pattern formation . in sectionsiv and v we define the pattern and present numerical results as well as an integral equation which provides an understanding of this sort of pattern formation .we define a model for which the equation is exact and discuss how the approximation can be improved . in sectionvi , we discuss the issue of temporal correlations with a view to seeing how they affect the pattern . in section vii, we present our results for an ordinary two - point correlation function in our model and show that activity - centered pattern formation has to be taken into account in order to understand some features in it . section viii deals with pattern formation in other extremal models and we conclude with a summary of our results in section ix .the extremal - model description of fluid - fluid interfaces in porous media is valid when the wetting is dominated by capillary forces , and thermal fluctuations are not important . in the extremal model ,the interface advances along the weakest barrier just ahead of it .the appealing feature of the model is that it is self - organised critical ; the dynamics , which involves searches for the global minimum at every step , automatically tunes the interface to a critical state at the depinning transition , without the necessity of fixing any external parameter . in the model proposed by sneppen @xcite, the random medium is modelled by a square lattice in which the sites are assigned random numbers @xmath14 $ ] .the random numbers could signify , for instance , the pore sizes in a porous medium .the interface is a directed path on this lattice and grows only at that perimeter bond with the smallest value of the random number ; after every such move , a local rearrangement process @xcite ensures the absence of very large slopes .extremal dynamics has also been proposed to describe very different situations for instance , the phenomenon of low - temperature dislocation creep @xcite , crack propagation @xcite and biological evolution @xcite .we will see in section viii that the sort of pattern formation we find in the interface model occurs in these models as well .we study a modified version of the sneppen model in this paper . in this version , hereafter referred to as the extremal bond model ( ebm ) , the interface is taken to be a directed path on a square lattice ( fig .3 ) , with tilted cylindrical boundary conditions @xcite which ensure that the mean slope is preserved . to every bond @xmath15 on the latticeis pre - assigned a fixed random number @xmath16 drawn from the interval [ 0,1 ] .the interface grows at that bond , in front of the interface , which carries the smallest random number .the local growth rules are the following ; if the chosen minimal bond has a positive ( negative ) slope , the sequence of links with negative ( positive ) slope just below ( on the left ) also advances , as illustrated in fig .this preserves the length of the interface as would happen in situations with very high surface tension .the local dynamics of interface adjustment is similar to that for the low - noise toom interface in @xcite ; the models are different in that the growth site is picked by the extremal rule in our case , while it is picked stochastically in the toom interface model .the ebm differs from the sneppen model in that the length of the interface is a strict constant of the motion and the @xmath17 s are associated with bonds rather than sites .while this modification does not change the values of any of the large - distance scaling properties of the sneppen model , it has a few advantages .the interface aligns along directed spanning paths in a percolation problem , just as in the sneppen model . in our case , the corresponding percolation problem is the diode - resistor percolation problem @xcite . on the square lattice , it is dual to the directed bond percolation problem @xcite which is relatively well - studied .another advantage is that the problem of interface growth in the ebm is conceptually simplified by the existence of a known one - to - one correspondence between the growing interface and a system of hard - core particles moving on a ring .the two - dimensional problem hence reduces to an effectively one - dimensional one .this also facilitates the numerics .the correspondence between the interface and the hard - core particles is detailed below .positive slope links of the interface are represented by particles ( @xmath18 ) and negative - slope links by holes ( @xmath19 ) ; see fig .the difference in height of the interface between sites @xmath20 and @xmath21 is given by @xmath22 .in front of each link of the interface is a bond with a random number @xmath23 assigned to it .correspondingly , the site @xmath24 with the particle ( hole ) representing this link carries a random number @xmath25 . just as for the interface , at each time step , activity is initiated at the site with the minimum @xmath25 .the update rules for the interface translate to the following dynamics for particles and holes .if the site with minimum @xmath25 contains a particle ( hole ) , it exchanges with the first hole ( particle ) to the left ( right ) .all sites hopped over , including the two which exchange the particle and hole , are refreshed by assigning a new set of @xmath25 s .this corresponds to the fact that the updated portion of the interface moves ahead and meets a fresh set of @xmath23 s on the square lattice .because the number of positive slope links ( and hence also the number of negative slope links ) is conserved for the interface , in the particle - hole terminology , this implies that the number of particles is conserved .hence we can define a density @xmath26 for particles on the one - dimensional lattice .this density determines the mean slope @xmath27 of the interface .an untilted interface corresponds to half - filling .the reference direction for determining tilt is the easy direction of directed bond percolation on the square lattice , which is along the @xmath28 line .tilt refers to any density away from @xmath29 , which implies a slope different from @xmath28 .the interface advances in a direction perpendicular to the direction of tilt and this translates to a nonzero current of particles on the ring .extremal models of interface depinning make use of a correpondence to the diode resistor percolation ( drp ) and directed percolation ( dp ) problems to predict various properties of the interface . in view of this , it is useful to recall some facts about the dp and drp processes . in the directed percolation problem ,bonds on a lattice are occupied with probability @xmath30 . atsome critical value @xmath31 an infinite directed path of occupied bonds ( in which every step is taken rightward or upward ) first forms along a definite direction ; on a @xmath32 square lattice , this is along the @xmath28 direction . for @xmath33 ,the network of these infinite paths forms an infinite connected cluster . for directed bond percolation on a square lattice the value of @xmath31is known to be @xmath34 @xcite .there are two distinct correlation lengths , @xmath35 along the easy direction and @xmath36 transverse to it , both of which diverge as @xmath37 : @xmath38 respectively .the values of these exponents are known to be @xmath39 and @xmath40 @xcite .suppose we have a single source point , and we ask which portion of the plane can be reached from it via occupied directed bonds . for @xmath41 this region is contained within a cone with opening angle @xmath42 where @xmath10 is the slope of the edge of the cone with respect to the @xmath28 direction ; the opening angle depends on @xmath30 .this relation can be inverted to find the critical probability @xmath43 _ viz . _the probability at which a connection first appears along the direction with slope @xmath44 .correlation lengths along and perpendicular to this direction have exponents @xmath45 and @xmath46 @xcite .we refer to the direction along @xmath47 as untilted ( @xmath12 ) ; any other slope is referred to as tilted . in the diode - resistor percolation problem, every bond is occupied by a `` diode '' ( a one - way connection ) with a probability @xmath48 or a `` resistor '' ( a two - way connection ) with a probability @xmath49 . on a square lattice ,the diodes all point up or right .let us ask which regions of the plane are connected to a given source point . if @xmath50 , a source point can reach the entire quadrant of which it is the left corner .as @xmath30 decreases from @xmath51 to @xmath31 , the opening angle @xmath52 of the connected region increases from @xmath53 to @xmath54 ; beyond this , the entire plane can be reached from the source point .the edge of the connected region is bordered by diodes pointing rightward and upward , which prevent it from spreading leftward and downward ( fig .4 ) . on the @xmath32 square lattice ,dp and drp are dual to each other @xcite .the dual to a drp configuration is constructed using the following rules .a diode in the drp lattice is crossed by a diode in the dual lattice , whereas a resistor is crossed by an insulator ( no connection ) in the dual lattice .thus we recover the directed percolation problem on the dual lattice .the opening angle of the cones in the two problems are related by @xmath55 . in the ebm, the random medium is modelled by considering a square lattice with every bond assigned a random number @xmath16 drawn from the interval @xmath56 $ ] .for a certain trial value @xmath57 , imagine occupying all bonds with @xmath58 by resistors , and the rest with diodes .we thereby generate a drp configuration with @xmath59 .when @xmath57 takes on the value @xmath60 , an infinite connected path of diodes is formed .such a path is called a ` stopper ' , and is significant for the dynamics of the ebm , as a moving interface with no overall tilt will align with such stoppers from time to time @xcite . when the interface aligns along a stopper ,all the bonds in front of it are larger than @xmath61 ( fig .5a ) . similarly a tilted interface with slope@xmath10 aligns along the edge of the cone with the same slope and all the bonds in front of it are expected to have a value larger than @xmath62 . since @xmath63 for @xmath64 , @xmath65 ( fig .even when the interface is evolving between two stoppers , only a small fraction of its overall length actually is in between ; the rest is still aligned with a stopper .the non - aligned fraction is expected to vanish in the thermodynamic limit .these expectations are confirmed by numerical studies of the ebm .as can be seen from the figures , the bonds in front of the interface are all mostly larger than a threshold value .consider an interface of slope @xmath10 aligned along a critical drp path of the same slope .it then moves forward by puncturing the path at the site with the least value of @xmath23 , which for an infinite system is exactly @xmath66 . on piercing through, a portion of the interface grows and fills out a loop of the infinite cluster while the rest of it remains pinned .however the interface motion within the loop is far from uniform .just as the critical cluster at @xmath43 impedes the growth of the interface on length scales of the order of the size of the system , near - critical clusters impede its motion at length scales of the order of but smaller than the loop size .one can think of these clusters as forming a finer network of connections within the network formed by the critical cluster at @xmath43 .while the interface is filling out a loop of the critical cluster , it encounters this finer mesh and as a result its motion is impeded temporarily . in what follows, we refer to these near - critical connections as ` sub - stoppers ' .a substopper can be characterized by the lowest value of @xmath16 on it , say @xmath67 , and also by the typical length scale @xmath68 over which it provides for effective pinning of the interface .these are related through @xmath69 .the above description of the evolution of the interface , contained as it is by networks of substoppers and stoppers , makes it clear that there are strong correlations between successive points of growth or forward motion .these correlations extend from small length scales up to scales of the order of the system size .figure 6 shows the plot of the location of the active site for 10,000 time steps for both the tilted and untilted cases .it can be seen that there are jumps on all scales in the active site position .the figure corroborates the description of interface motion given above .region ( 1 ) is a typical instance of the interface filling out a loop of size @xmath68 .it shows that there are jumps of all sizes upto a length @xmath68 , bearing out the substopper picture .region ( 2 ) on the other hand marks an instance when the interface has aligned along a stopper @xmath43 and hence there are jumps of all sizes upto the system size .more quantitatively , a measure of this long - ranged motion of the active site is the probability distribution @xmath70 that two consecutive locations of the active site are a distance @xmath68 apart .figure 7 shows @xmath70 for both the tilted and the untilted interface . in both cases , @xmath70 decays as a power for large @xmath68 : @xmath71 . in the untilted ( @xmath72 ) case, @xmath70 is a purely symmetric function because of @xmath73 symmetry .we find @xmath74 , which compares well with earlier determined values of @xmath54 for the sneppen model @xcite . in the tilted case ( @xmath75 ) ,@xmath76 is not a symmetric function ( fig .7 ) . as can be seen from the figure ,there is a larger number of small jumps to the right , but more jumps of large magnitude to the left .it is convenient to separately analyse the even and odd parts @xmath77 in order to find the exponents .we find that the even part , @xmath78 , decays asymptotically as @xmath79 with @xmath80 .the odd part @xmath81 changes sign ( as implied by the crossing of the curves in fig . 7 ) and asymptotically follows @xmath82 with @xmath83 .we verified that the values of @xmath84 and @xmath85 are the same for various @xmath75 .an interesting aspect of the function @xmath70 for the tilted interface is that @xmath86 vanishes in the thermodynamic limit . in terms of the active - site motion , this implies that though , for the tilted interface , the short - ranged jumps of the active site are mostly along the tilt , there are enough long - ranged jumps in the opposite direction to balance this .the cancellation gets better as the thermodynamic limit is approached .the tilt of the interface , therefore does not induce a net drift in the active site motion . because of the close connection of the interface growth problem to dp , several exponents associated with the growth and structure of the interface are thought to be related to the dp exponents @xmath87 and @xmath88 by scaling relations @xcite. the exponent @xmath54 in the untilted interface as well as the exponent @xmath89 in the tilted case have been argued to be related to the dp exponents by the scaling relation @xmath90 @xcite . besides this , for the tilted case, the exponents can also be obtained exactly @xcite . as mentioned earlier, the function @xmath70 does not contain all the information about the active - site motion .this is due to the presence of temporal correlations in the interface depinning model : the jump of the active site at the present instant is strongly correlated to the jumps before . as a result ,what is needed is the full distribtion function @xmath91 which is the conditional probability distribution that a jump of length @xmath68 occurs at @xmath92 given that a jump of length @xmath93 occurred at @xmath94 , a jump of length @xmath95 occured at @xmath96 and so on .the probability distribution @xmath70 is obtained by integrating out the other variables @xmath97 . in order to assess the importance of these temporal correlations , we numerically determined the conditional jump probability distribution @xmath98 , the conditional probabilty that a jump of length @xmath99 occurs given that at the previous instant , the active - spot jumped a distance @xmath68 .figure 8a presents our numerical measurement of this function for the ebm .after the initial short - ranged decay , which is a result of the local back - wetting move of the dynamics , the function is flat over a considerable range , and then decays as a power . as evident in the figure , the range of the flat region depends on @xmath68 but the value of the probability in the flat region becomes independent of @xmath68 for large @xmath68 . in figure 8bwe show the scaling plot of @xmath98 obtained by plotting the curves as a function of @xmath100 .the function evidently approaches a scaling form for large @xmath68 and @xmath99 .the scaling function is flat over a region and decays beyond as a power @xmath101 .we expect @xmath102 to hold as @xmath103 .although the measured value of @xmath104 in the range shown is larger ( @xmath105 ) , the bending apparent in the lower right portion of the curve is consistent with an approach to the value @xmath54 .qualitatively , the behaviour of this function may be understood thus .if the active site jumps a distance @xmath68 at the previous instant , one can think of the interface as pinned by a dp cluster with loops of average linear dimension @xmath106 .most of the interface would then be pinned while a portion of it fills out a loop of linear dimension @xmath106 .this would imply that on an average , any jump smaller than @xmath68 is equally likely . on length scales larger than @xmath68the motion is like the original problem and hence the jump probability decays as a power .this line of argument would imply that @xmath98 should be a scaling function of @xmath107 .figure 8b bears out this expectation .another equivalent way of understanding the function @xmath98 is using the ` backward - avalanche ' technique introduced in @xcite .a backward - avalanche is defined as follows .if at time @xmath108 a random number @xmath16 is picked as the minimum , the magnitude of the backward avalanche is @xmath109 if at time @xmath110 the random number picked was larger than @xmath16 but for time @xmath111 for @xmath112 the value of the random number picked was smaller .that is , one goes back in time to the first instant when the random number picked is larger than the present one and this interval of time is the magnitude of the backward - avalanche initiated at the present instant . if the active site hopped a distance @xmath68 in the previous instant , this would imply that the size of the spatial region affected by the backward avalanche initiated at the previous instant is @xmath106 . at the next time step any of three possibilities can occur .if the random number picked is _ smaller _ than the previous one , then the active site necessarily lies in the local region which has just been affected by the move .if the random number picked is larger , then either the random number picked belongs to the same backward - avalanche as the previous one , or it belongs to a bigger one ( which therefore encompasses the previous one ) .in the former case , the active site is equally likely to hop anywhere in the region affected by the present backward - avalanche ( @xcite ) and hence the function is flat upto a length @xmath106 . in the latter case ,active - site hops are in general larger , since the encompassing backward avalanche is larger .the probability of having the active site hop a distance @xmath113 is then related to the probability of finding a backward avalanche of this spatial extent , and hence decays as a power .another indication of the correlation in the jumps of the active site is the following quantity which is the analog of an avalanche , for jumps ( fig .let the avalanche be initiated at an instant @xmath110 by a jump of magnitude @xmath68 of the active spot .the `` jump - avalanche '' lasts as long as consecutive jumps in the active site are less than @xmath68 .that is , the avalanche is of duration @xmath109 if at time @xmath110 the active site hopped a length @xmath68 , and for @xmath109 consecutive instants after that , the hops in the active site are smaller than @xmath68 till at the @xmath114th instant , it hopped a distance greater than @xmath68 .the analog of this quantity for random numbers was first defined for models in the non - wetting invasion percolation regime @xcite and later also measured for the sneppen model @xcite .the jump - avalanche is a strong indication of the correlation in successive jumps of the active site . as fig .9 indicates , the function falls off as a slow power .this is an indication of the long - term memory in the system , as the function would decay exponentially if successive jumps had not been correlated .as discussed in the previous section , the motion of the active spot in this model is very correlated .one can then ask whether this motion reorganizes the shape of the interface ( or equivalently the arrangements of particles and holes in the particle model corresponding to the ebm ) in any specific way . below we will show that the correlated motion of the growth - spot in this model leads to the formation of a pattern in height gradients ( or equivalently the density of particles ) of the interface . the defining eq .[ eq : pattern ] for height gradients can be written in terms of densities as @xmath115 here @xmath116 is the density at site @xmath117 and the angular brackets denote a time average in the steady state .this pattern is linked to the formation of a structure which is very different from normal space - fixed patterns because it is referred to an origin @xmath118 which is moving around .in fact it can be discerned _ only _ when the origin moves .an ordinary space fixed average is translationally invariant and satisfies @xmath119 for any site @xmath117 .since the pattern is centered around the site of activity , we will refer to it as the `` activity - centered pattern '' ( acp ) .we studied @xmath11 by monte carlo simulation in the particle - hole representation .we studied systems of size @xmath3 ranging from @xmath120 to @xmath121 .the system was allowed to evolve through @xmath122 configurations before measurements were made . in order to speed up the algorithm to locate the site with the minimum @xmath23, we used a logarithmic - bin search procedure .steady state averages were computed using @xmath123 configurations .while this number of configurations was averaged over to get an accurate estimate of the decay exponent , even an average over about @xmath124 configurations indicates the presence of a strong density inhomogeneity clearly . in the untilted case ,the height profile is an even function ( fig .1a ) and @xmath125 is an odd function ( fig.10a ) decaying asymptotically as a power law @xmath126 with @xmath127 ( fig .10 , inset ) . in the tilted case as there is no @xmath128 symmetry , @xmath129 does not have a definite parity ( fig .it is useful to separately analyse the even and odd functions @xmath130 .the odd part decays as @xmath131 with @xmath132 ( fig .10 , inset ) .the even part @xmath133 where @xmath134 and @xmath135 as the lattice size @xmath136 ( fig . 11 ) .the density profiles of fig .10 correspond to the height patterns shown in figs . 1 and 2 .qualitatively , the reason for this time - averaged structure of the height profile can be seen as follows . in the untilted case , on average ,the active site is located at the peak ( fig .1a ) where @xmath16 s which have not been sampled earlier are most likely to occur ; such a region is thus more likely to contain small values of @xmath23 . in the tilted casethere is , in addition , a bootstrap effect at work .if @xmath137 , the active site is more likely to contain a particle . given the dynamics , regions to the left are more often refreshed , making the active site likely to move in this direction and find itself amidst a particle cluster .the directionality of the active site motion is evident in fig 6 .this leads to densities higher than @xmath138 on both sides of the active site ( fig .10b ) for @xmath139 sufficiently different from @xmath29 . that particles are picked more often than would be expected on the basis of the nominal density is evident from the fact that the value of the density at the active - spot is larger than @xmath138 . for the density pattern in fig .10b , which corresponds to @xmath140 , the value of the density at the origin is @xmath141 . in the next sectionwe develop an approximate theoretical description for the exponents describing the decay of the pattern in terms of the exponent @xmath54 which describes the long - ranged hops of the active site .as discussed in the previous section , a pattern in height gradients or densities is formed in the interface depinning model .it is related to the motion of the active site which rearranges the particles and holes in its wake .for example , if the active site were to remain stationary at a given site for a certain length of time , the dynamics is such that it would build up a pile - up of particles to the left of the site and a pile - up of holes to the right _ i.e. _ it would create a density shock around itself . in this section , we write down an integral equation which provides a description of the pattern in terms of @xmath70 , the probability that the active site hops a length @xmath68 in consecutive instants , as well as the local dynamics of interface readjustment .if the pattern is centered at @xmath118 at time @xmath6 , the dynamics causes two changes at the next instant : ( i ) a short - ranged readjustment of the interface changes the profile near @xmath118 .the average density change at site @xmath142 is modelled through a density - increment function @xmath143 which is short - ranged .( ii ) the active site jumps a distance @xmath144 .since the pattern is centered at the active site , the result of ( i ) followed by ( ii ) is that the average profile reproduces itself , except that it is centered at the shifted site @xmath145 .both effects are incorporated into the integral equation @xmath146 p(l ) dl \ , .\label{eq : acp}\ ] ] this equation can be solved using fourier transforms . in particular , the long - distance behaviour of the pattern @xmath11 is related to the decay of @xmath70 for large @xmath68 , resulting in a scaling relation for the exponent @xmath147 in terms of the exponent @xmath54 .the details of the analysis are given in section v b below .however , this equation provides only an approximate description . to understand the nature of the approximation made , we define a `` lvy - flight model '' , which is similar in spirit to the models of @xcite . as in these models, there is no explicit quenched disorder , but the effect of disorder is modelled by a long - ranged jump probability distribution . for this model , we show that the integral equation ( 3 ) holds exactly .further , we can gain considerable insight into the mechanism of activity - centered pattern formation by studying the effect of different decays of the jump function @xmath70 within the lvy - flight model .the lvy - flight model is defined as follows .consider a one - dimensional lattice with particles and holes .we assume that a jump probability distribution @xmath70 is specified a priori : if at @xmath148 the active site is located at a site @xmath117 , at the next instant , it can lie a distance @xmath68 away ( _ i.e. _ at @xmath149 ) with a probability @xmath70 .evidently there are no temporal correlations in this model , since at every instant , the jump length is chosen afresh from the distribution @xmath70 .spatial correlations in the active - site motion are however built in by hand since @xmath70 is given .once a particle or hole is picked for update , the local rules are assumed to be the same as in the extremal bond model _i.e. _ the particle ( or hole ) exchanges position with the nearest hole ( or particle ) to the left ( or right ) .we now derive an integral equation starting from the master equation for the lvy flight model .we first define a configuration @xmath150 to be a set of integers @xmath151 where @xmath152 is the size of the lattice and @xmath153 is the current position of the active site .the variables @xmath116 can take the values @xmath154 or @xmath51 depending on whether the site @xmath117 is occupied by a particle or is empty and @xmath153 can take any value between @xmath51 and @xmath152 .the total number of states is , therefore , @xmath155 where @xmath156 is the number of particles . in the usual manner, we characterise the steady state by a column vector @xmath157 .the entries of this column vector are the probabilities @xmath158 for the configuration @xmath150 where @xmath159 . to obtain the steady state , we need to solve the master equation @xmath160 where @xmath2 is an @xmath161 matrix connecting the different states .the dynamics that connects different states is the following : the particle ( hole ) at @xmath153 is exchanged with the nearest hole ( particle ) to the left ( right ) as specified earlier .subsequently the active site now hops from @xmath153 to @xmath162 with a probability @xmath70 .the diagonal elements of the matrix are @xmath163 .the off - diagonal elements are given by @xmath164 if configuration @xmath150 is connected to configuration @xmath24 by an elementary update and a jump of the active site of length @xmath165 and by @xmath166 otherwise .since the probability @xmath70 is normalised , ( @xmath167 ) the sum @xmath168 for each column of the matrix adds up to @xmath154 .this is a requirement for any stochastic matrix .every configuration @xmath150 can go to @xmath169 other configurations .each of these differ from @xmath150 in the positions of the particle and hole exchanged in the elementary update move , and also in the position of the active site .similarly there are @xmath152 configurations which feed into any configuration @xmath150 .the construction of these configurations is very similar to that carried out in the case of the low - noise toom - interface and related models @xcite .however , unlike in that case , here the transition probabilities are not all the same . as a resultthe steady state here is very different from the product - measure steady state found in @xcite and is difficult to characterise for general @xmath70 .however we are able to characterise one aspect of the steady state by defining a quantity @xmath170 in the following manner : @xmath171 that is , we add up the densities at a site @xmath117 for each configuration that occurs in the steady state , having first shifted the origin separately in each configuration so that the site @xmath117 for every state is one which lies @xmath117 away from the active site . from thisis subtracted the average density @xmath172 .the above equation is equivalent to eq .[ eq : pattern ] , with the time average in the latter being replaced by a weighted sum over configurations here . from examination of the @xmath2 matrix ,it is clear that in the case that @xmath70 is a constant independent of @xmath68 ( as in ordinary stochastic processes ) , all the @xmath173 s are equal . as a result , from the definition ,the acp vanishes .however , while this is a sufficient condition it is not necessary. there are instances ( see section viii ) when special symmetry considerations rule out any pattern formation .however , the presence of an acp definitely implies the presence of a nontrivial @xmath70 ; a pattern indicates correlations in the active - site motion .we will now derive eq .[ eq : acp ] for the lvy - flight model .note that the probability @xmath174 where the prime on the summation implies that only those configurations @xmath24 are considered in which the active site is @xmath68 sites away from its position in @xmath150 . further @xmath24 should transform to @xmath150 whenthe active - site is updated .this equation follows from the master equation .the sum is over all @xmath68 s .we can now substitute this in the rhs of eq .[ eq : patderv ] .@xmath175 the density at site @xmath117 in the configuration @xmath150 is multiplied by the sum of the probabilities of those configurations that lead to it after a local update ; a local exchange of a particle and hole , and a subsequent jump of the active - site of length @xmath68 . hence we can write @xmath176 here @xmath177 is a short - ranged function ( this is related to the ` density - increment ' function appearing in eq .[ eq : acp ] ) which depends on the local update rules .it is defined by eq .[ eq : patderv2 ] and is the difference between the two configurations @xmath150 and @xmath24 which are related by an update .this function is the same for this model as for the bond model ( fig .12 ) since it depends only on the local update rules . in the ebm ,it is a short - ranged function whose range is determined by the average length of particle and hole hops . later in this sectionwe will comment on circumstances in which this function can become long - ranged .substituting eq .[ eq : patderv2 ] in eq .[ eq : patderv1 ] , we find that the right hand side can be rewritten as @xmath178 p_{i } \ ,\label{eq : patderv3}\ ] ] using the definition of @xmath11 again , we finally obtain the integral equation [ eq : acp ] for @xmath129 in terms of @xmath70 , @xmath179 \ , \label{eq : inteq}\ ] ] here we have defined an averaged function @xmath180 . from particle conservation , it follows that @xmath181 . for the symmetric case of half - filling , @xmath182 particles and @xmath182 holes ,the function @xmath183 is strictly an odd function .however this does not hold when the number of particles is not equal to the number of holes ( fig .12 b ) .thus we have been able to show that the integral equation ( 2 ) is valid for the lvy - flight model . from the nature of the model ,it is clear that while spatial correlations in the active site motion are built in by hand , there are no temporal correlations in the length of subsequent hops of the active site .therefore the integral equation is exact only in such a case . in the next section , however , we carry over some of the predictions of the integral equation to the ebm and find that in some cases , it tallies quite well with numerical results . in sectionvi , we discuss briefly how to generalise eq .[ eq : inteq ] to include temporal correlations such as are present in the ebm .we now investigate the predictions of eq .[ eq : acp ] for @xmath11 in terms of a given active site hopping probability @xmath70 . and the short - ranged readjustment function @xmath177 .we solve the equation using fourier transforms .defining @xmath184 _ etc _ we find @xmath185 since we are mainly interested in the @xmath186 behaviour of eq .[ eq : ft ] , we do not need the full functional form of @xmath187 but only the leading order behaviour .given a function @xmath70 the integral equation predicts a corresponding @xmath11 .the large - distance behaviour of the acp thus depends on whether @xmath70 is short or long - ranged .we consider now three different cases for the function @xmath70 and solve for @xmath188 using equation [ eq : ft ] .we substantiate the predictions of the equation by numerically simulating the lvy flight model .* case 1 * : consider first the case of an infinite - ranged @xmath70 .the simplest case is when @xmath189 where @xmath152 is the number of sites in the lattice .this case corresponds to usual stochastic processes . in this case @xmath190 and @xmath191 .this implies that @xmath192 for @xmath193 and is therefore a very short - ranged function in space .* case 2 * : consider now the case when @xmath70 is a short - ranged function .we will consider the case when it is symmetric and hence for low @xmath194 @xmath195 . substituting this in eq .[ eq : ft ] we find that @xmath196 .this implies that @xmath197 .the particles and holes separate out completely and the active - site is located at the boundary between the two .this is easy to understand if we consider the limiting case when the active site is totally stationary . in this case , the action of the dynamics is to move all the particles from the right of the active site to its left .eventually , this leads to a total separation of particles and holes .this picture is modified only slightly when the active site executes a localised motion about any lattice site , and hence for a short - ranged @xmath70 ( fig .13 ) . * case 3 * : we now come to the case of interest for the interface depinning model _i.e. _ when @xmath70 decays as a slow power law .if @xmath198 , eq . [ eq : ft ] predicts that @xmath199 . the exponents @xmath200 and @xmath201 are related by the scaling relation @xmath202this is to be compared with the numerical estimate @xmath203 for the extremal bond model in the untilted case .consider now the case when none of the functions @xmath204 and @xmath205 has a definite parity .this is relevant for the tilted interface in the sneppen model .since @xmath206 is short - ranged , @xmath207 as @xmath186 .there is no @xmath208 term , as the elementary step of hopping a particle or hole conserves particle number , implying @xmath209 .the @xmath210 behaviour of @xmath211 is determined by the asymptotic power law decays of the even and odd parts @xmath212 as @xmath213 .thus we have @xmath214 .we might have expected @xmath215 , but in fact the mean velocity @xmath216 of the active site vanishes ( as mentioned in section iii ) implying @xmath217 .thus the integral equation predicts that to leading order , both @xmath218 and @xmath219 decay as powers @xmath220 @xmath221 , with @xmath222 and @xmath223 the prediction @xmath224 compares quite well with the numerically determined values @xmath225 for @xmath226 in the tilted case . for @xmath227, however , the numerically determined value of @xmath228 deviates substantially from the predicted value @xmath229 .the likely reason behind the discrepancy is explained in the next section .the integral equation ( 2 ) is exact for the lvy flight model , but only approximate for the ebm of interface depinning .here we briefly run over the nature of the approximations made , and possible directions for improvement .one sort of approximation is the neglect of correlations between lengths of successive jumps ; as we saw in section iii , these correlations are very marked .the extension of the integral equation to include such correlations is discussed in section vi b. as noted in case 3 of section vb for tilted interfaces ( @xmath230 ) , the pattern is not well represented by eq .( 2 ) even qualitatively .the reason for this is that the @xmath231 pattern in the ebm is formed due to a feedback effect which is missing in the lvy flight model . if @xmath232 , then within ebm dynamics , particles ( or holes if @xmath233 ) are picked more often than just @xmath234 times in @xmath235 triesthis condition is not incorporated in eq .[ eq : acp ] at all . to testhow important this effect is , we simulated the lvy flight model with the further constraint that if a site chosen for growth does not contain a particle , it is discarded and the search is continued till a site containing a particle is found .this leads to a pattern that much more resembles that in the @xmath230 ebm , in that the even part of the pattern is much more prominent than the odd part ( see fig .another point to note is that we have always taken the density - increment function @xmath236 to be short - ranged .insofar as we are interested in the ebm , this is certainly so .however , in the lvy - flight model , any power larger than 3 , results in a pattern like the one shown in fig . 13 , with a segregation of particles and holes and not one decaying as a power as predicted by the integral equation .the reason for this is that in this case , the mean squared distance @xmath237 covered by the active site in time @xmath6 is finite .hence for a large system , this is like a short - ranged motion and leads to particle hole segregation . in the process ,as particle clusters build up , the density - increment function is no longer short - ranged .lastly , while in the ebm , the probability distribution of jumps @xmath70 arises naturally from the dynamics , in the lvy - flight model we put in the spatial correlation in the active site motion by hand . in the following section , we try and remedy this point by modifying the integral equation by writing a set of coupled equations .a drawback of the lvy flight model is that the function @xmath70 has to be supplied from the outside .we try to rectify this by writing a set of coupled equations that predict @xmath11 and @xmath238 in terms of each other . in order to do this, we need to understand the probability that a certain portion of the pattern is visited by the active site . in the ebm ,the site visited last has the largest probability of being visited again since it is most likely to have the least random number .the more often a region is visited , the more the random numbers associated with sites in that region are refreshed and hence the more likely it is that the minimum random number lies in that region . since the dynamics creates a particle excess on the left of the active site and a corresponding paucity on its right , a density shock is built up in that region .the extent of the shock is directly proportional to the length of time spent by the active spot in that region .the larger the shock , the larger the density difference , and hence the larger the derivative of @xmath170 .hence the effect of a region being visited a large number of times is precisely that @xmath239 is large .but as explained above , the larger the number of times a site is visited , the larger is the probability of its being visited again .thus we conjecture that the probability of visiting a region a distance @xmath68 away from the active site is proportional to the magnitude of @xmath239 in that region , and write the following equations : @xmath240 \ , \label{eq : inteq1}\ ] ] and @xmath241 where @xmath242 is the appropriate normalising factor .equation [ eq : inteq1 ] is the integral equation that we derived in the last section .equation [ eq : coeq ] now defines @xmath70 in terms of the pattern @xmath170 .the details of the dynamics of particle - hole exchange in this model make this a reasonable premise as explained above .let us compare the results of the coupled equations with those for the untilted interface .if @xmath243 and @xmath244 .further , eq . [ eq : coeq ] predicts the relation @xmath245 .the solution is @xmath246 and @xmath247 .this compares quite well with the value of the exponents @xmath248 and @xmath249 found numerically for the ebm .the pattern , as defined earlier , is obtained by a straight time average over all configurations , taking care to shift the origin appropriately at every instant .it is interesting to ask how the pattern changes if we consider only those configurations that occur after very large jumps of the active site ( or equivalently , only those configurations for which the minimum random number is very close to the largest possible ) .a strong dependence of the pattern on the configurations averaged over is evident in fig .the pattern gets more squashed as the jumps in the active site leading to the configuration get larger .this implies that the pattern ceases to exist at stoppers and builds up again as the interface pierces through .the actual dynamics of pattern collapse and build - up is an interesting subject for further study .it should be recalled that long jumps occur much more infrequently than short jumps ( fig .7 ) ; the probability of occurrence of a stopper is thus very low .the time average in the definition of the pattern ( eq . [ eq : pattern ] ) is dominated by configurations between stoppers ( when the active site is moving within the loops of the directed percolation network ) , rather than those at stoppers ( when the active site is on the backbone of the network ) . the integral equation 3 can be modified to include correlations in time . to do this, we enlarge the definition of a configuration @xmath150 to include the active site at the previous instant as well . now the off - diagonal elements of the transition matrix elements @xmath250 if configuration @xmath24 has resulted as a consequence of a jump of the active site of length @xmath68 and is connected to configuration @xmath150 by an elementary update and a jump of the active site of length @xmath99 .the function @xmath251 is just the conditional jump probability already introduced in section iii .following the same procedure as before , we get an equation for @xmath252 , the pattern resulting from an active site hop of length @xmath253 : @xmath254 ~.\ ] ] keeping correlations upto one time step back gives the pattern a non - trivial dependence on the jump length @xmath99 .the integral equation [ eq : acp ] gives only a trivial dependence of @xmath11 on @xmath99 .this procedure can be further generalised by going back one more step in time and keeping the location of the active site two instants back .this leads to an equation of the following sort : @xmath255\ ] ] where @xmath256 is the conditional probability that the active site hops a length @xmath68 given that at the previous instant it hopped a distance @xmath257 and in the instant before that , a length @xmath258 .similarly , @xmath259 is the pattern formed when averaged over configurations that result after two consecutive jumps of @xmath257 and @xmath258 respectively . here@xmath260 .keeping the time sequence of jumps leads to an infinite hierarchy of equations .\psi_{l } ( r ) & = & \sum_{l_1,l_2 } p(l|l_1|l_2 ) [ \psi_{l_1,l_2 } ( r - l ) + \phi_{l_1,l_2 } ( r - l ) ] \\\psi_{l_1,l_2 } & = & \cdots \end{aligned}\ ] ] the integral equation [ eq : acp ] corresponds to curtailing this hierarchy at the first step by assuming that @xmath262 and @xmath263 .the activity - centred pattern @xmath11 defined in eq . 1 is a one - point correlation function , defined with respect to an erratically - moving origin . as a result, it has a strong effect on the customary , space - time averaged two - point correlation function @xmath264 here @xmath8 is a fixed site on the lattice , @xmath265 stands for a time average and \ { ... } stands for an average over all sites @xmath8 . numerical results for @xmath266 for a tilted interface show that it saturates at a value @xmath267 which decreases with increasing size @xmath3 , proportional to the system size @xmath3 , @xmath268 changes sign , as a result of the particle conservation sum rule @xmath269 . ]( fig 16 ) .this happens because subtracting out the quantity @xmath270 from the correlation function is not correct , since the presence of the acp causes a density inhomogeneity in the medium which can not be accounted for by subtracting out a constant quantity . to account for the presence of the pattern we consider the correlation function @xmath271 where @xmath272 is the location of the active site, @xmath117 is the distance from the active site and @xmath273 is the fluctuation around the average acp .a reasonable expectation is that the @xmath274 s are independent for large separations @xmath275 , _ i.e. _ @xmath276 as @xmath277 .now consider averaging the function @xmath278 over @xmath117 . on performing a space average over the right hand side of eq .[ eq : corrln ] , we note that @xmath279 thus this implies that @xmath280 approaches zero as @xmath281 .this predicts the saturation value of the correlation function @xmath282 to test this , we subtracted this estimate of the saturation value from @xmath283 and found that the saturation effect is in fact suppressed strongly ( fig . 16 ) , supporting our interpretation .it is possible that another slightly different definition of the pattern would eliminate the slight shoulder which remains in fig .16 , after subtraction of @xmath284 . usually ,if the two - point correlation function saturates as the separation between the two points is increased , the saturation value is associated with a nonzero value of the space - fixed average @xmath285 .the unusual aspect here is that there is saturation even though @xmath286 .we studied the manner in which @xmath287 approaches zero as @xmath277 . in the tilted case, we found that the correlation function decays exponentially , by studying a model where the rules are the same except that only particles are picked .this corresponds to a case of extreme tilt and is similar to the case studied by @xcite . in the untilted case, @xmath288 decays as a power law @xmath289 with @xmath290 .in the extremal bond model of interface depinning , we have described the activity - centered pattern in height gradients which forms as a result of correlated motion of the active site . however ,this is not the only sort of pattern that is formed .there is pattern formation also in the value of the average random number @xmath23 at a site , as a function of the distance from the active site . in analogy with eq .[ eq : pattern ] , we define this pattern in random numbers as @xmath291 here @xmath292 is the random number at a distance @xmath117 from the active - site and the time average @xmath293 and space average @xmath294 are performed over configurations in the steady state , as before .moreover , patterns are found in other extremal models as well .figures 17 , 18 and 19 show the @xmath23-patterns in the ebm , the bak - sneppen model of biological evolution @xcite and the zaitsev model of low - temperature creep @xcite respectively .numerically , it is difficult to directly extract the manner in which the patterns shown in figs .17 - 19 approach their asymptotic values , as fits to power law decays are very sensitive to the assumed saturation value .we avoided this problem by studying the fourier transforms of the functions , as the saturation value influences only the single fourier mode at @xmath295 . in all three cases ,the fourier transforms show evidence of power law behaviour as @xmath186 , implying power law approaches of the @xmath23-patterns to their respective saturation values in real space .however , we have not developed an analytical description of the @xmath23-patterns in any of these models .pattern formation thus seems to be generic to extremal models .however there are instances when considerations of symmetry rule out the formation of a pattern .this is exemplified in the following model , similar in spirit to that considered in @xcite .the rules of the dynamics are the same as in the ebm , random numbers are assigned to every site and the minimum is picked , except that there is no net current of particles as there was in the ebm .if the site picked is occupied by a particle , the particle exchanges place with the first hole to its left , and if the site is occupied by a hole , the hole too exchanges place with the first particle to its left . at halffilling , there is no net current .this model is the most symmetric of those considered so far and there is no density pattern formed .however , as in the bak - sneppen and the zaitsev models , there is a nontrivial pattern in random numbers in this model .we emphasize that the feature of the dynamics which is responsible for activity - centered pattern formation is the existence of correlations in the motion of the active site .extremal models constitute just one class in which there are such correlations .an example of another such class is models of certain types of reaction - diffusion systems , where the activity is quite constrained and correlated . in another physical context, it would seem that coherent structures which form in turbulent flows @xcite may well be described by activity - centered patterns . in our definition of the pattern ,we take an average over all times , keeping track of the moving structure .this is to be contrasted with previously used methods to identify such moving structures , based on the notion of conditional sampling , namely , averaging only over those time zones in which the activity is at a particular space - time location @xcite .our definition takes configurations at all times into account , but requires a shift of the origin at every instant .we have introduced a variant of the sneppen model of interface depinning the extremal bond model and have studied the effect of the dynamics of the growth process on the shape of the interface .our principal result is the observation that there is a nontrivial structure which forms in the interface , and which moves along with the active site .a simple time average of the height gradients , measured in a frame of reference which moves with the active site , defines the activity - centered pattern which serves to quantify the structure .our numerical study shows that the pattern has a tail which decays as a power law at large distances .an understanding of the mechanism underlying activity - centered pattern formation was obtained by writing an integral equation which relates the pattern to the probability distribution of active - site jump lengths .the integral equation could be derived by writing an extended master equation for the lvy flight model , making it clear that the equation is exact only when there are no temporal correlations between successive jump lengths . in the extremal bond model , however ,such correlations are strong .we have shown that a correct description of the pattern , involving temporal correlations , necessitates keeping an infinite hierarchy of equations . terminating this hierarchy at the very first step results in our integral equation .it would be of interest to understand how much one can better this description by keeping more steps in the hierarchy .the activity - centered pattern is a one - point function , and so enters the definition of two - point correlation functions .the physical point is that the density inhomogeneity caused by the pattern must be taken into account , by subtracting the relevant quantity from the density - density correlation function .if this is not done , and the square of the the nominal density @xmath138 is subtracted instead , the correlation function can exhibit a rather unusual sort of finite size effect .this is an interesting point since the space - fixed time - averaged density at a site is @xmath138 , and it is only the density as defined in eq .[ eq : pattern ] with respect to the moving active site which is different from the nominal density .yet the space - fixed two - point function is affected by this `` hidden '' pattern .the presence of the pattern clearly points to a non - homogeneity in the interface : the region around the active site looks very different on average from the region far away from it .for instance , we expect there to be a larger length of interface in a region of fixed size @xmath296 around the active site , than in a region opposite it .we monitored mean squared fluctuations of the height around the instantaneous average , in regions around and opposite the active site , and found a pronounced difference ( factor @xmath297 , for both tilted and untilted interfaces with @xmath298,@xmath299 ) .this effect is smaller at stoppers , in keeping with our finding that the pattern itself is suppressed there .this excess length of interface associated with the activity - centered pattern may provide a useful way to identify the active region in experiment . finally , it was pointed out that activity - centered pattern formation may occur in a wide variety of other physical contexts , ranging from low - temperature creep of dislocations to structures in turbulent flows .we have presented numerical evidence for this sort of pattern formation in a number of other extremal models .but more generally we expect activity - centered patterns to form whenever there are strong correlations between successive locations of the active site. 999 j. p. stokes , a. p. kushnick and m. o. robbins , phys .lett . * 60 * , 1386 ( 1988 ) .m. a. rubio , c. a. edwards , a. dougherty and j. p. gollub , phys .lett . * 63 * , 1685 ( 1989 ) . v. k. horvath , f. family and t. vicsek , j. phys .a * 24 * , l25 ( 1991 ) .s. he , g. l. m. k. s. kahanda and p. wong , phys .lett . * 69 * , 3731 ( 1992 ) .s. v. buldyrev , a .-barabasi , f. caserta , s. havlin , h. e. stanley and t. vicsek , phys .a. * 45 * , r8313 ( 1992 ) .j. zhang , y .- c .zhang , p. alstrom and m. t. levinsen , physica a * 189 * , 383 ( 1992 ) .s. krishnamurthy and m. barma , phys .lett . * 76 * , 423 ( 1996 ) .m. kardar , g. parisi and y. -c .zhang , phys . rev. lett . * 56 * , 889 ( 1986 ) .d. s. fisher in _ nonlinearity in condensed matter _ edited by a.r .bishop , d.k .campbell , p. kumar and s.e .trullinger ( springer , berlin,1987 ) , p. 339 .t. halpin - healy and y .- c .zhang , phys .rep . * 254 * , 215 ( 1995 ) .r. bruinsma and g. aeppli , phys .lett . * 52 * , 1547 ( 1984 ) .j. koplik and h. levine , phys . rev .b * 32 * , 280 ( 1985 ). d. a. kessler , h. levine and y. tu , phys .a * 43 * , 4551 ( 1991 ) .g. parisi , europhys .* 17 * , 673 ( 1992 ) .t. nattermann et al , j. phys ii france * 2 * , 1483 ( 1992 ) . o. narayan and d. s. fisher , phys . rev .b * 48 * , 7030 ( 1993 ) .zhang , j. physique * 51 * , 2129 ( 1990 ) .j. g. amar and f. family , j. phys .a * 24 * , l79 ( 1991 ) .e. medina , t. hwa , m. kardar and y .- c .zhang , phys .a * 39 * , 3053 ( 1989 ) .p. meakin and r. jullien , europhys . lett .* 9 * , 71 ( 1989 ) .j. g. amar , p .-m . lam and f. family , phys .a * 43 * , 4548 ( 1991 ) . m. h. jensen and i. procaccia , j. phys ii france * 1 * , 1139 ( 1991 ) . l. -h . tang and h. leschhorn ,a * 45 * , r8309 ( 1992 ) .k. sneppen , phys .69 * , 3539 ( 1992 ) .tang and h. leschhorn , phys .lett . * 70 * , 3832 ( 1993 ) .n. martys , m. cieplak and m. o. robbins , phys .. lett . * 66 * , 1058 ( 1991 ) ; n. martys , m. o. robbins and m. cieplak , phys . rev .b * 44 * , 12294 ( 1991 ) .r. lenormand and s. bories , c. r. acad .paris * b291 * , 279 ( 1980 ) .r. chandler , j. koplik , k. lerman and j. f. willemsen , j. fluid mech .* 119 * , 249 ( 1982 ) .d. wilkinson and j. willemsen , j. phys .a * 16 * , 3365 ( 1983 ) .r. lenormand and c. zarcone , phys .lett . * 54 * , 2226 ( 1985 ) ; see m. sahimi , rev .* 65 * , 1393 ( 1993 ) for an account of related experiments .l. furuberg , j. feder , a. aharony and t. jossang , phys .lett . * 61 * , 2117 ( 1988 ) .s. roux and e. guyon , j. phys .a * 22 * , 3693 ( 1989 ) .s. roux and a. hansen , j. de physique i * 4 * , 515 ( 1994 ) .k. sneppen and m. h. jensen , phys .lett . * 71 * , 101 ( 1993 ) .h. leschhorn and l. -h .tang , phys .e * 49 * , 1238 ( 1994 ) .s. maslov , m. paczuski and p. bak , phys .* 73 * , 2162 ( 1994 ) ; m. paczuski , s. maslov and p. bak , phys . rev .e * 53*. s. i. zaitsev , physica a * 189 * , 411 ( 1992 ) .p. bak and k. sneppen , phys .* 71 * , 4083 ( 1993 ) .j. m. kim and j. m. kosterlitz , phys .lett . * 62 * , 2289 ( 1989 ) .p. b. thomas and m. paczuski , condmat 9602023 .m. barma , j. phys .a * 25 * , l693 ( 1992 ) .b. derrida , j. lebowitz , e. r. speer and h. spohn , j. phys . a * 24 * , 4805 ( 1991 ) .d. dhar , m. barma and m. k. phani , phys .lett . * 47 * , 1238 ( 1981 ) .w. kinzel , in _ percolation structures and processes _ , eds .g. deutscher , r. zallen and j. adler ( hilger , bristol , 1983 ) , p. 425 .z. olami , i. procaccia and r. zeitak , phys .e * 49 * , 1232 ( 1994 ) . s. maslov and y. -c .zhang , phys .lett . * 75 * , 1550 ( 1995 ) .tang , m. kardar and d. dhar , phys .. lett . * 74 * , 920 ( 1995 ) .s. maslov , phys .* 74 * , 562 ( 1995 ) .binder , m. paczuski and m. barma , phys .e * 49 * , 1174 ( 1994 ) .k. sneppen and m. h. jensen , phys .e * 49 * , 919 ( 1994 ) .w. d. mc comb , _ the physics of fluid turbulence _ , ( clarendon press , oxford , 1990 ) .

<S> we study erratically moving spatial structures that are found in a driven interface in a random medium at the depinning threshold . </S><S> we introduce a bond - disordered variant of the sneppen model and study the effect of extremal dynamics on the morphology of the interface . </S><S> we find evidence for the formation of a structure which moves along with the growth site . </S><S> the time average of the structure , which is defined with respect to the active spot of growth , defines an activity - centered pattern . </S><S> extensive monte carlo simulations show that the pattern has a tail which decays slowly , as a power law . to understand this sort of pattern formation , we write down an approximate integral equation involving the local interface dynamics and long - ranged jumps of the growth spot . </S><S> we clarify the nature of the approximation by considering a model for which the integral equation is exactly derivable from an extended master equation . </S><S> improvements to the equation are considered by adding a second coupled equation which provides a self - consistent description . </S><S> the pattern , which defines a one - point correlation function , is shown to have a strong effect on ordinary space - fixed two - point correlation functions . </S><S> finally we present evidence that this sort of pattern formation is not confined to the interface problem , but is generic to situations in which the activity at succesive time steps is correlated , as for instance in several other extremal models . </S><S> we present numerical results for activity - centered patterns in the bak - sneppen model of evolution and the zaitsev model of low - temperature creep . 2 </S>

1208.1369

plasma turbulence plays important roles in fusion devices and various space and astrophysical situations , where it is an essential phenomenon underlying transport of mean quantities and particle heating @xcite . for these collisionless or weakly collisional plasmas, such turbulence requires a kinetic description in phase space , especially at small scales where dissipation takes place .turbulence theory in kinetic phase space is more than a simple extension of navier - stokes turbulence into higher dimensions as velocity space is not exactly equivalent to position space : for instance , there is no translational symmetry ( i.e. small velocities are not equivalent to large velocities ) , and there is always some large - scale velocity dependence imposed by the background distribution function ( e.g. on the scale of the thermal velocity for a maxwellian background ) ; see also ref . @xcite , sec .2.1 . however , in some simplified cases , classical fluid dynamical theories @xcite can be naturally extended into phase space @xcite . in magnetized plasmas , the gyrokinetic ( gk ) theory @xcite provides the minimal kinetic description of the low - frequency turbulence . in electrostatic gyrokinetics ,nonlinear interactions introduce a cascade of perturbed entropy ( which is proportional to the perturbed free energy at the sub - larmor scales ) to smaller scales both in position and velocity space @xcite . when the turbulence is restricted to two position - space dimensions , the system has two collisionless invariants .one of them is the free energy or entropy which is also an invariant in three dimensions ( 3d ) , and another , approximately related to kinetic energy in the long wave - length limit , is particular to two dimensions ( 2d ) . in thisregard , 2d gyrokinetic turbulence is analogous to 2d fluid turbulence , and indeed , reduces to it in a particular long wave - length limit @xcite .these two invariants can not share the same local - interaction space in a kolmogorov - like phenomenology , which leads to a dual cascade ( forward and inverse cascades ) @xcite .as the nonlinear term of 2d gyrokinetics is identical in form to that of 3d gyrokinetics , the understanding of the nonlinear interaction in purely 2d system will serve as a foundation for understanding general 3d magnetized plasmas @xcite .in this paper we focus on the freely decaying turbulence problem for the electrostatic 2d gk system .we first introduce the gk equation briefly and describe its basic nature in sec .[ sec : equations ] .we also review basic characteristics of the nonlinear phase mixing at sub - larmor scales that are reported in refs .sections [ sec : theory ] and [ sec : simulation ] are the main contents of the paper . in sec .[ sec : theory ] , we describe a phenomenological theory of the dual cascade in freely decaying turbulence based on the theory first developed for 2d fluid turbulence @xcite .there are three regimes for the turbulence : these correspond to weakly collisional , marginal , and strongly collisional cases . in the strongly collisional case ,dissipation acts strongly and the system decays to become more and more dissipative . in the marginal case ,the collisional dissipation balances with nonlinear ( inertial ) turnover and the system is preserved in this state . for weakly collisional cases ,the turbulence is , in a sense , able to escape the effects of dissipation , becoming less and less collisional in time , and tending asymptotically to a state of zero electrostatic energy decay .section [ sec : simulation ] presents the results of numerical simulation of the freely decaying turbulence .we demonstrate the inverse ( forward ) transfer of energy ( entropy ) in the direct numerical simulation and investigate the time - asymptotic decay laws of the two collisionless invariants , comparing with the theory developed in sec .[ sec : theory ] .we conclude with a summary of our results in sec . [sec : summary ] .we first introduce the gyrokinetic ( gk ) model briefly @xcite .since we are concerned with turbulence in magnetized plasmas , the dynamics of interest is much slower than particle gyromotion .the gyromotion is thus averaged over , eliminating gyroangle dependence from the system .the gk system has 3 spatial coordinates ( @xmath3 , @xmath4 , @xmath5 ) , and 2 velocity coordinates ( @xmath6 , @xmath7 ) , where @xmath8 and @xmath9 denote perpendicular and parallel directions to the background magnetic field , respectively .we assume the background plasma and magnetic field are uniform in space and time .it is necessary to distinguish between the particle coordinate @xmath10 and the gyrocenter coordinate @xmath11 .these coordinates are connected by the catto transform @xcite @xmath12 where @xmath13 is the unit vector along the background magnetic field and @xmath14 is the gyrofrequency .we further reduce the gk equation to 2d in position space , or 4d in phase space @xcite , by ignoring variation along the mean field ( @xmath15 ) .this not only reduces the dimension of the system but also removes one of the mechanisms of creating velocity - space structure linear parallel phase mixing ( landau damping ) , a much more familiar and better understood phenomenon than the nonlinear perpendicular phase mixing , on which we will concentrate in this paper . the resulting gk equation ( for the ions )is @xmath16 where @xmath17 is the gyroaverage holding the guiding center position @xmath11 constant , @xmath18 is the gyroaverage of the perturbed ion distribution function @xmath19 , @xmath20 is the electrostatic potential , @xmath21 is the background magnetic field ( aligned with the @xmath5-axis ) , and @xmath22 .the collision operator @xmath23 we use in our simulations describes pitch - angle scattering and energy diffusion with proper conservation properties @xcite ( see also appendix [ sec : collision ] ) .the potential @xmath20 in eq . is calculated from the quasineutrality condition : written in the fourier space , it is @xmath24 where the hat denotes the fourier coefficients , @xmath25 is the bessel function , representing , in fourier space , the gyroaverage at fixed particle position , @xmath26 , @xmath27 is the modified bessel function of the first kind , @xmath28 , @xmath29 is the ion thermal larmor radius , @xmath30 is the charge , @xmath31 and @xmath32 are the density and temperature of the background maxwellian @xmath33 , and @xmath34 and @xmath35 are the species indices .one may use @xmath36 for boltzmann - response ( 3d ) electrons or @xmath37 for no - response ( 2d ) electrons @xcite .hereafter we use the no - response electrons as in refs .@xcite because formally , the electrons can not contribute to the potential if @xmath38 exactly @xcite .this choice is not very important as it only introduces minor differences in various prefactors .the 2d electrostatic gk system possesses two quadratic positive - definite collisionless invariants @xcite , @xmath39 there are various ways to choose two independent invariants in our system . in refs .@xcite total perturbed entropy ( or free energy ) , @xmath40 ( see eq . ( 3.9 ) of ref .@xcite ) , is used in order to make the connection to thermodynamics .here we use the quantity @xmath41 for the sake of simplicity @xcite .in fact , @xmath42 averaged over @xmath11 is itself conserved ( that is , it is conserved for each value of @xmath6 ) @xcite . however , it is sufficient for the purposes of this paper to consider only the integrated quantity @xmath41 .one can adapt the arguments of ref .@xcite to show that the presence of conserved quantities @xmath41 and @xmath43 implies a dual cascade ( i.e. , both forward and inverse cascades ) @xcite . is comparable to or smaller than the larmor radius @xmath29 , the gyroaverage of the electric field induces a decorrelation of the distribution function at the velocity - space scale corresponding to the difference in larmor radii @xmath44 .this figure is reproduced with permission from ref .copyright ( 2009 ) by the american physical society ( http://link.aps.org/abstract/prl/v103/e015003 ) ., width=188 ] due to the neglect of the parallel streaming term , the creation of velocity space structure originates solely from the advection of the distribution function by the gyroaveraged @xmath45 drift [ the nonlinear term in eq . ] . for small - scale electric fields , particles with different gyroradiiexecute different @xmath45 motions because they `` see '' different effective potentials ; this leads to nonlinear phase mixing and other novel phenomena @xcite .as the turbulence cascades through phase - space , the excitation of fluctuations at spatial scale @xmath46 induces velocity structure of scale @xmath47 in the perpendicular velocity space which corresponds to the difference of the larmor radii @xmath48 ( see fig . [fig : mixing ] and refs .@xcite ) . in other words ,when the spatial decorrelation scale is @xmath46 , two particles with larmor radii separated by @xmath49 become decorrelated , since the gyroaveraged potentials these two particles feel are different .this nonlinear phase - space mixing effect was first pointed out by dorland and hammett @xcite . in refs .@xcite numerical simulations focused on the forward cascade of entropy , and showed a simultaneous creation of structures in position and velocity space in accordance with the theoretical prediction @xcite . $ ] from one of the forward cascade simulation reported in ref . @xcite .kinetic turbulence proceeds in the position and velocity space simultaneously .this figure is reproduced with premission from ref .copyright ( 2010 ) by the japan society of plasma science and nuclear fusion research ., height=207 ] figure [ fig : forward cascade result ] shows a normalized , time - averaged spectral density of entropy @xmath50 in a forward cascade simulation , where the @xmath6 structure is characterized by the hankel transform @xcite @xmath51 and @xmath52 denotes the fourier - hankel coefficients . in analogy with the reynolds number in fluid turbulence, we may introduce an amplitude - dependent dimensionless number @xmath53 @xcite , the ratio of collision time to nonlinear decorrelation time measured at the thermal larmor radius , which characterizes the smallest scales created in both position and velocity space by @xmath54 .thus , @xmath53 quantifies both how `` inertial '' the turbulence is ( in the sense of the reynolds number ) and how `` kinetic '' it is , because it measures the nonlinear turnover at the thermal larmor radius , which marks the beginning of the `` nonlinear phase - mixing range . ''the degree of freedom , corresponding to computational problem size , may also be characterized by @xmath53 , and scales as @xmath55 in three phase - space dimensions , consisting of two position - space and one velocity - space dimensions .we note here that the statistical description of our phase - space turbulence requires a 2d spectral space @xmath56 , in contrast to isotropic fluid turbulence , which requires only the scalar wave number. we will refer to this 2d spectrum @xmath57 in the following sections .with the use of the hankel transform for velocity space and the conventional fourier decomposition for position space , we may discuss the evolution of fourier - hankel modes of @xmath58 in @xmath59 space .figure [ fig : phenom ] depicts the simplest example of the evolution of @xmath60 in the freely decaying turbulence .space . a diagonal fourier - hankel mode ( indicated by the red circle )cascades towards small scales in position and velocity space according to the forward cascade of @xmath41 ( blue dotted arrows ) .nonlinear interaction conserves both @xmath41 and @xmath43 , so the forward cascade must be accompanied by the excitation of larger - scale modes on the diagonal ( green solid arrow ) , which corresponds to the inverse cascade of @xmath43 ., width=226 ] by applying the definition to eq . , we find that the fourier - hankel modes @xmath61 with @xmath62 ( purple shaded regions marked by `` no @xmath43 '' in fig . [fig : phenom ] ) give no contribution to @xmath63 due to the orthogonality of the bessel functions . in these regions , therefore , fluctuations associated with this part of @xmath41 can in principle cascade to small scales with no effect on the spectral distribution of @xmath43 . on the other hand , the spectrum of @xmath43 satisfies @xcite @xmath64 where @xmath57 is given by eq ., i.e. , while @xmath41 is the sum over the entire @xmath59-plane , the second invariant @xmath43 is only composed of the `` diagonal '' @xmath65 components ( dashed line in fig .[ fig : phenom ] ) . in the small - scale limit ( @xmath66 ) , we may approximate @xmath67 , so eq . implies @xmath68 which is indicated for the diagonal components in fig . [fig : phenom ] .as we will see , the decaying turbulence evolves to a state of local cascade . as a consequence of eq ., this state is characterized by an inverse cascade of @xmath43 and forward cascade of @xmath41 , which can be explained as follows @xcite .the energy of a diagonal fourier - hankel mode ( indicated by the red circle ) cascades towards small scales in position and velocity space according to the forward cascade of @xmath41 ( blue dotted arrows ) .nonlinear interaction conserves both @xmath41 and @xmath43 , so the forward cascade must be accompanied by the excitation of larger - scale modes on the diagonal ( green solid arrow ) , which corresponds to the inverse cascade of @xmath43 .simulation results reported in sec . [ sec : simulation ] indicate that this is the universal property of the time - asymptotic state of freely decaying turbulence . however , there are other interactions possible in the transient and/or driven cases ( for more details on such processes , see refs .@xcite ) . in this section ,we derive scaling laws for the decay of collisionless invariants based on some simple phenomenological arguments .the underlying assumptions are as follows .first , we assume that collisionless invariants are dominated by a single scale @xmath69 , in the same way energy in fluid turbulence is dominated by the `` energy - containing scale . ''then , in terms of the amplitude at this scale , @xmath70 where @xmath71 and @xmath72 are the rms values of the distribution function and potential associated with the scale @xmath69 .second , we assume that the evolution of @xmath69 is governed by the inverse cascade along the diagonal @xmath73 , so defining @xmath74 , we have @xmath75 both assumptions are found to be valid in the numerical simulations described in sec .[ sec : simulation ] .depending on the strength of collisions compared to that of turbulent dynamics , we may derive several different scaling laws . in order to quantify the collisionality, we characterize the instantaneous turbulent state via a sub - larmor version of the dimensionless number introduced in ref .@xcite : it is the ratio of the collision time to the nonlinear decorrelation time @xmath76 at the scale @xmath69 @xcite , @xmath77 we have taken the collisional decay rate to scale as @xmath78 because the gk collision operator is second order in velocity and spatial derivatives @xcite and @xmath73 . note that in going from the second to the third expression in eq . , we used eq . and @xmath79 valid in the @xmath80 regime from the form of the convective derivative @xmath81 .note that the factor of @xmath82 is introduced due to the large argument expansion of the bessel function associated with the gyroaverage of @xmath72 .in analogy with the microscopic reynolds number @xcite , the initial value and the time evolution of @xmath0 provide a natural way to classify various physical regimes . in the following paragraphs, we describe three different decay laws , classified using eq . .they are the weakly collisional ( @xmath83 ) , marginal ( @xmath84 ) and strongly collisional ( @xmath85 ) cases . herethe constant @xmath2 denotes the value of @xmath0 that divides these three regimes , which corresponds to the kinetic version of the `` critical reynolds number '' in ref .@xmath2 is a universal constant by conjecture , which applies to the asymptotic state of the freely decaying turbulence as we explain below ( see also sec . [sec : num decay law ] and fig .[ fig : dorland number ] ) .[ [ weakly - collisional - case ] ] weakly collisional case + + + + + + + + + + + + + + + + + + + + + + + in the asymptotic limit where collision frequency becomes negligible ( @xmath86 ) , the second invariant @xmath43 does not decay at all as it is transferred to larger scale where dissipation is inactive . on the other hand , the first invariant @xmath41 is transferred to smaller scales , and is dissipated by small but finite collisions there .the decay rate of @xmath41 is determined by the rate of its transfer to small scales , thus @xmath87 where we used @xmath76 for the characteristic time of nonlinear transfer at scale @xmath69 [ see eq . ] .then from @xmath88 and @xmath89 , we obtain @xmath90 where we used eqs . and .in this limit eq. implies that @xmath0 increases in time as @xmath91 .[ [ marginal - case ] ] marginal case + + + + + + + + + + + + + as collision frequency becomes large , collisional damping at scale @xmath69 becomes important .when it balances with the nonlinear transfer , the two terms in @xmath92 become comparable , where we have taken the collisional decay rate to scale as @xmath78 . from eqs . , and @xmath88 , we obtain the following decay laws of collisionless invariants @xmath93 substitution of eq . into eq. immediately leads to a constant @xmath0 which equals the marginal dimensionless number @xmath2 .it is noted that eq .implies that the second invariant also decays by collisions in a consistent manner : @xmath94 [ [ strongly - collisional - case ] ] strongly collisional case + + + + + + + + + + + + + + + + + + + + + + + + + in this case we may regard the turbulence to be fairly dissipative .we find that @xmath41 and @xmath43 do not individually satisfy decay laws as powers of @xmath95 .however , eq . applies , as does the analogous equation describing the collisional decay of @xmath41 , @xmath96 these equations imply [ with the help of eq . ]decay laws for both @xmath97 and the ratio @xmath98 : @xmath99 in this case we can deduce that @xmath0 decays in time .it is noted that although eq . holds for both marginal and strongly collisional cases , the individual decay laws for @xmath41 and @xmath43 may vary with @xmath0 [ i.e. eq .is not satisfied here ] ; however , the decay law for the ratio @xmath98 is robust .in this section we show the results of numerical simulations performed using the mpi - parallelized nonlinear gyrokinetic code astrogk @xcite .all simulations are made in two spatial dimensions ( @xmath3 and @xmath4 ) and two velocity dimensions ( energy @xmath100 and pitch angle @xmath101 ) @xcite .the system size is restricted to @xmath102 , so as to focus on the sub - larmor regime .time is normalized by the initial turnover time @xmath103 where @xmath104^{1/2}$ ] , and @xmath105 is the wave number at which initial spectrum is peaked [ see eqs . and below ] .our biggest run ( run d in table [ run table ] ) used 9,216 processor cores for about 50 wall - clock hours . in order to investigate the freely decaying turbulence , we prepared initial conditions peaked at a high wave number @xmath105 ( @xmath106 ) and made a series of simulations for varying initial wave number @xmath105 and collision frequency @xmath107 .six runs were made for two kinds of velocity distribution functions ( described below ) and are listed in table [ run table ] . .index of the runs described in sec .[ sec : simulation ] . [ cols="^,^,^,^,^,^,^",options="header " , ] [ run table ] the different initial velocity distributions allow us to vary the initial ratio of invariants . for each initial velocity distribution, we made weakly and strongly collisional simulations .[ [ coherent - velocity - distribution - runs - ad ] ] coherent velocity distribution ( runs a d ) + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + the first type of the initial velocity distribution is bessel - like , with a maxwellian envelope @xmath33 , @xmath108 j_0\left ( \frac{{k_{\perp}}{v_{\perp}}}{\omega } \right ) f_0 , \label{coherent init cond}\ ] ] where the width of the wave - number peak is @xmath109 . from quasi - neutrality [ eq . ] , it can be deduced that small - scale velocity oscillation whose period in velocity space is comparable to @xmath110 are needed to produce a finite potential ( see the discussions in sec . [sec : phenomenology ] ) .indeed , such oscillatory structure can be found in the eigenmodes of the entropy mode @xcite , which can be unstable at quite large @xmath111 . the fourier - hankel spectral density corresponding to eq .is concentrated at @xmath112 as plotted in the left panel of fig .[ fig : coherent 2d - spectra ] corresponding to @xmath113 , and represented by the red spot in fig .[ fig : phenom ] . for eq .the initial ratio of the invariants is @xmath114 which we note is the same as what is predicted for the time - asymptotic state , given by eq . .[ [ random - velocity - distribution - runs - e - and - f ] ] random velocity distribution ( runs e and f ) + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + the second type is a random velocity distribution that represents velocity scales @xmath115 , @xmath116 \nonumber \\ & \times \frac{1}{n } \sum_{j=1}^n ( 2 \delta_j - 1 ) \sqrt{p_j{v_\textrm{th } } } j_0 ( p_j { v_{\perp } } ) f_0 , \label{random init cond}\end{aligned}\ ] ] where the width of the wave - number peak is @xmath109 , @xmath117 , @xmath118 and @xmath119 are homogeneous random numbers in @xmath120 , and @xmath121 is the number of random modes for each @xmath122 .the factor of @xmath123 is introduced to cancel the same factor of the bessel function in the asymptotic regime ( @xmath124 ) . as is shown in the left panel ( @xmath113 ) of fig .[ fig : random 2d - spectra ] , eq . corresponds to a high - density band parallel to the @xmath125 axis in the @xmath59 spectral space . in this casethe initial ratio of the invariants is @xmath126 note that while eq .is used to represent the velocity structure of a coherent mode , the distribution is designed to mimic the random nature of the velocity space that develops from forward cascade ( see ref .@xcite ) .the time evolution of the 2d spectrum @xmath60 [ see eq . ] for runs d and f ( table [ run table ] ) is shown in figs .[ fig : coherent 2d - spectra ] and [ fig : random 2d - spectra ] , respectively . in fig .[ fig : coherent 2d - spectra ] , the spectrum is concentrated around @xmath127 at @xmath128 [ see also eq . and table [ run table ] ] .the spectral density is transferred diagonally to the lower - left corner of the @xmath59 space .since the high-@xmath59 components suffer strong collisional dissipation , the upper - right energy content damps quickly .the remaining lower - left transfer dominates after @xmath129 , and inverse cascade follows .nonlinear transfer of the invariants may be directly monitored as is done for the forward cascade simulation in refs .following ref .@xcite , we define a shell filtered function by @xmath130 where @xmath131 .then the evolution of energy in shell @xmath132 is described as @xmath133 where we introduced an energy transfer function @xmath134 which measures the rate of energy transferred from shell @xmath135 to shell @xmath132 . here@xmath136 denotes the spatial volume of the domain .the entropy transfer function is defined in a similar manner in ref .@xcite and is recaptured here in the present notation : @xmath137 note that the shell filtering is performed on @xmath20 and @xmath58 in eqs . and , respectively , so that @xmath138 and @xmath139 both satisfy antisymmetry under exchange of @xmath132 and @xmath135 .snapshots of the normalized energy transfer function @xmath140 and entropy transfer function @xmath141 at @xmath142 are shown for the simulation of coherent initial condition ( run d ) in fig .[ fig : coherent ktrans ] . and ( b ) entropy transfer function @xmath141 at @xmath142 for rund [ see eqs . , and table [ run table ] ] ., title="fig : " ] and ( b ) entropy transfer function @xmath141 at @xmath142 for run d [ see eqs . , and table [ run table ] ] ., title="fig : " ] at this time the peak of the wave - number spectra ( not shown ) is located at @xmath143 .figure [ fig : coherent ktrans](a ) shows that the energy transfer is ( 1 ) well localized along the diagonal ( meaning local - scale interaction ) , ( 2 ) has strong positive values at @xmath144 , @xmath145 , @xmath146 , ( 3 ) has corresponding negative values at @xmath147 , @xmath148 , @xmath149 , and ( 4 ) disappears rather quickly at high wave - number shells ; namely , the energy is transferred from large wave - number shells ( small scales ) to small wave - number shells ( large scales ) around the spectral peak , showing clear evidence of the inverse cascade of @xmath43 .this inverse transfer of @xmath43 creates the peaked , high - density region at the diagonal of @xmath59 space , which propagates along the diagonal toward the small @xmath59 regime as seen at @xmath150 and @xmath151 of fig .[ fig : coherent 2d - spectra ] . at an initial transient stage ,however , we observe differences including nonlocal transfer , which is discussed elsewhere @xcite .note that as the contribution to @xmath43 only comes from the @xmath65 component of the distribution function , the transfer in velocity space proceeds in conjunction with that of position space , effectively `` unwinding '' fine structure in position and velocity space simultaneously .this is a striking feature of the phase - space cascade . on the other hand , from fig .[ fig : coherent ktrans](b ) , the entropy transfer ( 1 ) is well localized along the diagonal ( meaning local - scale interaction ) , ( 2 ) has a positive peak at @xmath152 and corresponding negative peak at @xmath153 , ( 3 ) extends to larger wave - number shells contrary to the transfer of @xmath138 ; namely , the entropy is mostly transferred from small wave - number shells ( large scales ) to large wave - number shells ( small scales ) , showing clear evidence of the forward cascade of @xmath41 .this forward transfer of @xmath41 creates the broad off - diagonal spectra seen at @xmath150 and @xmath151 of fig .[ fig : coherent 2d - spectra ] , similar to the forward cascade simulation @xcite ( see also fig . [fig : forward cascade result ] ) .however , it is interesting to note the reversed coloring around @xmath154 of fig .[ fig : coherent ktrans](b ) , which is located outside of the closest diagonal grids that show forward cascade of @xmath41 .this denotes the inverse transfer of @xmath41 associated with the strong inverse cascade of @xmath43 ; however , its magnitude is less than @xmath155 of the peak of the forward transfer of @xmath41 [ note also the difference of the scale on the color bar in fig .[ fig : coherent ktrans](a ) and ( b ) ] . axis . ] in the case of random initial condition , the initial random velocity distribution is characterized by a vertical band in @xmath59 space ( see fig . [fig : random 2d - spectra ] ) , covering @xmath156 ( note that @xmath157 as shown in table [ run table ] , run f ) .most of the energy in the band is transferred to higher wave number and then dissipated by strong collisions as it is not associated with @xmath43 ( recall the discussions in sec .[ sec : phenomenology ] ) . for larger scales ( @xmath158 ) , excitation of fourier - hankel modes is concentrated about the diagonal component ( @xmath65 ) , as seen at @xmath159 and @xmath160 . as can be expected from the time evolution of the 2d spectra ( compare figs .[ fig : coherent 2d - spectra ] and [ fig : random 2d - spectra ] ) , random - initial - condition cases and the coherent - initial - condition cases show similar transfer after the transient phase ( that is , the transfer functions resemble those of fig .[ fig : coherent ktrans ] ) .figure [ fig : slices along p ] shows several slices of fig .[ fig : random 2d - spectra ] .these slices are taken at the peak of the 2d spectra at each time .except at @xmath161 , the diagonal component @xmath162 is excited spontaneously with a nearly gaussian form , which is in contrast to the initial @xmath125-spectrum consisting of a broad band occupying @xmath163 .thus we conclude that the peaked excitation about the diagonal is not merely a reflection of a similarly peaked initial @xmath125-spectrum .furthermore , the peaking of the spectrum around @xmath164 , combined with eq . , justifies the approximation given in eq . .on the other hand , the gaussian peak is surrounded by a fairly broad spectrum of an order of magnitude smaller amplitude , which is due to the excitation of random velocity fluctuation arising from the small - amplitude forward cascade [ see fig . [fig : coherent ktrans](b ) ] . note that at @xmath161 , the spectrum has a fairly broad peak because the scale of the peak is approaching the system size ( due to a background maxwellian with the thermal velocity @xmath165 ) . in both cases , the spectra share some qualitative features with the runs reported in ref .@xcite ( see also fig .[ fig : forward cascade result ] ) : at large @xmath122 and @xmath125 , the spectral density is broadly distributed about the diagonal , which is expected from the forward cascade of @xmath41 @xcite .however , in each case there is a highly peaked diagonal component ( whose scale is denoted by @xmath69 in sec .[ sec : decay law ] ) that tends to be the dominant contribution to the total value of the invariants at the later stage .this is the component generated by the inverse cascade and is discussed in more detail in the following sections .the time evolution of the dimensionless number @xmath0 is shown for each of the runs in fig .[ fig : dorland number ] . for the runs indexed in table [ run table ] .runs a and e correspond to the strongly collisional case , b corresponds to the marginal case and c , d and f correspond to the weakly collisional case . ] depending on the long - time behavior , we may classify the runs into strongly collisional , marginal and weakly collisional cases as described in sec .[ sec : decay law ] .run b corresponds to the marginal case as @xmath0 approaches a constant ( @xmath166 ) in @xmath167 . runs a and e are strongly collisionalas @xmath0 decreases in time and runs c , d and f are weakly collisional as @xmath0 increases in time . the evolution of @xmath0 differs among weakly collisional runs but approaches the theoretical limit @xmath91 as @xmath0 increases .figure [ fig : ratio law ] shows the time evolution of the ratio of two collisionless invariants for strongly collisional ( runs a and e ) and marginal ( run b ) cases .initially , the ratio is of the order of @xmath168 or @xmath169 depending on the initial velocity distributions [ see eqs . and ] .coherent cases show an initial phase of constant ratio ( @xmath170 for run a and @xmath171 for run b ) , which stems from the fact that the initial condition is almost monochromatic with high-@xmath59 , and that both collisionless invariants initially decay at the same rate due to the collisional damping of the distribution function at the scale @xmath172 .the decay law seems consistent with both strongly collisional and marginal cases at the later stage .the case with the random initial condition tends to take a longer time to approach the theoretical line as the initial ratio , @xmath173 [ see eq .] , is much smaller than the asymptotic ratio @xmath174 .a slight deviation from the power law is observed at the last stage of the simulation for runs a and b , which is due to the fact that the cascade has reached the largest wave length of the system around @xmath175 , due to the finite size of the simulation box .the time evolution of individual collisionless invariants are shown in fig .[ fig : decay law ] for the marginal ( run b ) and weakly collisional ( run d ) cases .the marginal run ( b ) shows a reasonable agreement with the theoretical expectation .the weakly collisional run shows a significantly slower decay than the marginal case , but still faster than the asymptotic ( @xmath86 ) limit .computational resources beyond those available for this study will be needed to unequivocally confirm the asymptotic decay laws .the self - similarity of the spectrum may be investigated with the approach of chasnov for 2d fluid turbulence @xcite .chasnov defined an instantaneous length scale of the decaying turbulence in terms of the ratio of the two invariants , energy @xmath176 and enstrophy @xmath177 .this is possible because of the relationship that hold between them at each scale , namely vorticity @xmath178 is a first - order derivative of velocity @xmath179 . herewe may define an instantaneous length scale in terms of the ratio of the two collisionless invariants @xmath41 and @xmath43 , based on eq . .in general @xmath41 and @xmath43 can be nearly independent due to the extra degree of freedom arising from the velocity space ; however , the present argument is valid when the spectral density is sufficiently concentrated along the diagonal , @xmath180 .we normalize the spectra in velocity space as well as in position space . on dimensional grounds , we may define @xmath181 where @xmath97 and @xmath182 are the inverses of the characteristic scale length in position and velocity space , respectively , determined by [ see eq . ] , @xmath183 the wave number @xmath122 and velocity wave number @xmath125 are normalized to @xmath97 and @xmath182 , @xmath184 and the wave - number and velocity - space spectra are defined by @xmath185 and eq . .when we apply the normalization to simulation results , one would expect a good coincidence for the same value of @xmath0 .namely , if the theory is correct , the normalized spectra at different times should collapse in the time - asymptotic regime of the marginal case ( run b , see also fig .[ fig : dorland number ] ) .we first show the normalized spectra defined by eqs . in fig .[ fig : collapsed spectra for run b ] for run b ( see table [ run table ] ) . ) ., title="fig : " ] ) ., title="fig : " ] ) ., title="fig : " ] as one can easily expect from the 2d spectra ( fig .[ fig : coherent 2d - spectra ] ) and the ratio ( fig .[ fig : ratio law ] ) , the spectral peak moves towards larger scales [ to smaller @xmath59 ] and we may expect self - similar spectra in the later stage of the simulation ( from figs . [fig : dorland number][fig : decay law ] , we see that the asymptotic regime is attained in the range @xmath187 ) .after the ratio @xmath98 approaches the power - law behavior ( @xmath188 , see fig .[ fig : ratio law ] ) , wave - number spectra show promising coincidence which indicates a good self - similarity of the decaying turbulence .notice especially the amazing coincidence of the tail at @xmath189 and @xmath190 , both in the time - asymptotic regime as shown in figs .[ fig : dorland number][fig : decay law ] .we also confirm that the peak of the spectra moves towards large scales roughly with the scaling @xmath191 in this time regime .the last one ( @xmath192 ) is offset by a small amount because of the fact that the spectral peak has reached the system size .the wave - number spectra at the right of the peak show slopes somewhat steeper than the theoretical prediction of the forward cascade @xcite .this agrees with the expectation from 2d fluid turbulence @xcite , as the dimensionless number @xmath0 is not asymptotically large in the marginal case .as @xmath0 becomes large , the spectral slope becomes shallower and the forward cascade spectra with the slope @xmath193 for @xmath194 and @xmath195 for @xmath196 ( see ref .@xcite ) is expected to be recovered .figure [ fig : collapsed spectra for run d ] shows the normalized spectra for run d. ) ., title="fig : " ] ) ., title="fig : " ] ) ., title="fig : " ] comparison with fig .[ fig : collapsed spectra for run b ] clearly indicates that the spectral slopes in fig .[ fig : collapsed spectra for run d ] are indeed closer to the theoretical expectation of the forward cascade , for both spectra of @xmath197 and @xmath198 .collapse of the spectra is as good as fig .[ fig : collapsed spectra for run b ] .however , as @xmath0 continues to grow for @xmath167 ( see fig .[ fig : dorland number ] ) , the spectral slope does change slightly , gradually approaching the theoretical prediction ( @xmath199 and @xmath200 ) as shown in fig .[ fig : collapsed spectra for run d ] . at @xmath201 the spectral peak has reached the system size and the spectra become offset .we presented theoretical and numerical investigations of electrostatic , freely decaying turbulence of weakly - collisional , magnetized plasmas using the gyrokinetic model in 4d phase space ( two position - space and two velocity - space dimensions ) .landau damping was removed from the system by ignoring variation along the background magnetic field .nonlinear interactions introduce an amplitude - dependent perpendicular phase mixing of the gyrophase - independent part of the perturbed distribution function , which creates structure in @xmath202 comparable in size to spatial structure ( see fig . [fig : mixing ] ) .since our 2d ( in position space ) system possesses two collisionless invariants [ entropy , or free energy , see eq . ; and energy , see eq . ] , a dual cascade ( forward and inverse cascades ) takes place when the initial condition consisted of small - scale fluctuations in position as well as in velocity space such as in eqs . and . as the dual cascade proceeds , the peak of the spectra moves towards large scales in both position and velocity space as shown in figs .[ fig : coherent 2d - spectra][fig : random 2d - spectra ]. nonlinear transfer is diagnosed by the direct numerical simulation , which shows a clear evidence of inverse ( forward ) transfer of energy ( entropy ) ( see fig . [fig : coherent ktrans ] ) . in the inverse cascade , the velocity space spectrumis highly focused due to the fact that energy comes from coherent structure in the velocity space [ see eq . and fig .[ fig : slices along p ] ] , which is in contrast to the broad distribution of velocity scales excited at each wave number in the forward cascade @xcite .following an example from 2d navier - stokes turbulence @xcite , a phenomenological theory of decay is presented ( sec .[ sec : decay law ] ) as well as the numerical simulation ( sec .[ sec : num decay law ] ) .several types of asymptotic decay have been identified in numerical simulation , which match up well with the phenomenological theory using a classification based on the kinetic dimensionless number @xmath0 [ see eq . ] . when @xmath0 takes a marginal value , decay laws of both invariants are identified[ see eq . , figs .[ fig : dorland number ] and [ fig : decay law ] ] . in the weakly collisional regimethe invariants decay more slowly [ see fig . [fig : decay law ] ] ; and in the asymptotic limit where the collision frequency becomes negligible ( but finite ) , the entropy ( or free energy ) decays as @xmath203 while the energy stays constant [ see eq . ] . in this paper we focused on the time - asymptotic regime of the freely decaying turbulence .although there is a range of behavior depending on the strength of dissipation , the cases are unified by some common features .the most prominent of these features is the dual cascade , whereby the invariant @xmath43 cascades inversely to large scales while the invariant @xmath41 cascades to small scales .the transient and driven cases , hosts to a broader range of phenomena , have recently been explored in other works @xcite .numerical simulations were performed at nersc , ncsa , tacc , and iferc csc .the authors thank drs .a. schekochihin and s. cowley for fruitful discussions .this work was supported by the u.s .doe center for multiscale plasma dynamics , the leverhulme trust international network for magnetised plasma turbulence , wolfgang pauli institute , and jsps kakenhi grant number 24540533 .the collision operator that we use in our numerical simulations is a linearized model operator described in refs .it consists of pitch - angle scattering and energy diffusion , augmented by a few correction terms introduced for conservation of momentum and energy .symbolically it is written as @xmath204 where @xmath205 and @xmath53 denotes the diffusion - type operator in pitch angle and energy , respectively , @xmath206 and @xmath207 are momentum restoring terms associated with @xmath205 and @xmath53 , respectively , and @xmath43 is the energy restoring term associated with @xmath53 .specifically , @xmath208,\ ] ] and @xmath209 where @xmath210 and @xmath211 are collision frequencies and @xmath212 is the pitch angle .it is noted that @xmath205 handles the smoothing in @xmath213 space while @xmath53 does it for energy , so with both of them we can cover the smoothing in the whole two - dimensional plane of @xmath6 and @xmath7 .the detailed expression of collision frequencies and conservation terms are shown in refs .@xcite .essential features of the collisionless dynamics may be described in a 3d phase space @xmath214 , however , we use four phase - space dimensions @xmath215 to account for the fact that collisions isotropize the velocity space . herethe hankel transform is defined by the three - dimensional velocity integral .when there is no dependence on gyroangle and maxwellian dependence on @xmath7 , we will obtain the standard hankel transform with respect to @xmath6 times @xmath216 .

<S> in magnetized plasmas , a turbulent cascade occurs in phase space at scales smaller than the thermal larmor radius ( `` sub - larmor scales '' ) [ phys . </S><S> rev . </S><S> lett . * 103 * , 015003 ( 2009 ) ] . when the turbulence is restricted to two spatial dimensions perpendicular to the background magnetic field </S><S> , two independent cascades may take place simultaneously because of the presence of two collisionless invariants . in the present work , freely decaying turbulence of two - dimensional electrostatic gyrokinetics </S><S> is investigated by means of phenomenological theory and direct numerical simulations . a dual cascade ( forward and inverse cascades ) </S><S> is observed in velocity space as well as in position space , which we diagnose by means of nonlinear transfer functions for the collisionless invariants . </S><S> we find that the turbulence tends to a time - asymptotic state , dominated by a single scale that grows in time . </S><S> a theory of this asymptotic state is derived in the form of decay laws . </S><S> each case that we study falls into one of three regimes ( weakly collisional , marginal , and strongly collisional ) , determined by a dimensionless number @xmath0 , a quantity analogous to the reynolds number . </S><S> the marginal state is marked by a critical number @xmath1 that is preserved in time . </S><S> turbulence initialized above this value become increasingly inertial in time , evolving toward larger and larger @xmath0 ; turbulence initialized below @xmath2 become more and more collisional , decaying to progressively smaller @xmath0 . </S>

1309.5076

as computers become more powerful , there is a move to numerically simulate larger and larger model networks of neurons @xcite . while simulation is useful for confirming observed behavior it is not as helpful in determining the mechanisms underlying the behavior .the tools of dynamical systems theory , such as bifurcation analysis can be useful in this regard when studying single neuron or small network models .however , they are not viable for large networks , especially if the neurons are not identical .thus , a common approach is to try to extrapolate large network behavior from detailed analysis of the behavior of individual cells or small networks @xcite .this can be problematic as networks can have behavior that is not present in individual cells .for example , individual neurons that are only capable of tonic firing when isolated may burst when coupled in a network @xcite .further , large networks may exhibit behavior not present in smaller networks .for example , dur - e - ahmad et al . studied bursting in networks ranging in size from 2 cells to 100 .they found that bursting occurred in a larger range of parameters for larger networks ( * ? ? ?* figure 7 ) . given the role of bursts in networks of neurons ,it is important to understand how a network transitions ( bifurcates ) from a non - bursting behavior , to bursting .bursting has been suggested to be a fairly important and information dense firing mode for neurons .for example , single bursts can induce long term potentiation and depression in the hippocampus , which are important proccesses for learning and memory @xcite .additionally , bursts have been found to carry more information about an animals position in space then isolated spikes alone as place fields have been found to be more accurately defined when considering bursts alone @xcite . additionally , the mechanism we analyze in this paper , adaptation induced bursting has also been suggested as a biologically plausible mechanism for the generation of grid cells via oscillatory interference of the bursting @xcite . however , adaptation induced bursting is a network level phenomenon and we can not apply bifurcation analysis directly to a network of adapting neurons .mean - field theory offers one approach to bridge this gap . in applying this theory ,one usually derives ( or suggests ) a low dimensional system of differential equations that govern the moments of different variables across the network @xcite .for example , for a network of all - to - all coupled izhikevich neurons @xcite , one can derive a two dimensional system of differential equations for the mean adaptation variable , and the mean synaptic gating variable @xcite .mean - field systems enable one to conduct network level bifurcation analysis and hence to test different hypotheses about large network behavior .for example , hemond et al .@xcite found that when uncoupled , the majority of hippocampal pyramidal neurons in region ca3 do not display bursting .this is contradictory to the observation that burst firing is ubiquitous this region ( * ? ? ?* section 5.3.5 ) .a possible explanation for these contradictory observations is that bursting is a network level phenomenon .this hypothesis has been tested using bifurcation analysis of the mean - field system derived from a network of izhikevich neurons @xcite .in particular , it was shown that for a network of identical all - to - all coupled neurons fit to experimental data from @xcite for ca3 pyramidal cells , bursting occurs for a large range of synaptic conductances and applied currents if the spike frequency adaptation in the neurons is sufficiently strong .hemond et al .@xcite also observed that the pyramidal neurons in their study were heterogeneous , in particular , the neurons had different degrees of spike frequency adaptation . when the study of @xcite was extended to a network of two homogeneous subpopulations of hippocampal neurons with different degrees of spike frequency adaptation ,the mean - field equations predict , and numerical simulations confirm , that the region in the parameter space where bursting occurs decreases in size @xcite .this would seem to indicate that adaptation induced bursting may not be robust to heterogeneity , however , it is unknown how robust this network level bursting is to heterogeneity in different parameters .an extension of the mean - field system in @xcite to large networks of heterogeneous neurons , is needed to fully analyze the robustness of adaptation induced bursting .the application of mean - field analysis to large networks of heterogeneous neurons has been limited to networks of one dimensional integrate and fire neurons with heterogeneity in the applied current@xcite .hansel and mato @xcite analyze a network of all - to - all coupled quadratic integrate and fire neurons consisting of two subpopulations : one excitatory and one inhibitory .they showed analytically that the tonic firing asynchronous state can lose stability either to a synchronous tonic firing state or a bursting state .vladimirski et al .@xcite analyze a network of all - to - all coupled linear integrate and fire neurons subject to synaptic depression .this model was used to study network induced bursting in the developing chick spinal cord .their derivation of the mean - field model is based on temporal averaging of the fast synaptic gating variable , in addition to the usual network averaging .this results in a model that only involves the distribution of slow synaptic depression variable and the distribution of firing rates .vladimirski et al .note that , for their model , increased heterogeneity tends to make population induced bursting more robust .they also note that one can not understand the behaviour of their network with a single slow , network averaged synaptic depression variable .the motivation for the present paper is to explore the effect of heterogeneity in parameters on network induced bursting when adaptation is the primary negative feedback acting on the individual firing rates . to this end, we introduce a set of mean - field equations for networks of heterogeneous two - dimensional integrate and fire neurons .while the specific neural model we consider is for neurons with adaptation , our derivation is quite general and could be applied to other integrate and fire models .in contrast with @xcite our derivation of the mean - field model does not use temporal averaging thus we end up with a mean - field system which involves the network averaged synaptic gating variables as well as the distribution of adaptation variables .we also allow for heterogeneity in more than one parameter .our approach is a generalization of that used for homogeneous networks of two dimensional integrate and fire neurons @xcite , however , in the heterogeneous case it turns out that there are actually multiple mean - field models , as different assumptions can be made during the derivation .this leads us to three distinct mean - field systems , each derived under different assumptions , and used for different purposes .together , these sets of equations allow us to do bifurcation analysis on large networks , as in the homogeneous case .we show that the bifurcation structure of the heterogeneous network differs both qualitatively and quantitatively from the homogeneous network case .we discuss the implications of this for network induced bursting in the hippocampus .when considering a homogeneous network , the mean - field variables are a good approximation for the variables of every neuron .however , this is not the case for a heterogeneous network .if the heterogeneity is large , then the neuron variables may be also widely distributed , rendering information about the first moments less useful .this also implies that the behavior of any individual neuron is less predictable with a mean - field system than it was in the homogeneous case .one of our mean - field systems addresses this problem , giving information about the distributions of the variables instead of just the mean . when considering a model for a specific heterogeneous network of neurons , one stumbling block is determining the distribution of parameters .estimates of the distribution can be made through direct intracellular recording of a sufficient number of neurons , and conventional measurements of the biophysical properties ( membrane capacitance , voltage threshold , etc . ) .unfortunately , this is a very time consuming and intensive process .what is needed is a way of measuring the biophysical parameters of multiple neurons simultaneously .there are a few ways to sample multiple neurons such as multi - unit recordings using tetrode arrays , or two - photon microscopy techniques @xcite .however , these techniques typically only tell us about spike times of large ( dozens to hundreds of neurons ) networks @xcite . while an impressive accomplishment, this still does not tell us anything directly about the biophysical properties of the neurons that caused those spikes . in this paperwe use a mean - field system to determine an approximate distribution of firing rates for a network given a known distribution of parameters .this is an unusual state of affairs for a mean - field system , as these kinds of systems seldom give information about entire distributions .more importantly however , using a mean - field system , we can invert a distribution of steady state firing rates ( which can be obtained from multi - unit recordings ) to obtain a distribution of parameters .in fact , this can be done at the individual neuron level , to determine the parameter value for any particular neuron .this allows one to estimate different biophysical parameters , which are difficult to measure at the network level , using easy to measure firing rate distributions .however , the assumptions required for the numerical accuracy of the estimation are fairly strong .the plan for our article is as follows .section 1.1 introduces the general class of adapting two - dimensional integrate and fire neurons used in our network models .this class was introduced by touboul @xcite , who also completed the bifurcation analysis of the single , uncoupled neuron .population density methods are briefly introduced in section 1.2 , as a population density equation serves as an intermediate step to obtain our mean - field models .section 2 begins with a review of mean - field theory and the equations for the homogeneous network .this is followed , in sections 2.1 - 2.3 , by the derivation of the three mean - field systems for the heterogeneous network .a comparison of numerical simulations of these mean - field systems and the full network is the subject of section 2.4. applications of mean - field theory to networks with a single heterogeneous parameter can be found in section 3 including bifurcation analysis ( section 3.1 ) , distributions of parameters and firing rates ( section 3.2 ) and using mean - field theory for parameter estimation from firing rate data ( section 3.3 ) .applications of mean - field theory to networks with multiple sources of heterogeneity are included in section 4 .a discussion of our work and its implications can be found in section 5 .we consider a network of two - dimensional integrate and fire models of the form @xmath0 where @xmath1 represents the nondimensionalized membrane potential , and @xmath2 serves as an adaptation variable .time has also been non - dimensionalized .the dynamical equations ( [ de1])-([de2 ] ) are supplemented by the following discontinuities @xmath3 this particular notation was formally introduced by touboul @xcite , along with a full bifurcation analysis of this general family of adapting integrate and fire neurons .members of this family include the izhikevich model @xcite , the adaptive exponential ( adex ) model @xcite , and touboul s own quartic model @xcite .the methods of this paper can be applied to a network of any particular neuron belonging to this general family , and thus all derivations are done for this model . for the numerical examples , however , we only consider the izhikevich neuron : @xmath4 in dimensionless form , this is given by - with @xmath5 in addition to dimensionless versions of the discontinuities .the application to other neural models is straight forward , see @xcite where the homogeneous mean - field theory has been derived and tested for both the adex and the izhikevich models .networks of these neurons can be coupled together through changes in the synaptic conductance .the synaptic conductance of post - synaptic neuron @xmath6 due to presynaptic neurons @xmath7 is given by @xmath8 where @xmath9 denotes the maximal synaptic conductance of neuron @xmath6 and @xmath10 denotes the total proportion of postsynaptic ion channels open in the membrane of neuron @xmath6 .the time dependent variable @xmath11 represents the proportion of postsynaptic ion channels open in the membrane of neuron @xmath6 as a result of the firing in neuron @xmath12 . the changes in @xmath11 that occur after a spike are often modeled as transient pulses . for example ,if neuron @xmath12 fires its @xmath13th action potential at time @xmath14 , then the variable @xmath11 at time @xmath15 is given by @xmath16 there are different functions proposed for @xmath17 in the literature including the simple exponential , the alpha synapse , and the double exponential .we primarily consider the simple exponential synapse @xmath18 which is governed by the ordinary differential equation @xmath19 in the rest of the paper , we assume all - to - all connectivity and that the synaptic parameters @xmath20 , and @xmath21 are the same for every synapse . in this casewe may set @xmath22 for all @xmath6 , as each postsynaptic neuron receives the same summed input from all the presynaptic neurons .then , using and , the network of all - to - all coupled neurons that we consider is given by the following system of discontinuous ode s : @xmath23 @xmath24 for @xmath25 .in the examples , we consider one or more parameters as the sources of heterogeneity .however , to simplify the notation in the derivations , we use the vector @xmath26 to represent all the heterogeneous parameters . then , denoting the state variables @xmath1 and @xmath2 as the vector @xmath27 ,we can write the equations for the individual oscillator as @xmath28 given a specific heterogeneous parameter , @xmath29 and @xmath30 may not depend on @xmath26 , or all of the components of @xmath26 .however , for the sake of simplicity , we include the dependence in both equations .our numerical examples are restricted to the izhikevich neural model , and we primarily consider the driving current @xmath31 of each neuron , the synaptic conductance @xmath32 and the adaptation jump size , @xmath33 as the source of heterogeneity .however , the mean - field equations we derive can be applied to any of the two - dimensional adapting integrate and fire models , with any heterogeneous parameter or set of parameters .finally , we note that in many applications , @xmath34 is a small parameter , and thus the @xmath35 term can be dropped in @xmath30 .we do this in all our numerical studies .however , one can can still derived appropriate mean - field equations if this term is present ( see discussion in @xcite ) , and thus we have left the term in the derivations .the population density function , @xmath36 determines the density of neurons at a point in phase space , @xmath27 , at time @xmath15 .consider first the case of a homogeneous network , i.e. , all the oscillators have the same parameter values , denoted by @xmath26 . in the limit as @xmath37 , one can derive the following evolution equation for the population density function : @xmath38 where @xmath39 is given by @xmath40 and must satisfy the boundary condition @xmath41 in the same limit , the differential equation for @xmath42 converges to @xmath43 where the integral term is actually the network averaged firing rate , which we denote as @xmath44 .derivations of equation ( [ continuityequation ] ) can be found in various sources @xcite .equation is frequently referred to as the continuity equation and it has various applications besides its use as an intermediate step in mean - field reductions .for example , the equation has been used to determine the stability of the asynchronous tonic firing state by various authors @xcite .these papers predominantly consider homogeneous networks of linear integrate and fire neurons .the exception is the work of hansel and mato @xcite which considers heterogeneity in the applied current .one can study stability of various firing states using spectral analysis or other analytical treatments of this equation @xcite .however , these approaches are too complicated for the models we consider in this paper . now consider a heterogeneous network where the parameters vary from oscillator to oscillator , but are static in time. then one can rewrite the equations for the individual oscillator as @xmath45 in this case the flux contribution due to @xmath26 is 0 , and the evolution equation for the network is given by @xmath46 the density now has the vector of parameters , @xmath26 , as an independent variable .the flux consists of the vector @xmath47 , with @xmath26 as an independent variable , as opposed to a fixed constant .if the parameters are time varying however , the final component of the flux will be non - zero .the equation for @xmath42 is also different in the heterogeneous case : @xmath48 while the evolution equation is an exact representation for the network in the large network limit , it is difficult to work with analytically. additionally , as the dimensions of the pde become large , it becomes difficult to find numerical solutions efficiently @xcite .however , mean - field reductions of the network can be used to reduce the population density pde to a system of nonlinear switching odes that governs the moments of the distribution .unlike the pde , the system of odes is tractable using bifurcation theory , at least numerically .furthermore , we show that in the heterogeneous case , the resulting mean - field systems can yield more information than just the type of bifurcation that the network can undergo . in the homogeneous case ,the mean - field system of equations for an all - to - all coupled izhikevich network was derived in @xcite .we present here a quick summary of this derivation . in order to derive a mean - field system of equations ,one first needs to reduce the pde for @xmath49 by a dimension .this is done by first writing the density in its conditional form : @xmath50 and then integrating the continuity equation with respect to @xmath2 .this yields the one dimensional pde @xmath51 where the flux has been redefined to @xmath52 one can now make a first order moment closure assumption , @xmath53 , and derive an approximate ode for @xmath54 , which yields the system @xmath55 where the subscript on the density function has been dropped for convenience .the details , and validity of the first order moment closure assumption that is used can be found in @xcite .we note however that the work in @xcite was primarily with leaky integrate and fire neurons , as opposed to the two - dimensional adapting class we consider here .however , it is a necessary assumption to proceed analytically .if we assume that the adaptation time constant , @xmath56 is large , one can apply a quasi - steady state approximation to derive a system of switching ode s for @xmath54 and @xmath42 : @xmath57^{-1 } & : h(\langle w \rangle , s)\geq 0 \\ 0 & : h(\langle w \rangle , s)<0 \end{array } \right.\label{switch1 } \end{aligned}\ ] ] the switching manifold for the system , @xmath58 is given by : @xmath59 note that @xmath58 depends on the parameter(s ) of the model , and thus for the heterogeneous case , we make this dependence explicit by writing @xmath60 .as the computation for @xmath61 is somewhat lengthy and is only outlined in the discussion of @xcite , we have placed it in appendix a. for the izhikevich neuron , equations - become @xmath62^{-1 } & : h(\langle w \rangle , s)\geq 0 \\ 0 & : h(\langle w \rangle , s)<0 \end{array } \right.\label{izswitch1 } \\ h(\langle w \rangle , s ) & = & i - \langle w \rangle -\frac{(\alpha+gs)^2}{4 } + ge_rs . \label{hiz } \end{aligned}\ ] ] note that in this case , we can evaluate @xmath63 explicitly : @xmath64 in addition to an approximation to @xmath65 @xmath66 a comparison of solutions of these equations and the full network are shown in figure [ fig1 ] for both the tonic firing and the bursting case .this system of equations is valid when @xmath67 , however the magnitude of @xmath21 is also significant . while in the original derivation of @xcite , @xmath68 was suggested as a criterion for validity , this is not actually necessary .one merely requires that @xmath21 not be significantly smaller then @xmath69 , the time scale of the pde .the reason for this is that if the time scale of the ode for @xmath42 is smaller than that of the pde then the quasi - steady state approximation must be applied to the ode for @xmath42 as well .the requirements on the time constants are carried forward in the heterogeneous case . in our models ,the timescale of the ode for @xmath42 is typically between that of the pde and that of the ode for @xmath2 , and we have not applied the quasi - steady approximation to @xmath42 .applying a quasi - steady state approximation to both @xmath42 and the reduced pde yields a more compact system , which is just an ode for @xmath70 , however the analysis does not get any simpler .the reason for this is two - fold : the ode for @xmath54 remains non - smooth , and the firing rate now has to be implicitly solved at each time step .thus , it is more convenient to apply the quasi - steady state approximation only to the partial differential equation .when parameter heterogeneity is added into the mix , it turns out that there are multiple mean - field " systems of equations that can be derived , by applying different assumptions on the conditional moments .we outline three different assumptions that can be made , and the resulting system of mean - field equations in each case .we begin by writing out the density function in the conditional form @xmath71 the continuity equation is then given by @xmath72 simple integration with respect to @xmath26 yields the reduced continuity equation @xmath73 this step is valid for all the non - dimensionalized models we consider as they are all linear in their dimensionless parameters ( see @xcite ) .the flux has also been redefined upon integration to @xmath74 we now apply the moment closure assumption @xmath75 to yield the following pde : @xmath76 it should be clear that this is equivalent to the continuity equation for a homogeneous network with parameter values fixed at @xmath77 .thus , the associated mean - field system is identical to the homogeneous case , only with the parameters fixed at @xmath78 .this is the simplest assumption one can make in the heterogeneous case .for example , if we treat @xmath79 as the source of heterogeneity for a network of izhikevich neurons , with distribution @xmath80 , then the resulting mean - field system is @xmath81 note that @xmath79 in equations - has been replaced by @xmath82 in equations ( [ switch12])-([vm1 ] ) .we treat this system as the baseline mean - field model for comparison purposes , in addition to direct numerical simulations of the network , and we denote this system of equations as mean - field one ( mfi ). we should expect this system to be an adequate approximation to the actual network for narrowly centered distributions of the parameter heterogeneity ( small values of the variance , @xmath83 ) .this set of differential equations is representative of a common approach taken when fitting actual neurons . in this approach , multiple estimates of parameters or measurements taken from multiple neuronsare averaged to yield a single parameter value , which is really the mean parameter value , @xmath84 .simulations of homogeneous , large networks are then run with the parameters fixed at their mean values . as we shall see in subsequent sections ,the behavior of a simulated heterogeneous network can differ substantially from that of mfi .to derived our second mean - field system , we begin by writing the density function in the alternative conditional form @xmath85 next we integrate the continuity equation with respect to @xmath2 .this yields the following system @xmath86 where the last term vanishes as @xmath87 is assumed to be vanishing on the boundary , and @xmath88 we now make the first order moment closure assumption @xmath89 . then to complete the system, we must derive a differential equation for @xmath54 : @xmath90 note that we have made the approximation @xmath91 in addition to dropping the @xmath92 terms .additionally , the substitution on line comes from the boundary condition . applying a quasi - steady state approximation to the pde yields the following equation for the steady state voltage independent flux , @xmath93 : @xmath94^{-1}\rho_\beta(\bm\beta ) & \text{if } \quad h(\langle w \rangle , s,\bm\beta)\geq 0\\ 0 & \text{if}\quad h(\langle w \rangle , s,\bm\beta)<0 \end{cases}. \label{jbeta}\ ] ] we interpret the ratio @xmath95 as the parameter dependent ( or conditional ) network averaged firing rate , @xmath96 , based on the fact that @xmath97 in other words , the distribution of parameters induces a distribution of firing rates across the network , and the network averaged firing rate is the mean of the distribution . in summary ,the resulting mean - field equations are given by : @xmath98^{-1 } & : h(\langle w \rangle , s,\bm\beta)\geq 0 \\ 0 & : h(\langle w \rangle , s,\bm\beta)<0 \end{array } \right.\label{sswitch1 } \\h(\langle w \rangle , s,\bm\beta ) & = & i -\langle w \rangle+ \min_{v}(f(v ) + g(e_r - v)s ) \label{hmfii}\\\langle v\rangle & = & \int_{\bm\beta}\langle v|\bm\beta \rangle \rho_{\beta}(\bm\beta)\,d\bm\beta\label{vm2}\end{aligned}\ ] ] where the forms of @xmath99 and @xmath100 depend on which specific neural model is used , and the equation for @xmath101 can be found in appendix a. note that the distribution of firing rates is not computed explicitly in these equations , only the conditional firing rates , @xmath102 , are computed. however , we show in section [ mf3sec ] that a distribution for the steady state firing rates of the network can be computed using @xmath96 .we refer to equations - as mean - field two ( mfii ) .it should be noted that mfi and mfii effectively differ in the order in which the integrations are carried out . in mfi ,we integrate with respect to @xmath26 first , and then apply the first order moment closure assumptions @xmath103 and @xmath53 . in mfii, we integrate with respect to @xmath2 first , and then apply the moment closure assumption @xmath104 . furthermore, if @xmath105 does not actually depend on the heterogeneous parameter @xmath26 , such as when the heterogeneity is in @xmath33 , then mfi and mfii are identical .in fact the first order moment closure assumption used here can be weakened .this leads to the mean - field " system in the next subsection , which is a different kind of system than mfi and mfii .suppose that instead of assuming that @xmath89 , we make the weaker assumption that @xmath106 .it turns out that this assumption yields a pde , even when one makes the quasi - steady state approximation , as we now show . applying this weaker moment closure assumption to yields the following simplification of the continuity equation : @xmath107 application of the quasi - steady state approximationnow yields @xmath108^{-1}\rho_\beta(\bm\beta ) & : h(\langle w|\bm\beta \rangle , s,\bm\beta)\geq 0 \\ 0 & : h(\langle w|\bm\beta \rangle , s,\bm\beta)<0 \end{array } , \right.\label{sswitch12 } \\w|\bm\beta \rangle , s,\bm\beta ) & = & i -\langle w|\bm\beta \rangle + \min_{v}(f(v ) + g(e_r - v)s).\end{aligned}\ ] ] an equation for the time variation of @xmath109 can be derived in a similar manner to the last section , yielding the following mean - field system : @xmath110^{-1 } & : h(\langle w|\bm\beta\rangle , s,\bm\beta)\geq0 \\ 0 & : h(\langle w|\bm\beta\rangle , s,\bm\beta)<0 \end{array } \right.\label{mf3r } \end{aligned}\ ] ] note that @xmath54 can be computed via : @xmath111 we denote this system as mean - field three ( mfiii ) .note that the equation for @xmath112 is actually a pde .thus , while this system should be more accurate than mean - field ii , it has the drawback of being more difficult to analyze .the dependence on @xmath26 forces one to discretize over a mesh in @xmath26 in order to work numerically with this system .this makes numerical bifurcation analysis more difficult .however , as we shall show later , an approach that yields similar information to direct bifurcation analysis can be used with mfiii . the equation for @xmath101can once again be found in the appendix .recall that simulations of a homogeneous network and the corresponding mean - field system are shown in figure [ fig1 ] .note that the network undergoes a bifurcation from tonically firing to bursting as the amount of applied current @xmath113 is decreased , with all other parameter values held fixed .further simulations show that if @xmath113 is decreased below @xmath114 then all neurons in the network are quiescent ( non - firing ) . to determine and compare the validity of the three mean - field systems, we have run a series of numerical simulations of these systems and of an actual network containing 1000 neurons .the parameter values for the individual neurons can be found in table [ table1 ] .they are based on those given in @xcite which were fit to data for hippocampal ca3 pyramidal neurons from @xcite .these are the parameter values we use for the rest of this paper , unless otherwise indicated . to begin , we consider heterogeneity only in the applied current .the distributions are assumed to be normal with mean @xmath82 and standard deviation @xmath115 .we varied the values of the mean and standard deviation and found that the accuracy of the mean - field approximations depends on where the mean is relative to the different bifurcation regions and on the size of the standard deviation .[ cols= " < , < , < , < " , ]building on the mean - field framework for networks of homogeneous oscillators , we extended the mean - field approach to networks of heterogeneous oscillators .this was accomplished through the derivation of three separate mean - field systems , mfi , mfii and mfiii , with differing applications and regions of validity .we successfully applied numerical bifurcation analysis to mfi and mfii to aid in the understanding of the different behaviors that heterogeneous networks can display , and how they transition between these different types of behaviors .more importantly however , we have surpassed the natural limitation of mean - field systems : that they can only provide information about the first moments . with a few additional tools , we used mfiii to derive information about distributions of firing rates , and even parameters , given some basic knowledge .other researchers @xcite have derived firing rate distributions for heterogeneous networks , however these have been derived under differing assumptions .for example , the heterogeneous mean field systems studied by hansel and mato ( equations ( 5.5)-(5.7 ) in @xcite ) have similar integral terms to our mfii however they are firing rate models .our models are current / conductance based models .the difference between these two types of equations arises from which time scale is the fastest , that of the synaptic current , or the firing rate .if the firing rate time scale is assumed to be the fastest , then a differential equation for the synaptic current can be obtained ( as in our case ) .if the time scale of the synaptic current is assumed to be the fastest , then one obtains firing rate equations , as in @xcite .the fact that these two different limits result in different kinds of equations were first highlighted in @xcite ( section 7.2 ) .additionally , no adaptation is contained in the rate models in @xcite . finally , it is likely that the firing rate models shown in @xcite can not display period doubling bifurcations as they have a similar structure to mfii , which misses out on the more complicated bifurcations of the actual network that mfiii can reproduce , due to its pde nature .the model of vladimirski et al .@xcite is formulated in terms of an input - output relation for the synaptic conductance , so has a different structure than ours .it involves a distribution of the synaptic depression variable so has some aspects similar to our mfiii , however , no pde governing the evolution of this variable is derived .dur - e - ahmad et al .@xcite studied adaptation induced bursting in a network of homogeneous izhikevich neurons , with parameters determined from experimental data on ca3 pyramidal neurons .they showed that , if the adaptation is strong enough , network bursting occurs in large regions of the parameter space consisting of the synaptic conductance , @xmath116 , and the applied current , @xmath113 .in @xcite we showed that the transition from tonic firing to bursting involves a saddle - node bifurcation of non - smooth limit cycles , followed by a grazing bifurcation and a subcritical hopf bifurcation .for fixed @xmath113 greater than rheobase but sufficiently small , there is one transition from tonic firing to bursting at a low @xmath116 value and another from bursting back to tonic firing at a higher @xmath116 value .thus the bursting region is a closed semi - circular region in the @xmath116 , @xmath113 parameter space . in @xcitewe showed that the size of this bursting region is reduced if the network is split into two homogeneous subnetworks , one strongly adapting and one weakly adapting . here, we used the tools we developed to investigate how this adaptation induced network bursting is affected by heterogeneity in the parameters .somewhat surprisingly , we have found that adaptation induced network bursting is not very robust to heterogeneity .this has been confirmed by direct simulations of the full network , bifurcation analysis using mfii and analysis of the proportion of bursting neurons in the network using mfiii .this lack of robustness is caused by two changes to the homogeneous case : 1 .the low @xmath116 hopf bifurcation point moves towards higher values , thereby decreasing the size of the bursting region 2 .the low @xmath116 hopf bifurcation switches from subcritical to supercritical .this has two effects : * the bifurcation direction changes , eliminating the bursting at conductance values less than the bifurcation value * the initial limit cycles created by the bifurcation are small amplitude oscillations in the firing rate as opposed to full bursts , thus the transition to bursting moves to even higher conductance values .further , in networks with both weakly and strongly adapting neurons , heterogeneity caused the high @xmath116 hopf bifurcation value to decrease when the proportion of strongly adapting neurons is reduced .let us now put our results in the context of experimental results on the ca3 region .bursting is often seen in these studies ( * ? ? ?* section 5.3.5 ) . when the neurons have their synaptic inputs blocked however , the majority ( @xmath117% ) of these pyramidal neurons do not display bursting , but different degrees of spike frequency adaptation @xcite .thus , it would seem that adaptation induced network bursting should play a role in the ca3 network .however , the biophysically important part of the parameter region is in the low @xmath116 region @xcite . when this fact is taken in conjunction with our results described above, this would seem to weaken the case that adaptation induced network bursting is the only source of bursting in ca3 networks .some other mechanism seems necessary . in their study of hippocampal ca3 pyramidal neurons , hemond and colleaguesnote that roughly 20% of pyramidal neurons were intrinsically bursting .that is , the neurons burst without any synaptic input for some input current .it may by possible that a small subpopulation of intrinsically bursting neurons can facilitate bursting in the rest of the network , however this would depend on the conductance values connecting this particular subpopulation to the rest of the network .this hypothesis can be tested relatively easily using a mean - field approach .all that is required is to fit a two - dimensional adapting model to the intrinsically bursting neurons .this is feasible , as has been previously noted , all the two - dimensional adapting neurons can be turned into intrinsically bursting neurons by simple parameter changes @xcite .the conductance parameter connecting this subpopulation to the rest is best treated as a bifurcation parameter , with some estimate of the range in which it lies in from physiological data . while an intrinsically bursting subpopulation is the most promising avenue of study with regards to hippocampal bursting , synaptic depressionhas also been shown to induce bursting in oscillators that can not otherwise display this behavior . in a model of the developing chick spinal cord, @xcite found that heterogeneity actually makes the bursting more robust , as opposed to less as we have found .thus it is possible the synaptic depression induced bursting is more robust to heterogeneity than adaptation induced bursting .however , in this study the heterogeneity was via a uniform distribution in the applied current ( as opposed to the gaussian distributions we consider ) and typically @xmath118 was close to rheobase , which could also be factors in their results .in addition to area ca3 in the hippocampus , adaptation induced bursting has also been suggested as a possible mechanism for the generation of velocity controlled oscillators ( vco s ) in the entorhinal cortex by @xcite .the vco s burst at frequencies that vary with the velocity of the animal . when a subset of vco s signals are linearly added to a readout neuron , an interference pattern emerges and a grid cell is formed .@xcite use a recursively coupled network of homogeneous izhikevich neurons with adaptation variables given by @xmath119 , and @xmath120 , and these parameters are strong enough so that adaptation induced bursting can occur .the network acts a single velocity controlled oscillator , and the burst frequency is set to vary with the velocity of an animal .this is done by fixing the @xmath116 parameter at a specific value , and inverting the @xmath121 curve where @xmath122 is the frequency of bursts , and @xmath79 is the homogenous applied current to each neuron .grid cells can be generated by using multiple networks and linearly adding their output currents to a read - out neuron .this was done under uncorrelated noisy inputs arriving to each neuron . however , @xcite state that synchrony in the noise ( which can come from the animals velocity signal for example ) coming to each vco network can disrupt grid - cell formation . here , we have shown that a heterogeneous network of oscillators can still maintain a network level oscillation rate , even if the individual neurons have different behaviors .the network level behaviors are predicted from the mean - field systems .given the fact that the individual neurons are heterogeneous , any synchronized noise input into the individual neurons should become increasingly desynchronzied by the differing resposnes of the individual neurons . as a network level oscillation exists and the heterogeneity will likely desynchronize any noise coming to the individual oscillators ,this is a plausible means of generating velocity controlled oscillators .we leave this particular application of mean - field theory for future work . in either of these applications ,a mean - field system may yield valuable insights as to the mechanisms of bursting and the parameter regions they occur in . by carefully choosing the appropriate bifurcation parameters , and accounting for the level of heterogeneity in the neurons in the network , one can determine the bifurcation types and behaviors neurons in these different networks display , in addition to estimates of the different distributions to yield insights about the real cells .when the quasi - steady state approximation is applicable , a very convenient method emerges for computing the moments of @xmath1 .in particular , the quasi - steady state approximation not only yields the steady state flux , but it also yields the steady state density : @xmath125 where @xmath126 is the steady state flux ( @xmath63 for @xmath127 ) . obviously with the density , we can compute the moments of any arbitrary function of @xmath1 .for example , @xmath61 is given by @xmath128 however , once we cross the switching manifold , equation ( [ mevo1 ] ) is no longer valid .an approximation has to be made to compute @xmath123 in this region .in particular , we will assume that @xmath129 , and that the dynamics of @xmath123 are fast relative to @xmath42 and @xmath54 .if so , then the dynamics of @xmath61 are approximately given by @xmath130 and we can solve the steady state equation for @xmath61 @xmath131 however , we have to be careful here .based on the assumptions on @xmath132 we know that there are only two solutions to this system , and we need to solve for the stable one . for the izhikevich neuron for example ,this is given by @xmath133 computation of the integral [ mev1 ] for the izhikevich neuron , in conjunction with equation [ mev2 ] yields the following equation for @xmath61 over the entire @xmath134 plane : @xmath135 the same formulas can actually be used for @xmath136 , with the interpretation that they are now explicit functions of @xmath26 in particular , one can show that under tonic firing , @xmath137 is given by the same formula as @xmath61 when @xmath138 . additionally , the formula for @xmath61 for @xmath139 is also a satisfactory approximation for @xmath101 .to once again reiterate that while the formula for the mean @xmath101 under heterogeneity is the same as the formula for @xmath123 under homogeneity , this does not imply that @xmath140 .in fact , to compute @xmath123 , one has to compute the traditional integral for it : @xmath141performing direct simulations of a network is fairly straight forward , but it is somewhat difficult to use the results of the simulations to automatically classify a given neuron , let alone the entire network , as bursting or tonically firing .we considered various classifiers and opted to use the ratio of the largest to the smallest interspike interval for a neuron when the network has reached steady state . for a bursting neuron ,the ratio of its largest interspike interval ( which is the interburst interval ) to its smallest interspike interval should be large .thus for the @xmath142 neuron we define @xmath143 where the max / min is determined after a suitable time period of steady state behaviour ( either bursting or tonically firing ) .we set a critical ratio , @xmath144 , and classify neuron @xmath6 as bursting if @xmath145 is higher than this ratio. the critical ratio is typically taken to be 2 .this is the single neuron classifier .we use the single neuron classifier to make an estimate for the total proportion of bursting neurons in the network as follows @xmath146 where @xmath147 is the heaviside function .we can use @xmath148 to classify a network by picking some specific critical value of @xmath148 as a threshold for a classifier , however it is more informative to simply plot the contours of @xmath148 for a set of simulations run over a mesh in the bifurcation parameters of interest .in @xcite we considered multiple subpopulations for heterogeneous networks by simulating discrete homogeneous subpopulations within a network .for example , we simulated two subpopulations with two different sets of parameters corresponding to weakly adapting and strongly adapting neurons as a first attempt at studying inhomogeneous networks .however , a more realistic way of analyzing subpopulations in heterogeneous networks , is through the use of mixed distributions .a mixed random variable @xmath149 is a function of two or more random variables .for example , if @xmath150 and @xmath151 are random variables with probability density functions @xmath152 and @xmath153 , respectively , then @xmath154 is a mixed random variable with probability density function @xmath155 now consider a network where the adaptation jump size @xmath33 comes from two different subpopulations . in one subpopulation, @xmath156 is large , and in the other , @xmath156 is small .we can simulate these two subpopulations within an individual ( all - to - all ) coupled network using heterogeneity , and the density function @xmath157 where @xmath158 denotes the proportion of strongly adapting neurons and @xmath159 denotes the strongly adapting subpopulation , and @xmath160 denotes the weakly adapting subpopulation .note that these densities have higher moments than simply @xmath156 , and it is the magnitude of these moments that determine whether or not the density in the heterogeneous parameter we are considering is bimodal , indicative of multiple subpopulations . using this approach we can analyze how the bimodality affects the steady state distribution properties .we can also use more of the data gained from real networks for the purposes of simulation , as opposed to arbitrarily classifying neurons as strongly adapting or weakly adapting .the parameters of the individual density functions can be approximated , along with @xmath161 , using standard statistical approaches ( maximum likelihood estimation , for example ) .more importantly however , the same mean - field equations apply to a unimodal or a multimodal distribution of heterogeneity .the procedure in section [ mf3sec ] requires not only the estimation of @xmath162 , but also the calculation of its inverse and the derivative of this inverse . to calculate @xmath162 the mean field equations - are discretized in @xmath163 and the resulting equations are numerically integrated to steady state , which yields the steady state value of @xmath164 at each mesh point , @xmath165 .the inverse @xmath166 is calculated via numerically inverting the steady state @xmath167 as a function of @xmath165 as follows .the matlab function interp1 is used to interpolate values of @xmath163 given values of @xmath168 using the steady state @xmath169 mesh points .the derivative of the inverse is calculated using a finite - difference approximation over the mesh : @xmath170 this is then used to find @xmath171 at each mesh point via equation . to implement the computations in section [ invsec ] ,equations - are discretized in @xmath163 and numerically integrating to compute the steady state value of @xmath162 at each mesh point , @xmath172 .this is then used to compute @xmath173 through a first order finite - difference approximation over the discrete mesh : @xmath174 these two quantities are then used as approximations of @xmath175 and @xmath176 in equation to find @xmath177 at each mesh point . to estimate the parameter values for individual neurons , we take the discretized steady state firing rate , @xmath178 , calculated as indicated above .we then invert the functional relationship between @xmath178 and @xmath165 , and interpolate the @xmath163 values of the individual neurons using their firing rate , i.e. , we treat @xmath179 and @xmath165 as the @xmath180 points to be interpolated .this yields an estimate of the parameter values for each individual neuron , unlike in the approach for section [ mf3sec ] which yielded an estimate of the overall distribution .note that all that is required for this computation is knowledge of the steady state firing rate for each neuron .we use the griddata and the interp1 functions in matlab to perform the interpolation .p. hemond , d. epstein , a. boley , m. migliore , g.a .ascoli , and d.b .distinct classes of pyramidal cells exhibit mutually exclusive firing patterns in hippocampal area ca3b ., 18(4):411424 , 2008 .skinner , h. bazzazi and s.a .campbell two - cell to n - cell heterogeneous , inhibitory networks : precise linking of multistable and coherent properties0 18(3 ) : 343 - 352 .strogatz and r. e. mirollo .stability of incoherence in a population of coupled oscillators ., 63:613635 , 1991 .

<S> we analytically derive mean - field models for all - to - all coupled networks of heterogeneous , adapting , two - dimensional integrate and fire neurons . </S><S> the class of models we consider includes the izhikevich , adaptive exponential , and quartic integrate and fire models . </S><S> the heterogeneity in the parameters leads to different moment closure assumptions that can be made in the derivation of the mean - field model from the population density equation for the large network . </S><S> three different moment closure assumptions lead to three different mean - field systems . </S><S> these systems can be used for distinct purposes such as bifurcation analysis of the large networks , prediction of steady state firing rate distributions , parameter estimation for actual neurons , and faster exploration of the parameter space . </S><S> we use the mean - field systems to analyze adaptation induced bursting under realistic sources of heterogeneity in multiple parameters . </S><S> our analysis demonstrates that the presence of heterogeneity causes the hopf bifurcation associated with the emergence of bursting to change from sub - critical to super - critical . </S><S> this is confirmed with numerical simulations of the full network for biologically reasonable parameter values . </S><S> this change decreases the plausibility of adaptation being the cause of bursting in hippocampal area ca3 , an area with a sizable population of heavily coupled , strongly adapting neurons . </S>

1305.5640

the encrypted key exchange ( eke ) protocol integrates the asymmetric and symmetric encryption algorithms to ensure two parties sharing a secret key@xmath0}$ ] .we first evaluate the post - quantum security of the eke protocols via analyzing a simplified eke protocol .suppose two participants @xmath1 and @xmath2 preshare a password @xmath3 , they intend to establish a secret key @xmath4 via eke protocol .a simplified eke protocol is as follows@xmath5}$ ] : let @xmath6 be the encryption / decryption algorithms using password @xmath3 as its key , the public domain parameters are group @xmath7 , and @xmath8 with order @xmath9 .@xmath10 randomly chooses @xmath11 , and computes @xmath12 then she sends @xmath13 and @xmath14 to @xmath15 .@xmath15 randomly chooses @xmath16 , and computes @xmath17 then she sends @xmath18 and @xmath19 to @xmath10 .@xmath10 computes @xmath20 and @xmath15 computes @xmath21 finally , @xmath10 and @xmath15 establish a key string @xmath4 with eke protocol . for an eke protocol , the attacker(eve )can get all the ciphertexts transmitted in the channel . as the password @xmath3 is not available to eve , she can not impersonate @xmath15 to interact with @xmath10 and get the session key @xmath4 .in order to get @xmath4 , eve has to perform a password - guessing attack as follows : she generates a candidate password @xmath22 to decrypt @xmath23 , @xmath24 , and obtains @xmath25 or @xmath26 during this stage , eve can not check whether her candidate session key @xmath27 is correct .she can only assess a @xmath27 is correct by checking a decrypted plaintext is meaningful .suppose eve has a quantum computer . for each candidate @xmath28, eve has to execute the discrete logarithm computation once .let @xmath29 be a lower bound of time cost by a cycle of the discrete logarithm computation , and the time to perform an elementary quantum logic operation is @xmath30 .it is well known that @xmath30 has a lower bound : @xmath31 then the time of a discrete logarithm computing cycle @xmath29 will be @xmath32 where @xmath33 is the number of quantum gate executed serially in a discrete logarithm algorithm .it can be seen that the value of @xmath33 depends on the improvement of paralization of quantum discrete logarithm algorithm .although we can not calculate the exact value of @xmath33 , a rough estimation for a lower bound of @xmath33 is easy , then we have @xmath34 that is , the cycle of a discrete logarithm computation is no less than @xmath35 seconds .the physical security of a protocol@xmath36}$ ] indicates that the protocol will never be broken by an adversary living in the real physical world .let @xmath37 be the ultimate number of times a quantum computer can finish the discrete logarithm algorithm during 100 years , we have @xmath38 for each candidate @xmath28 , eve has to execute the quantum algorithm of discrete logarithm at least once to get @xmath39 or @xmath40 , so the length of the password @xmath3 should satisfy @xmath41 , that is , for resisting attack with several(@xmath42 ) quantum computers , a 68-bit password is enough to ensure the security of an eke protocol within 100 years .above analysis is based on some universal parameters of single - qubit quantum gate operations .now we focus on the analysis of cnot gate of ion - trap quantum computers .ion - trap quantum computer is proposed by cirac and zoller in 1995@xmath43}$ ] , it is considered one of the most promising scheme of physical implementation of quantum computer . to compute a discrete logarithm meaningful in cryptanalysis , the quantum computer must at least have @xmath44 qubits . in an ion - trap quantum computer ,the distance between two neighboring qubits is @xmath45 10 @xmath46 m , and the characteristic distance of two qubits executing the cnot operation is about @xmath47 . in an ion - trap ,the mutual coulomb force between two ions is @xmath48 where @xmath49 denotes the number of the electrons in an ion , and @xmath50 denotes the distance between the two ions . in the ion - trap quantum computer ,the force to an ion can be expressed as @xmath51 @xmath52 where @xmath53 denotes the riemann zeta - function , @xmath54 , so @xmath55 it is proper to reckon the propagation velocity of interaction between different qubits carried by photons via evaluating the velocity of the acoustic waves in the trap based partially on some parameters of the stationary wave .for the ion - trap , the dispersion relation of the acoustic waves is @xmath56 where @xmath57 are the frequency and the wavelength of the acoustic wave , respectively , and @xmath58 is the mass of every ion . as @xmath59 ,equivalent to monochromatic approximation , the dispersion relation can be simplified as @xmath60 then we obtain the velocity of the acoustic waves : @xmath61 let @xmath62 , the velocity will be @xmath63 so the time a cnot operation takes will be greater than @xmath64 in a discrete logarithm computation considered , the number of cnot operations executed serially is @xmath65}$ ] . in view of the fault - tolerant structure of the ion - trap quantum computer , especially the concatenated quantum error - correcting coding related with threshold theorem , we know that @xmath66 then we can infer that the lower bound of time finishing a discrete logarithm computation is @xmath67 that is , during one second , an ion - trap quantum computer can not finish the discrete logarithm computation @xmath68 times , then the attacker can not execute the discrete logarithm computation @xmath69 times in 100 years(@xmath70 seconds ) .therefore , in 100 years , an eke protocol with a 40-bit password can resist the attacker with several(@xmath42 ) ion - trap quantum computers .although physical parameters of other kinds of quantum computers are different from that of the ion - trap quantum computer , the conclusion is similar .let us estimate the length of a password that ensures the physical security of the eke protocol .suppose the size of a quantum computer is about one meter , then the upper bound of number of the quantum computers employing by any adversary is @xmath71 with all these quantum computers the attacker can run the discrete logarithm algorithm about @xmath72 times within 100 years .when the length of the password @xmath3 is @xmath73 , the attacker has to call the algorithm for averagely @xmath74 times , this is beyond the attacker s maximal calculating ability .88 is roughly the number of bits of 11 characters of asc@xmath75 , which is an usual option of a password in nowadays internet .our conclusion is : the attacker can not break eke protocol during 100 years even if he owns a huge number of ion - trap quantum computers , an 88-bit password is enough to achieve the physical security of a eke protocol .the eke schemes use both symmetric and public key encryption algorithm to provide secrecy and entity authentication for the users of network@xmath76}$ ] . by executing an eke protocol ,two participants @xmath1 and @xmath2 can identify each other and generate a common key @xmath4 .the eke protocol is as follows : \1 . @xmath10randomly generates the public key / secret key pair , @xmath77 and @xmath78 .then he uses the password @xmath3 to encrypt @xmath77 by symmetric encryption and sends to @xmath15 @xmath79 \2 .as @xmath15 is aware of the password @xmath3 , he decrypts @xmath80 to extract @xmath77 .afterwards , @xmath15 randomly generates a session key @xmath4 , and encrypts it with the public key @xmath77 and the password @xmath3 respectively by asymmetric and symmetric encryption , then sends to @xmath10 @xmath81 \3 . after receiving @xmath82 ,@xmath10 uses the password @xmath3 and the secret key @xmath78 to extract @xmath4 , then generates a random string @xmath83 and uses @xmath4 to encrypt it , then she sends to @xmath15 @xmath84 \4 .@xmath15 extracts @xmath83 with @xmath4 , then generates another random string @xmath85 , and sends to @xmath10 @xmath86 \5 .@xmath10 decrypts @xmath87 and gets @xmath88 , then she checks whether @xmath83 is correct .if the check succeeds , @xmath10 sends to @xmath15 @xmath89 \6 .@xmath15 decrypts @xmath90 to get @xmath85 , then he checks whether @xmath85 is correct .if the check pass , the protocol is accomplished , and @xmath10 and @xmath15 can use @xmath4 as the session key for communication .an eke protocol can be achieved with various public key algorithm .the one based on diffie - hellman protocol can be described as follows(the public domain parameters are group @xmath7 and @xmath91 with order @xmath92 which are public to all the users ) : \1 .@xmath10 randomly chooses a random string @xmath93 and sends to @xmath15 @xmath94 \2 . @xmath15randomly chooses a random string @xmath95 and computes : @xmath96 .he randomly chooses a string @xmath85 , then computes and sends to @xmath10 @xmath97 \3 .@xmath10 decrypts @xmath98 to get @xmath99 , further she computes @xmath4 and uses it to extract @xmath85 by decrypting @xmath90 with @xmath4 .she generates another random string @xmath83 , and sends to @xmath15 @xmath86 \4 .@xmath15 decrypts @xmath87 to get @xmath83 and @xmath85 .once he ensures that the string @xmath85 is correct , he sends to @xmath10 @xmath84 \5 .@xmath10 decrypts @xmath100 to get @xmath83 .once she ensures that the string @xmath83 is correct , the protocol is completed .now the two parties can use @xmath4 as the session key for communication . in this diffie - hellman - based eke protocol, the attacker can get all the ciphertexts transmitted in the channel : @xmath101 , @xmath98 , @xmath90 , @xmath100 , and @xmath87 . after generating a candidate password @xmath28 ,eve can computes @xmath102},\ \r_b'=\log_g\!{[d_{p'}(e_p(g^{r_b } ( \text{mod}\ q)))]},\ ] ] and further computes a candidate session key @xmath103 in order to avoid the attacker s achieving @xmath101 directly , @xmath10 can send the encrypted string @xmath104 to @xmath15 .then , the ciphertexts eve obtains turn to be @xmath104 , @xmath98 , @xmath90 , @xmath100 and @xmath87 . after generating a candidate password @xmath28 ,eve can computes a candidate key @xmath27 as @xmath105},\ k^{\prime}=(d_{p'}(e_p(g^{r_a } ( \text{mod}\ q))))^{r_b'}\ ] ] or @xmath106},\ k^{\prime}=(d_{p'}(e_p(g^{r_b } ( \text{mod}\ q))))^{r_a'}.\ ] ] after getting @xmath27 , eve can use it to decrypt @xmath90 , @xmath100 , @xmath87 and obtains @xmath107 @xmath108 then verifies @xmath27 by checking whether @xmath109 is equal to @xmath110 .we can see that , in this diffie - hellman - based eke protocol the entity authentication can help eve to check whether her candidate key is correct , at the same time , it will not reduce the number of of times executing the discrete logarithm computation . in this eke protocol , for each candidate password @xmath28 , the attacker has to execute the discrete logarithm computation once , no matter whether @xmath111 is encrypted with the password @xmath3 .therefore , a 40-bit password is enough to resist an attacker with several(@xmath42 ) quantum computers for 100 years , and an 88-bit password is enough to achieve a physical security of this protocol . in the diffie - hellman - based eke protocol ,the attacker eve an get all the ciphertexts transmitted in the channel : @xmath101 , @xmath98 , @xmath90 , @xmath100 and @xmath87 .once eve recovers some old session keys , she may use these messages to launch some attacks about the password @xmath3 .an enhanced eke protocol@xmath112}$ ] can efficiently forbid the attacker from using the old session key @xmath4 to get any information about the password @xmath3 . in the enhanced eke protocol, the string @xmath96 is used for key exchange .we denote the final session key as @xmath113 in step 2 of the diffie - hellman - based eke protocol , b generates another random string @xmath114 and sends to @xmath10 @xmath115 then , in step 3 , @xmath10 generates another random string @xmath116 and sends to @xmath15 @xmath117 finally , both @xmath10 and @xmath15 can get the session key @xmath118 .this key can be used to encrypt messages between @xmath10 and @xmath15 , and @xmath4 is just used for encrypting random strings .we can see that in this enhanced eke protocol , the attacker eve can not directly get any information about the password @xmath3 from the session key @xmath113 because @xmath3 is not directly used to encrypt any string that can be used directly to derive @xmath113 . in order to identify the password @xmath3 ,eve should get the correct encryption key @xmath4 at first .suppose eve has already got an old session key @xmath113 , he generate a candidate key @xmath27 to computes @xmath119 @xmath120 @xmath121 then checks whether @xmath27 is correct by checking whether @xmath122 is equal to @xmath113 .after getting the correct @xmath4 , she generates a candidate password @xmath28 , then computes @xmath123},\ k'=g^{r_a\cdot r_b'},\ ] ] now eve can check whether @xmath28 is correct by checking whether @xmath27 is equal to @xmath4 .compared with the simple diffie - hellman - based eke protocol , in this enhanced eke protocol , the security of the password @xmath3 will not only depends on the length of the password @xmath3 , but also depends on the length of the key @xmath4 , and @xmath4 can be much longer than @xmath3 . while @xmath4 is large enough , eve can not use the old session key @xmath113 to extract information about the password @xmath3 .in addition , in this protocol the attacker also has to execute the discrete logarithm computation at least once for each candidate @xmath28 , so the calculated amount of the attacker is not reduced , the physical security of the protocol is still ensured , a 40-bits password is long enough to resist the attacker with several(@xmath42 ) ion - trap quantum computer with 100 years , and an 88-bits password is long enough to achieve the physical security of the protocol .the original eke protocol has a defect : both the two parties @xmath10 and @xmath15 should be aware of the password @xmath3 in advance .actually , most of identification systems based on password just store the hash value of the password , rather than the password itself . in an augmented eke(a - eke ) protocol@xmath124}$ ] , the one - way hash value of the password @xmath3 is used as the super password .a diffie - hellman - based a - eke protocol is as follows : \1 .@xmath10 computes the one - way hash value of the password @xmath3(denoted by @xmath125 ) and randomly chooses a random string @xmath83 , then she sends to @xmath15 @xmath126 \2 .@xmath15 chooses a random string @xmath85 , and sends to @xmath10 @xmath127 \3 . both @xmath10 and @xmath15compute the session key @xmath128 in order to confirm that she really owns the password @xmath3 , @xmath10 sends to @xmath15 @xmath129 as @xmath15 is aware of the strings @xmath4 and @xmath125 , he can easily decrypt @xmath130 and verify the signature . only the one who has the password @xmath3 and the session key @xmath4can generate the cipher @xmath130 . as analyzed in the simple eke protocol in sec.1, the attacker also has to run the discrete logarithm algorithm once for each candidate @xmath131 in the a - eke protocol , so the physical security of this protocol is ensured by an 88-bit password .it can be seen that the attacker can not get any information about the password @xmath3 .in addition to the eke protocol , other hybrid cryptosystems are of interest for post - quantum cryptography , such as shamir s no - key protocol .shamir s no - key protocol can be regarded as a kind of key generation protocol@xmath132}$ ] . in this protocol ,two participants @xmath10 and @xmath15 exchange three messages through a public channel without authentication to transmit a secret key .the basic no - key protocol is as follows(a prime number @xmath92 is a public parameter that chosen in advance ) : \1 . @xmath10 randomly generates the secret key @xmath4 and a random number @xmath133 , and computes @xmath143 , then uses the password @xmath3 to encrypt it by a given symmetric encryption algorithm and sends to @xmath15 @xmath144 for this encrypted no - key protocol , eve can get all the ciphertexts transmitted in the channel : @xmath153 , @xmath148 , @xmath151 . as password @xmath3 is not available , eve can not get @xmath4 via impersonating @xmath15 to interact with @xmath10 . in order to obtain @xmath4, she has to perform a password - guessing attack as follows : eve randomly generates a candidate password @xmath22 , then uses it to decrypt @xmath153 , @xmath148 and @xmath151 , and obtains @xmath154 , @xmath155 and @xmath156 .then , she try to extract @xmath157 from @xmath158 $ ] , and @xmath159 from @xmath160 $ ] . finally , she computes @xmath161 @xmath162 and verifies whether the candidate password @xmath28 is correct by checking whether @xmath163 is equal to @xmath164 .it is obvious that the computational complexity of this attack depends on the length of password @xmath3 . for each candidate @xmath28, eve has to execute the discrete logarithm computation twice . as analyzed in sec.1 , in order to resist an attacker with an ion - trap quantum computer , the length @xmath165 of the password @xmath3 should satisfy : @xmath166 thus , for an attacker with several(@xmath167 ) ion - trap quantum computers , a 40-bit password is enough to ensure the physical security of the protocol within 100 years .account of the attacker s maximal ability of computing in physics , an 88-bit password is enough .the hybrid cryptosystems integrating both asymmetric and symmetric algorithms have already been a common choice in the practice of cryptography .the problem is to evaluate their security in quantum computing environment , and to choose appropriate parameters in accordance with their security requirements . for the developing of post - quantum cryptography ,this is in fact one of the most direct and effective option .00 s. m. bellovin and m. merritt , encrypted key exchange : password - based protocols secure against dictionary attacks , proceedings of the 1992 ieee computer society conference on research in security and privacy , 72 - 84 , 1992 d. r. stinson , cryptography theory and practice , crc press , chapman and hall , 2005 m. a. nielsen and i. l. chuang , quantum computation and quantum information , cambridge university press , cambridge , england , 2000 l. yang , c. xiang and b. li , qubit - string - based bit commitment protocols with physical security , arxiv : quant - ph/1011.5099v3 , 2010 j. i. cirac and p. zoller , quantum computations with cold trapped ions , phys .lett , * 74*(20 ) : 4091 - 4094 , 1995 p.w. shor , algorithms for quantum computation : discrete logarithms and factoring , proceedings of the 35th annunal symposium on foundations of computer science , ieee computer society , washington dc , 124 - 134 , 1994 p. w. shor , polynomial - time algorithms for prime facrotization and discrete logarithms on a quantum computer , siam journal on computing , * 26*(5 ) : 1484 - 1509 , 1997 p.w. shor , fault - tolerant quantum computation , proceedings of the 37th symposium on foundations of computer science , los alamitos , 15 - 65 , 1996 r. cleve and j. watrous , fast paralell circuits for the quantum fourier transform , proceedings 41st annual symposium on foundations of computer science , 526 - 536 , 2000 ( also see arxiv : quant - ph/0006004 ) b. schneier , applied cryptography : protocols , algorithm , and source code in c , john wiley & sons , inc , new york , 518 , 1996 d. p. jablon , extended password key exchange protocols immune to dictionary attack , proceedings of wet - ice97 , 248 - 255 , 1997 m. steiner , g. tsudik and m. waidner , refinement and extension of encrypted key exchange , acm operating systems review , july 1995 s. m. bellovin and m. merritt , augmented encrypted key exchange : a password - based protocol secure against dictionary attacks and password file compromise , proceedings of the symposium on security and privacy , 72 - 84 , 1992 a. j. menezes , p. c. van oorschot and s. a. vanstone , handbook of applied cryptography , crc press , boca raton fl , 1997 l. yang , b. y. yang and y. f. chen , full quantum treatment of rabi oscillation driven by a pulse train and its application in ion - trap quantum computation , ieee j. quantum electronics , * 49*(8 ) , 641 - 651 , 2013 ( also see arxiv : quant - ph/1010.5986 )

<S> we investigate the post - quantum security of the encrypted key exchange(eke ) protocols based on some basic physical parameters of ion - trap quantum computer , and show that the eke protocol with a 40-bit password will be secure against a quantum adversary with several ion - trap quantum computers . </S><S> we present a password encrypted no - key protocol to resist middle - man attack , and prove that it is also with the post - quantum security . </S><S> the analysis presented here is probably of general meaning for the security evaluation of various hybrid cryptosystems . </S><S> gbksong encrypted key exchange , post - quantum security , physical security of protocol the development of quantum computer is threatening modern cryptography , especially the public - key cryptosystems . </S><S> the post - quantum security of cryptographic protocols becomes of general interest . </S><S> since there is no efficient quantum attack so far to the cryptographic algorithms based on hard problems of the lattice , coding and multivariate objects , many researchers turning their interest to related cryptography and cryptanalysis problem . </S><S> here we consider to resist this threat along another line : since a quantum computer is a physical system , its gate operation rate is limited by some basic physics parameters , thus some kinds of cryptographic algorithms may keep their security against a quantum adversary . </S><S> we shall explain this through analyzing several examples in details . </S>

hep-ph0110012

the experimental possibility of studying the phase transitions of qcd via ultrarelativistic heavy ion collisions with the current effort at rhic and the forthcoming program at lhc motivates a theoretical effort to understand the dynamical aspects of phase transitions at high temperature .qcd is conjectured to feature two phase transitions , the confinement - deconfinement ( or hadronization ) and the chiral phase transitions .detailed lattice studies@xcite seem to predict that both transitions occur at about the same temperature @xmath14 .while lattice gauge theories furnish a non - perturbative tool to study the thermodynamic equilibrium aspects of the transition the _ dynamical _ aspects can not be accessed with this approach . in a condensed matter experimentthe temperature is typically a control parameter and it can be varied sufficiently slow so as to ensure that a phase transition occurs in local thermodynamic equilibrium . in an ultrarelativistic heavy ion collisionthe current theoretical understanding suggests that a thermalized quark - gluon plasma may be formed at a time scale of order @xmath15 with a temperature larger than the critical .this quark - gluon plasma then expands hydrodynamically and cools almost adiabatically , the temperature falling off as a power of time @xmath16 until the transition temperature is reached at a time scale @xmath17 depending on the initial temperature@xcite .whether the phase transition occurs in local thermodynamic equilibrium or not depends on the ratio of the cooling time scale @xmath18 to the relaxation or thermalization time scale of a fluctuation of a given wavelength @xmath19 , @xmath20 .if @xmath21 then the fluctuation relaxes in time scales much shorter than that of the temperature variation and reaches local thermodynamic equilibrium . if , on the other hand , @xmath22 the fluctuation does not have time to relax to local thermodynamic equilibrium and freezes out . for these fluctuationsthe phase transition occurs _ very fast _ and out of equilibrium .thus an important _ dynamical _ aspect is to understand the relaxation time scales for fluctuations .a large body of theoretical , experimental and numerical work in condensed matter physics reveal that while typically short wavelength ( @xmath23 ) fluctuations reach local thermal equilibrium , near a critical point long wavelength fluctuations relax very slowly , and undergo _ _ critical slowing down__@xcite . a phenomenological description of the dynamics near a phase transition typically hinges on the time - dependent landau - ginzburg equation which is generalized to include conservation laws@xcite . in the simplest case of a non - conserved order parameter , such as in a scalar field theory with discrete ( ising - like ) symmetry , the time dependent landau ginzburg equation is purely dissipative .while phenomenological , this approach has proved very succesful in a variety of experimental situations and is likely to provide a suitable description of the dynamics for macroscopic , coarse - grained systems such as binary mixtures , etc.@xcite . the phenomenological approach based on the timedependent landau - ginzburg equations which are first order in time derivatives seem to provide a suitable description of coarse - grained macroscopic dynamics in _ non - relativistic systems_. however it is clear that such approach is not justified in a relativistic quantum field theory , since the underlying equations of motion are second order and time reversal invariant .in particular in the case of purely dissipative time dependent landau ginsburg ( phenomenological ) description@xcite , frequency and momenta enter with different powers in the propagators , and at the mean field ( or tree ) level , this results in a dynamical scaling exponent @xmath24 .this situation must be contrasted to that of a relativistic quantum field theory where at tree level ( mean field ) frequencies and momenta enter with the same power in the propagator leading to a dynamical scaling exponent @xmath25 .furthermore , critical slowing down is automatically built in the phenomenological description , even at _ tree level _ as a consequence of dissipative equations of motion@xcite .clearly this is not the case in a relativistic quantum field theory . for a detailed discussion of the differences of non - equilibrium dynamical aspects between the time dependent landau ginzburg approach and quantum field theory see ref .@xcite .there are important non - equilibrium consequences of slow dynamics near critical points .if the cooling time scale is much shorter than the relaxation time scale of long - wavelength fluctuations , these freeze out and undergo spinodal instabilities when the temperature falls below the critical@xcite during continuous ( no metastability ) phase transitions .these instabilities result in the formation of correlated domains that grow in time@xcite with a law that in general depends on the cooling rate@xcite . in ultrarelativistic heavy ion collisions during the expansion of the quark gluon plasma the critical point for the chiral phase transition may be reached .if long wavelength fluctuations freeze out shortly before the transition , the ensuing instabilities may lead to distinct event by event observables@xcite in the pion distribution as well as in the photon spectrum at low energies@xcite .thus an important aspect of the chiral phase transition is to establish the relaxation time scales of long - wavelength fluctuations , and whether critical slowing down and freeze out of long - wavelength fluctuations can ensue . in the strict chiral limit with massless up and down quarks, qcd has an @xmath26 symmetry which is spontaneously broken to @xmath27 at the chiral phase transition , the three pions being the goldstone bosons associated with the broken symmetry .it has been argued that the low energy theory that describes the chiral phase transition is in the same universality class as the heisenberg ferromagnet , i.e , the @xmath28 linear sigma model@xcite .this argument has been used@xcite to provide an assessment of the dynamical aspects of low energy qcd based on the phenomenological time - dependent landau ginzburg approach to dynamical critical phenomena in condensed matter@xcite . while the universality arguments are appealing , a more microscopic understanding of dynamical critical phenomena in _ quantum field theory _is needed and has begun to emerge only recently@xcite .in reference@xcite a wilsonian renormalization group extended to finite temperature was implemented in a scalar quartic field theory . in this approachonly one loop diagrams enter in the computation of the beta functions , and the imaginary part of the self energy , which arises first at two loops order for @xmath29 is accounted for by an imaginary part in the effective quartic coupling@xcite . thereit is found that the relaxation rate of zero momentum fluctuations @xmath30 reveals critical slowing down in the form @xmath31 .our study in this article is complementary to that program in that we focus on the dynamical aspects of long - wavelength quasiparticles with @xmath32 . as discussed in@xcite and in detail below this is closer to the _ classical _ regime . * the goals : * in this article we study the _ dynamical _ aspects of quasiparticles near the critical point in a scalar quartic field theory by implementing a renormalization group program at high temperature . while the renormalization group has been generalized to finite temperature in various formulations@xcite , mainly to study critical phenomena associated with finite temperature phase transitions in field theory , only _static _ aspects were studied with these approaches .instead we focus on _ dynamical _ aspects , in particular the dispersion relations , and relaxation rates of long - wavelength excitations at and near the critical point . already at the technical levelone can see the differences : to understand dynamical aspects , in particular relaxation , a consistent treatment of absorptive parts of self - energies is required .this aspect is notoriously difficult to implement in a wilsonian approach in euclidean field theory@xcite .reference@xcite proposes a method to circumvent this problem but a complete treatment that manifestly includes absorptive parts of self - energy contributions is still lacking in this approach .other approaches using the euclidean version of the renormalization group adapted to finite temperature field theory were restricted to static quantities@xcite and in fact , as it will be seen in detail below , miss important phenomena that will be at the heart of the results presented here .* brief summary of results : * long - wavelength phenomena at high temperature @xmath33 implies a dimensional reduction from the decoupling of matsubara modes with non - zero frequency@xcite .the coupling in the dimensionally reduced theory is @xmath34 , where @xmath35 is the quartic coupling . by dimensional reasons ,the perturbative expansion in four space - time dimensions is in terms of the dimensionless ratio @xmath36 with @xmath37 the typical momentum scale , which is strongly relevant in the infrared . as a result ,a perturbative approach to studying long - wavelength phenomena breaks down .this is manifest in the breakdown of the quasiparticle picture in naive perturbation theory ( see@xcite and below ) . in @xmath38 space - time dimensions , the effective coupling in the high temperature ,long wavelength limit is @xmath39 .we implement an @xmath0 expansion around _ five _ space - time dimensions and a renormalization group resummation program at high temperature with @xmath40 near the critical point , with @xmath41 the typical frequency and momentum scales .we analyze the high temperature behavior of the relevant graphs and find that it is dominated for @xmath42 by the zero matsubara mode while the sum of the nonzero modes gives subdominant contributions .the effective renormalized coupling is driven to an infrared stable fixed point @xmath43 , which for small @xmath0 allows a consistent perturbative expansion near the fixed point .an important feature that emerges clearly in this approach , and that has been missed in most other treatments of renormalization group at finite temperature , is the _ anisotropic _ scaling between spatial and time directions , which is manifest in a non - trivial renormalization of the speed of light .this is a consequence of the fact that in the euclidean formulation at finite temperature time is compatified to @xmath44 , thus space and time or momentum and frequency play different roles .this results in a novel _ dynamical _ critical exponent , @xmath11 which determines the anisotropic scaling .the renormalization group leads to scaling in the infrared region in terms of anomalous dimensions which can be computed systematically in the @xmath0 expansion .in particular , to lowest order in @xmath0 we find for a scalar theory with discrete symmetry @xmath45,which describes a new universality class for dynamical critical phenomena in quantum field theory .all dynamical aspects , such as the relaxation rates , and dispersion relations depend on this critical exponent , while the static aspects are completely described by the usual critical exponents .we provide a renormalization group analysis of the quasiparticle properties near the critical point , such as their dispersion relation and width , complemented with an explicit evaluation to lowest order in the @xmath0 expansion .the main results obtained in this article are the following : at @xmath46 we find that the dispersion relation and width of quasiparticles of momentum @xmath2 is @xmath3 and @xmath47 respectively with a vanishing group velocity of quasiparticles in the long wavelength limit highlighting the collective nature of the quasiparticle excitations .for @xmath48 but @xmath49 we find @xmath50^{\frac{1}{2}}$ ] and @xmath51 with @xmath9 the finite temperature correlation length @xmath10 . in the case of @xmath12 symmetrywe find to lowest order in the epsilon expansion that the dynamical exponent @xmath52 .critical slowing down emerges near the critical point and in the @xmath0 expansion @xmath53 confirming the quasiparticle picture .we discuss some relevant cases of threshold singularities in which the usual ( breit - wigner ) parametrization of the quasiparticle propagator is not available since the real part of the inverse green s function vanishes at the quasiparticle frequency with an anomalous power law . in sectionii we introduce the model and discuss the breakdown of naive perturbation theory . in section iiiwe introduce the @xmath0 expansion and analyze the static case . in sectioniv the renormalization aspects and the anisotropic scaling is analyzed in detail .section v presents the renormalization group in the effective , dimensionally reduced theory both at and near the critical point .this section contains the bulk of our results which are summarized in section vi .our conclusions and a discussion of potential implications are presented in section vii .the high temperature behavior of the relevant diagrams is computed in the appendices .the low energy sector of qcd with two massless ( up and down ) quarks is conjectured to be in the same universality class as the @xmath28 heisenberg ferromagnet@xcite described by the @xmath28 linear sigma model . furthermore , since we are interested in describing the dynamical aspects associated with critical slowing down and freeze out of long - wavelength fluctuations just before the chiral phase transition , we focus on @xmath54 . while our motivation for studying critical slowing down stems from the experimental program in ultrarelativistic heavy ion collisions , the questions are of a fundamental nature . to understand the dynamical aspects near the critical point , we focus on the simpler case of a single scalar field theory , and we will recover the case of @xmath12 symmetry at the end of the discussion .we thus focus on the theory described by the lagrangian density @xmath55 where the subscripts in the mass and coupling refer to bare quantities .the case of @xmath56 components in the unbroken phase @xmath57 differs from the single scalar field by combinatoric factors that change the critical exponents quantitatively .these factors will be at the end of the calculation to obtain an estimate of the critical exponents for the @xmath12 theory in section [ sec : summary ] .we are interested in obtaining the relaxation properties of long wavelength excitations near the critical temperature , which in this scalar theory @xmath58 with @xmath59 being the renormalized mass and coupling .thus the regime of interest for this work is @xmath60 with @xmath61 being the momentum and frequency of the long wavelength excitation .as it will become clear below it is convenient to work in the matsubara representation of finite temperature field theory , which is more amenable to the implementation of the ( euclidean ) renormalization group. in the matsubara formulation euclidean time @xmath62 is compactified in the interval @xmath63 whereas space is infinite , bosonic fields are periodic in euclidean time and can be expanded as@xcite @xmath64 thus we see that while the spatial momentum is a continuum variable , the matsubara frequencies are discrete as a consequence of the compactification of euclidean time .this feature of euclidean field theory at finite temperature will be seen to lead to _ anisotropic _ rescaling between space and time and therefore , as it will be clear below , new _dynamical _ critical exponents . anticipating anisotropic rescaling , we then introduce the bare speed of propagation @xmath65 of excitations in the medium by writing the euclidean lagrangian in the form @xmath66 where we have introduced the effective renormalized _ temperature dependent _ mass @xmath67 and the counterterms , in particular the mass counterterm @xmath68 is adjusted order by order in perturbation theory so that the inverse two point function obeys @xmath69 the critical point is defined with ( the inverse susceptibility ) @xmath70 .we will begin our study by focusing our attention on the critical theory for which @xmath71 .we will later consider the theory near the critical point but in a regime in which @xmath72 .thus , the general regime to be studied is @xmath73 . in order to highlight the nature of the infrared behavior when @xmath74 we focus first on the critical theory in the static limit , when the matsubara frequencies of all external legs in the n - point functions vanish . for the purpose of understanding the nature of the infrared physics in the static limit we will set @xmath75 in this and next section , and we will recover this variable , the speed of light , when we study the dynamics in section [ section : dynamics ] .we consider first the @xmath77 scattering amplitude , or four point function , to one - loop order in three spatial dimensions .the full expression is given by @xmath78 + { \calo}(\lambda^3_0 ) \ ; , \label{oneloopscattamp } \end{aligned}\ ] ] where @xmath79 and @xmath80 , @xmath81 \left [ ( { \vec q}+{\vec p})^2 + ( 2 \pi \ ; t)^2 \ , ( n+m)^2 \right ] } \ ; , \ ] ] @xmath82 . since the external momentum @xmath83 it is clear from the above expression that the dominant infrared behavior of @xmath84 is determined by the zero matsubara frequency in the sumas it will be explicitly shown below the contribution from the non - zero matsubara frequencies will introduce a renormalization of the bare coupling which in the limit @xmath85 is independent of the external momentum ( this will be seen explicitly in the next section ) . keeping only the zero internal matsubara frequency and carrying out the three dimensional integral explicity we find @xmath86 thus , defining the effective coupling constant at the symmetric point @xmath87 where @xmath88 in the static limit one finds that in the infrared limit @xmath89 @xmath90 + \mathcal{o}(\lambda^3_0 ) \label{lambdaeff}\ ] ] two important features transpire from this expression : i ) the factor @xmath91 can be explained by dimensional arguments : in the matsubara formulation for each loop there is a factor @xmath33 from the sum over internal matsubara frequencies .the infrared behavior for @xmath92 is obtained by considering only the zero internal matsubara frequency in the loop .this integral has only one scale and since @xmath35 is dimensionless in three spatial dimensions the one - loop contribution must be proportional to @xmath91 .a similar argument shows that for a diagram with m internal loops , and transferred momentum scale @xmath37 there will be a power @xmath93 from the matsubara sums , the infrared behavior is obtained by the contribution with _ all _ the internal matsubara frequencies equal to zero , which by dimensional power counting must be of the form @xmath94 . therefore a diagram with m - internal loops will contribute to the scattering amplitude with @xmath95 . in taking only the zero internal matsubara frequencywe are assuming that the internal loop momenta are cutoff at a scale @xmath96 .thus , at the critical point the most important infrared behavior is that of the _ dimensionally reduced _ three dimensional theory@xcite .the reason for this dimensional reduction is clear : at finite temperature @xmath33 the euclidean time is compactified to a cylinder of radius @xmath97 for transferred momenta @xmath37 the spatial resolution is on distances @xmath98 .therefore for @xmath99 thus the compactification radius is effectively zero insofar as the long distance ( infrared ) physics is concerned .we will study below the contribution from the non - zero matsubara frequencies .ii ) for a transferred momentum scale @xmath37 perturbation theory breaks down for @xmath100 since the contribution from higher orders is of the form @xmath95 .this suggests that a resummation scheme is needed to study the infrared limit .this situation is similar to that in critical phenomena where infrared divergences must be summed and the renormalization group provides a consistent and systematic resummation procedure . we can obtain a hint of how to implement the renormalization group in finite temperature field theory in the limit when @xmath33 is much larger than any other scale ( masses , momenta and frequency ) by realizing that , from the argument presented above , the perturbative expansion is actually in terms of the dimensionless coupling @xmath101 .therefore from ( [ lambdaeff ] ) we can write @xmath102 + \mathcal{o}(g^3_0)\ ] ] we can improve the scattering amplitude via the renormalization group by considering the rg-@xmath103 function @xmath104 the first term ( with the minus sign ) just displays the scaling dimension ( for fixed @xmath34 ) of the effective coupling , that this dimension is @xmath105 is a consequence of the dimensional reduction since @xmath34 is the effective dimensionful coupling of the three dimensional theory .we thus see that the renormalization group improved coupling runs to the infrared fixed point @xmath106 as the momentum scale @xmath107 . comparing with the renormalization group beta function of critical phenomena@xcitewe see that this is the wilson - fischer fixed point in three dimensions , again revealing the dimensional reduction of the low energy theory .the resummation of the effective coupling and the fixed point structure can also be understood in the large n limit@xcite .as described in@xcite the large @xmath56 limit can be obtained by replacing the interaction in the lagrangian density for @xcite @xmath108 with @xmath109 , and the form of the quartic coupling has been chosen for consistency with the notation of reference@xcite .the leading order in the large @xmath56 limit for the scattering amplitude is obtained by summing the geometric series of one - loop bubbles in the s - channel ( only this channel out of the three contribute to leading order in the large n limit ) , each one proportional to @xmath56 which is the number of fields in the loop . as a resultone finds that the effective scattering amplitude at a momentum transfer @xmath37 is given by@xcite @xmath110 thus , introducing the dimensionless effective coupling @xmath111 one finds that @xmath112 i.e , the effective coupling constant goes to the three dimensional fixed point@xcite .the two point function in the static limit is given by @xmath113 where @xmath114 stands for the two loop sunset diagram at zero external matsubara frequency and the counterterm @xmath115 will cancel the momentum independent , but temperature dependent parts of the self - energy . the two loop self energy for external momentum @xmath116 and matsubara frequency @xmath117 is given by @xmath118 [ k^2 + \omega^2_j ] \left [ ( { \vec p}+{\vec k}+{\vec q})^2 + ( \omega_l+\omega_j+\omega_m)^2 \right ] } \ ] ] with @xmath119 .the static limit is obtained by setting @xmath120 ( @xmath121 ) . in this limit the dominant contribution in the infrared for @xmath85 arises from the term @xmath122 in the sum .the terms @xmath123 for which we can take @xmath124 ( since @xmath125 ) will be cancelled by the counterterm .a straightforward calculation leads to @xmath126 + { \cal o}(\lambda^3)\ ] ] where @xmath37 is a renormalization scale .this expression clearly reveals the effective coupling @xmath127 which becomes very large in the limit @xmath128 .clearly we need to implement a resummation scheme that will effectively replace the bare dimensionless coupling constant by the effective coupling that goes to a fixed point in the long - wavelength limit , and also ensure that at this fixed point the effective coupling is _ small _ so that perturbation theory near this fixed point is reliable .this is precisely what the renormalization group combined with the @xmath0-expansion achieves in critical phenomena@xcite .the two - loop contribution to the self - energy for @xmath129 is obtained from a dispersive representation of the self - energy in terms of the spectral density @xmath130 the spectral density @xmath131 has been obtained in reference@xcite in the high temperature limit@xcite in @xmath76 . using the expression given in@xcite for the spectral density at two loops in the high temperature limit , and after some lengthy but straightforward algebra , we find @xmath132\ ] ] carrying out the dispersive integral ( [ dispersive ] ) and subtracting off the terms that are independent of @xmath133 that are absorbed by the counterterm we find @xmath134\ ] ] clearly , the static limit @xmath135 of the self - energy coincides with ( [ defg2if ] ) .the two point function is therefore given by @xmath136-i\rho(p,\omega)\ ] ] with @xmath137 given by ( [ specdens ] ) .there are several features of this expression that are noteworthy : * it is clear that for @xmath138 the two loop contribution is much larger than the tree level term @xmath139 , this already signals the breakdown of perturbation theory in the high temperature regime when @xmath138 in the dynamical case .* consider the _real part _ of the two point function as @xmath140 i.e , near the mass shell .@xmath141 + \frac{1}{12 } \left(\frac{\lambda t}{4\pi p } \right)^2 \ln\left(\frac{2p}{\mu } \right ) \right\}\ ] ] this expression reveals that @xmath142 is _ not _ the position of the mass shell of the ( quasi)-particle .the coefficient of @xmath143 hints at wave function renormalization but the fact that the two point function _ does not _ vanish at this point prevents such identification .furthermore , we see that the term that does not vanish at @xmath144 hints at a momentum dependent shift of the position of the pole , i.e , a correction to the dispersion relation .however for @xmath145 both contributions are non - perturbatively large and the analysis is untrustworthy . * now consider the _ width _ of the quasiparticle , + @xmath146 + so that @xmath147 for @xmath145 .this signals the breakdown of the quasiparticle picture .+ a similar analysis reveals the breakdown of perturbation theory away , but near the critical point with @xmath148 .the imaginary part of the two - loops self - energy at @xmath149 and in terms of the temperature dependent mass @xmath150 can be obtained straighforwardly in _ _ three spatial dimensions__@xcite in the limit @xmath151 .it is found to be@xcite @xmath152 consequently , at two - loops order the width of the zero momentum quasiparticle in three spatial dimensions is given by@xcite @xmath153 * this behavior is different from that of gauge theories at high temperature where low order fermion or gauge boson loops are infrared safe and determined by the hard thermal loop contributions@xcite .this is so because for fermions there is no zero matsubara frequency while in the case of gauge bosons , the vertices are momentum dependent .while the second term inside the bracket in expression ( [ gamma2pw ] ) is determined by landau damping is infrared finite and is similar to the leading contribution in the hard thermal loop program@xcite , the first term arises from the three particle cut .the dependence of this term on the renormalization scale @xmath37 arises from the subtraction of the mass term at the critical point and reveals the infrared behavior . furthermore , in the hard thermal loop program@xcite one finds thermal masses of order @xmath154 with @xmath155 the gauge coupling , and widths of order @xmath156 ( up to logarithms ) so that @xmath53 in the weak coupling limit , while in the scalar theory under consideration ( [ gammaaway ] ) suggests that @xmath157 in naive perturbation theory .* we note at this stage that the high temperature limit @xmath158 of the self - energy calculated from the spectral representation ( [ dispersive ] ) can be _ directly _ obtained by computing @xmath159 in the matsubara representation given by ( [ sigmatfull ] ) by setting the internal matsubara frequencies @xmath160 and analytically continuing @xmath161 .+ we highlight this observation since it will be the basis of further analysis in what follows : _ the high temperature limit of the self energy @xmath158 can be obtained by setting the internal matsubara frequencies to zero and analytically continuing in the external matsubara frequency _ ,i.e , @xmath162}}\right|_{\omega_m\rightarrow -i\omega+0^+}\ ] ] we provide one and two loops examples of this statement in appendices [ sec : appendix1 ] and [ sec : appendix2 ] and formal proof of this statement to one loop order in appendix [ sec : formalproof ] .+ the result for the width of the quasiparticle at two loops was anticipated in references@xcite .this width is purely _ classical _ since the product @xmath34 is independent of @xmath163@xcite .this result for the damping rate of long wavelength quasiparticles in the critical theory is in striking contrast with that for @xmath164 which has been studied in detail in@xcite .for @xmath164 the thermal mass is @xmath165@xcite while the two loop contribution to the imaginary part of the self energy for @xmath166 is still @xmath167 .thus , the damping rate of long wavelength excitations is @xmath168 in the weak coupling limit .therefore for @xmath164 long wavelength excitations are true weakly coupled quasiparticles with narrow widths .the analysis of the previous section points out that naive perturbation theory at finite temperature breaks down at the critical point for momenta @xmath169 .the reason for this breakdown , as revealed by the analysis of the previous section , is the following : in four space - time dimensions the quartic coupling is dimensionless , however each loop diagram in the perturbative expansion has a factor @xmath33 from the sum over the matsubara frequencies . after performing the renormalization of the mass including the finite temperature corrections and setting the theory at ( or near ) the critical point, the effective expansion parameter for long - wavelength correlation functions is @xmath34 which has dimensions of momentum .if a given diagram has a momentum transfer scale @xmath37 the effective _ dimensionless _ expansion parameter is therefore @xmath36 which becomes very large for @xmath100 , i.e , the effective coupling is strongly relevant in the infrared .the analysis based on the rg beta function ( [ betag],[fixedpoint ] ) suggests that the effective coupling @xmath170 is driven to the three dimensional ( wilson - fischer ) fixed point in the infrared , obviously a consequence of the dimensional reduction in the high temperature limit .this is confirmed by the large n resummation of the scattering amplitude ( [ scatamplargen],[fixedpointlargen ] ) . _if _ the value of the coupling at the fixed point is @xmath171 then a perturbative expansion _ near the fixed point _ would be reliable , however the value of the coupling at the fixed point is @xmath172 which of course is a consequence of the fact that for fixed @xmath34 the effective coupling scales with dimension of inverse momentum in the infrared .this situation is the same as in critical phenomena for theories that are superrenormalizable , in which the infrared divergences are severe .the remedy in critical phenomena is to study the perturbative series via the @xmath0-expansion wherein the value of the coupling at the fixed point is @xmath173@xcite and sum the perturbative series via the renormalization group .we now implement this program in the high temperature limit . in _ five _ space - time dimensions the quartic coupling @xmath35 has canonical dimension of inverse momentum ,therefore the product @xmath34 that occurs in the perturbative expansion in the dimensionally reduced low energy theory is _dimensionless_. then in a perturbative expansion at ( or very near ) the critical point we expect that infrared divergences will be manifest in the form of logarithms of the momentum scale in the loop .this implies that the effective coupling is _marginal_. considering the theory in @xmath174 spatial dimensions and one euclidean ( compactified ) time dimension the effective coupling of the dimensionally reduced theory , @xmath34 , has dimensions of @xmath175 with @xmath37 being a momentum scale .therefore the effective dimensionless coupling for diagrams with a transferred momentum scale @xmath37 is @xmath176 .thus , for fixed @xmath33 the scaling dimension of this effective coupling is @xmath177 , hence we expect a non - trivial fixed point at which the coupling @xmath178 .therefore for @xmath179 we can perform a systematic perturbative expansion near the fixed point .this is the spirit of the @xmath0 expansion in critical phenomena which , when combined with a resummation of the perturbative series via the renormalization group has provided a spectacular quantitative and qualitative understanding of critical phenomena@xcite . while dimensional regularization and the @xmath0 expansion have been used to study the dimensionally reduced high temperature theory insofar as thermodynamic quantities is concerned , i.e , static phenomena@xcite , we emphasize that our focus is to study _ dynamics _ at and near the critical point , which is fundamentally different from the studies of static phenomena in these references . as a prelude to the study of the dynamics, we now revisit the scattering amplitude at one loop level and the self - energy at two loops level in @xmath174 _ spatial _ dimensions at high temperature in the static limit .the one loop self - energy is momentum independent and is absorbed in the definition of the thermal mass@xcite which is set to zero at the critical point .the are two main purposes of this exercise : the first is to quantify the role of the higher matsubara modes and secondly to obtain a guide for the infrared running of the coupling constant . the one loop contribution in the static limit , i.e , when the external matsubara frequencies are zero is given by @xmath180 \left [ ( { \vec q}+{\vec p})^2 + ( 2 \pi \ ; t\;n)^2 \right ] } ~~;~~ d= 4-\epsilon \ ; , \ ] ] in the high temperature limit the nonzero matsubara terms give subdominant contributions .this property can be argued in different manners .for example , the @xmath181 terms in eq.([unlazoepsi ] ) can be interpreted as feynman diagrams in @xmath182 dimensions with mass @xmath183 .such contributions are negligible for @xmath184@xcite .we find for high temperatures ( see appendix [ sec : appendix1 ] ) , @xmath185 \ ; , \ ] ] where @xmath186 stands for the @xmath187 contribution to the sum ( [ unlazoepsi ] ) * plus * the dominant high temperature limit of the sum over the @xmath181 terms , which is obtained by setting @xmath124 .separating the @xmath188 mode and setting @xmath124 in the contribution from the sum over @xmath189 we find @xmath190\ ] ] with @xmath191 riemann s zeta function which has the following properties@xcite @xmath192 where @xmath193 stands for the euler - mascheroni constant .near four space dimensions @xmath194 we find @xmath195 + { \cal o}(\epsilon)\ ; .\ ] ] there is _ no pole _ in @xmath0 and the argument of the logarithm reveals that @xmath33 acts as an ultraviolet cutoff .the reason that there is no pole in epsilon as @xmath194 is that the poles in @xmath0 should be independent of temperature and should be those of the zero temperature theory .however , in dimensional regularization one loop integrals have no poles in odd space - time dimensions@xcite . on the other hand near three space dimensions @xmath196 we find @xmath197 + { 1 \over ( 4\pi)^2 \ ; ( \epsilon - 1 ) } \ ; .\ ] ] where the pole term at @xmath198 corresponds to the usual coupling constant renormalization .this divergent term is temperature - independent , as expected , the @xmath199 is reminiscent of an upper momentum cutoff for the high temperature limit .the first term is precisely what we obtained in equation ( [ hir ] ) from setting the internal matsubara frequency to zero , i.e , the result of the dimensionally reduced theory . after subtracting the pole near three space dimensions ,the first term gives the leading infrared contribution in the limit @xmath200 whereas the logarithm is subleading .this expression coincides with that given in references@xcite . in these referencesthe four dimensional high temperature static theory was studied and a systematic analysis of feynman diagrams in the dimensionally reduced theory ( three dimensions ) was performed .the @xmath201 in our expressions should be mapped onto the @xmath202 for comparison with the results in these references .for @xmath203 we can neglect the terms of the form @xmath204 in eqn .( [ epsihit ] ) in the limit @xmath205 . and for @xmath206 we find that the static scattering amplitude at a symmetric point ( [ sympoint ] ) in the limit in which the temperature is much larger than the external momentum scales is given by @xmath207\ ] ] the factor @xmath33 arising from the matsubara sum is such that @xmath34 has dimensions of @xmath208 so that in @xmath38 space time dimensions @xmath209 is _dimensionless_. thus , introducing the _ dimensionless _ renormalized coupling @xmath210 we find @xmath211 therefore this effective coupling in the infrared limit is driven to a non - trivial fixed point @xmath212 hence for @xmath179 the fixed point theory can be studied perturbatively .this of course is the basis of the @xmath0 expansion in critical phenomena@xcite and will be the important point upon which our analysis will hinge .as mentioned above the one loop contribution to the self energy is momentum independent and absorbed into the definition of the thermal mass .the two loop contribution in the static limit in @xmath213 _ spatial _ dimensions is given by @xmath214[p^2+\omega^2_m][(p+q+k)^2+(\omega_l+\omega_m)^2]}\ ] ] we now introduce two feynman parameters , separate the @xmath215 term from the matsubara sums and take @xmath85 in the sums with @xmath216 to find @xmath217^{-\epsilon}+ g^2 t^{2}\left[\frac{t^2}{\mu^2 } \right]^{-\epsilon } \mathcal{c}(d)\ ] ] with @xmath218 only depending on the dimensionality .the second term @xmath219 does not depend on the momentum and is therefore cancelled by the mass counterterm which defines the critical theory .therefore for @xmath203 but small we find @xmath220= k^2\left[1+\frac{g^2}{12\epsilon}-\frac{g^2}{12 } \ln\left(\frac{k^2}{\mu^2}\right ) \right]+ \mathcal{o}(g^3)\ ] ] we introduce the wave function renormalization in dimensional regularization by the usual relation @xmath221 and choose @xmath222 therefore the renormalized two point function in the static , high temperature limit is given by @xmath223+\mathcal{o}(g^3)\ ] ] the infrared behavior is obtained by resumming the perturbative series via the renormalization group@xcite which leads to the scaling form of the two point function in the infrared limit @xmath224 with @xmath225 this is the anomalous dimension to lowest order in @xmath0 @xcite . the analysis of the static case above has highlighted several important features of the infrared behavior near the critical point , which determines the strategy for studying the dynamical case : * a : ) * the infrared behavior in the limit when @xmath226 with @xmath2 the typical momentum of the feynman diagram is determined by the dimensionally reduced theory obtained by setting the _ internal _ matsubara frequencies to zero .* b : ) * naive perturbation theory breaks down in three space dimensions because the dimensionless coupling is @xmath36 with @xmath37 the external momentum scale in the feynman diagram , while a large @xmath56 or renormalization group resummation suggests a non - trivial infrared stable fixed point , the coupling at this fixed point is of @xmath227 . *c : ) * just as in critical phenomena the perturbative expansion can be systematically controlled in an @xmath0 expansion around _four spatial dimensions _ corresponding to a theory dimensionally reduced from @xmath38 space time dimensions .the effective dimensionless coupling of the dimensionally reduced theory ( four dimensional ) is @xmath34 which is marginal .this combination is independent of @xmath163-this can be seen by restoring powers of @xmath163 @xmath228 - so that the low energy , dimensionally reduced theory is _classical_. in @xmath174 spatial dimensions the effective dimensionless coupling @xmath229 is driven to the infrared stable wilson - fischer fixed point of @xmath173 by the renormalization group trajectories .thus , the strategy to follow becomes clear : we will now study the dynamics by including the contribution from the external matsubara frequency , focusing on the infrared behavior for @xmath158 near the critical point in a systematic @xmath0 expansion around _ four _ spatial dimensions .we note that the theory in @xmath38 dimension is formally non - renormalizable in the ultraviolet , however this is irrelevant for the infrared which is the region of interest here .the analysis provided above in the limit @xmath194 clearly shows that near five space - time dimensions there are no poles in dimensional regularization in one loop diagrams as expected@xcite .the potential poles are replaced by @xmath199 .the low - energy theory must be understood with a cutoff of @xmath230 and the dimensionally regularized integrals in 5 space time dimensions clearly display this cutoff in the argument of logarithms .the long wavelength @xmath231 an the @xmath194 limits do not commute : keeping the subleading terms in the high temperature limit and taking @xmath194 results in that poles in @xmath0 actually translate into logarithms of the cutoff @xmath33 , on the other hand , keeping @xmath232 and small the @xmath233 limit can be taken and the subleading high temperature corrections vanish .clearly it is the _ latter _ limit the one that has physical relevance , since eventually we are interested in studying the infrared behavior of the physical theory in three space dimensions . hence in what follows we consider the long - wavelength limit for @xmath203 but small and approach the physical dimensionality @xmath234 in a consistent @xmath0 expansion improved via the renormalization group .this is the strategy in classical critical phenomena as well where for @xmath203 and small the ultraviolet cutoff can be taken to infinity . at this stageit is important to highlight the difference between the main focus of this work and that in references@xcite .the work of references@xcite studies the _ static _ limit of the dimensionally reduced theory near _ three _ spatial dimensions arising from the high temperature limit of a four dimensional euclidean theory compactified in the time direction .in contrast we here focus on studying the _ dynamics _ in the limit when @xmath158 which as emphasized by the analysis above will be studied in an @xmath0 expansion in a dimensionally reduced theory near _ four _ space dimensions .the limit of physical interest @xmath196 must be studied by improving the perturbative expansion via the renormalization group@xcite and eventually by other non - perturbative resummation methods , such as pade approximants or borel resummation that will extend the regime of validity of the @xmath0 expansion@xcite .we now turn to the dynamics .our main goal is to study the feasibility of a quasiparticle description of low energy excitations at and near the critical point .of particular interest is the dispersion relation as well as the damping rates of these excitations .this information is contained in the two point function @xmath235 which is the inverse propagator , analytically continued to @xmath236 .the region of interest is @xmath158 and if the theory is ( slightly ) away from the critical point @xmath237 as well . in principle for a fixed @xmath238 or fixed external matsubara frequencies in the external legs of @xmath239 functions , one must perform the sum over the internal matsubara first and then take the analytic continuation .however , as was shown above in detail in the static case , the most infrared singular contribution arises from setting the internal matsubara frequencies to zero . that this is also the case in the dynamicscan be seen by considering a diagram with @xmath240 internal lines , rerouting the external matsubara frequency through one of the lines .all of the lines are equivalent since rerouting the external matsubara frequency corresponds to a shift in one of the sums .the other @xmath241 contain propagators in which the internal matsubara frequency acts as a mass of @xmath242 .these are the superheavy modes in the description of references@xcite .the contribution that is dominant in the infrared is from the region of loop momenta @xmath243 which is largest when the mass of the propagator is zero , i.e , the zero matsubara frequency .keeping non - zero matsubara frequencies in any of the @xmath241 legs will lead to subleading contributions in the limit @xmath158 .once the internal matsubara frequencies had been set to zero we can analytically continue the external matsubara frequency to a continuous euclidean variable @xmath244 to obtain the euclidean two point function .the dispersion relation and damping rate are obtained by further analytical continuation @xmath245 . as anticipated in section([section : model ] ) because euclidean time is compactified and plays a _different _ role than the spatial dimensions , we must consider the anisotropic lagrangian density ( [ lagrangianeuclidean ] ) which includes the velocity of light multiplying the derivatives with respect to euclidean time .if this velocity of light is simply a constant it can be reabsorbed into a trivial redefinition of the time variable .however , as it will become clear below , this velocity of light acquires a non - trivial renormalization as a consequence of the anisotropy between space and time directions at finite temperature and will run with the renormalization group .thus , the euclidean propagator is generalized to @xmath246 we begin by studying the scattering amplitude , now as a function of external momenta and frequencies .the one loop contribution is determined by the function @xmath247 given by eqn .( [ unlazo ] ) , which for @xmath248 is given by ( see appendix ) @xmath249 \left [ 1 + { \cal o}\left(\frac{p^2 , \frac{s^2}{v^2_0}}{t^2}\right)\right ] \ ; , \ ] ] where @xmath250 stands for the @xmath187 contribution to the sum ( [ unlazo ] ) * plus * the dominant high temperature limit of the sum over the @xmath181 terms which is obtained by setting @xmath251 @xmath252 } + \frac{\gamma\left(1 + \frac{\epsilon}2 \right ) \ ; \zeta(\epsilon ) } { 8 \ ; \pi^{2+\frac{\epsilon}2 } \ ; \epsilon } \ ; t^{1 - \epsilon } \;,\ ] ] this integral is computed in the appendix a with the result @xmath253 where @xmath254 stands for the hypergeometric function . for @xmath232 and @xmath255 we can neglect the second term , since it is proportional to @xmath256 .we note that the infrared dominant contribution can be written in the form @xmath257 the factor @xmath33 thus combines with the coupling @xmath35 to give the effective coupling of dimension @xmath175 in @xmath258 spatial dimensions . for @xmath203 but small and neglecting the second term in ( [ 4ptonu ] )can be expanded in @xmath0 leading to @xmath259\ ] ] where we introduced the dimensionless bare coupling @xmath260 we remark that one can not take @xmath194 in this expression since in this limit the pole is actually cancelled by the second term in ( [ 4ptonu ] ) above .as emphasized above , this expression must be understood for @xmath203 but small so that the contributions of the form @xmath261 for @xmath262. therefore , the expression above must be understood in the sense that i ) : @xmath263 with fixed @xmath203 and ii ) : @xmath179 and the resulting expressions have a laurent expansion for small @xmath0 . neglecting the contribution from the non - zero matsubara frequencies which will be absorbed by the mass counterterm in the definition of the critical temperature ( or @xmath264 away from the critical point ) and also neglecting terms that vanish in the limit @xmath265 the dominant contribution in the infrared to the two loops self - energy is @xmath266 } \nonumber \\ = & & \frac{\lambda^2_0 t^2}{6(4\pi)^d}~\frac{\gamma^2(\frac{d}{2}-1)\gamma(3-d)}{\gamma(d-2)}~\int^1_0 dx ( 1-x)^{d-3}x^{1-\frac{d}{2}}\left[xp^2+\frac{s^2}{v^2_0 } \right]^{d-3 } \end{aligned}\ ] ] while this expression can be written in terms of hypergeometric functions , it is more convenient to expand it in @xmath0 with the result that @xmath267+\mathcal{o}(g^2\epsilon^0,g^2\epsilon , g^3)\ ] ] with @xmath268 given by eqn .( [ dimenlessbare ] ) .the forms of the two and four point functions immediately suggests a renormalization scheme akin to the familiar one used in critical phenomena@xcite with _ one important difference _ : we see from eqn .( [ gammaeps ] ) that the velocity of light @xmath65 must also be renormalized .the wave function renormalization is introduced as usual via @xmath269 the renormalized mass as a function of temperature is defined as @xmath270 this definition , however , defines the inverse susceptibility or correlation length , rather than the pole mass , the critical theory is defined by @xmath271 .coupling constant and velocity of light renormalization are achieved by @xmath272 @xmath273 the renormalization conditions that determine the constants @xmath274 are @xmath275 consistently with the @xmath0 expansion we choose the renormalization constants @xmath274 in the minimal subtraction scheme to lowest order , since keeping higher powers of the coupling or @xmath0 results in higher order corrections in the @xmath0 expansion .to lowest order , one loop for the 4-point function and two loops for the two point function , we find from the results ( [ dim5pt ] ) and ( [ gammaeps ] ) @xmath276 thus , the renormalized two point function reads @xmath277+\frac{s^2}{v^2_r}\left[1- \frac{g^2}{3\epsilon}\ln\left ( \frac{s^2}{v^2_r\mu^2}\right)\right]+\mathcal{o}(g^3,g^2\epsilon)\ ] ]before we embark on the resummation program via the renormalization group , it is important to highlight two important features * the contributions that are dominant in the infrared in the limit @xmath265 correspond to the terms with internal matsubara frequencies equal to zero , the non - zero matsubara frequencies give subleading contributions for @xmath203 .this in turn results in that the dependence on temperature is solely through the effective coupling @xmath278 . this can be seen from the fact that each loop has a factor t from the matsubara sum over the internal loop frequencies as well as one power of the coupling constant @xmath35 , by dimensional reasons the dimensionless coupling is obtained by multiplying by @xmath279 . *the velocity of light @xmath280 always enters in the form @xmath281 since this is the form that enters in the propagators and the renormalization conditions above .the bare n - point functions are independent of the renormalization scale @xmath37 , and this independence leads to the renormalization group equations ( we now suppress the subscript r understanding that all quantities are renormalized ) @xmath282\gamma^{(n)}\left(p_1,\frac{s_1}{v } ; p_2,\frac{s_2}{v}\ldots , p_n;\frac{s_n}{v } ; g,\mu\right ) = 0\ ] ] with @xmath283 to lowest order we find @xmath284~v+ \mathcal{o}(g^3,g^2\epsilon ) \label{betav2lup}\\ & & \gamma=\frac{g^2}{6}+ \mathcal{o}(g^3,g^2\epsilon ) \label{gamma2lup } \end{aligned}\ ] ] while we can write down the general solution of the rg equation ( [ egr ] ) for an arbitrary @xmath56-point function , our focus is to understand the quasiparticle structure which is obtained from @xmath285 .since @xmath285 has dimension two , it follows that @xmath286 therefore @xmath287 this scaling property then leads to @xmath288\gamma^{(2)}(e^t p,\frac{e^t s}{v},g,\mu)=0\ ] ] which combined with the rg equation ( [ egr ] ) leads to the following equation that determines the scaling properties of the two point function @xmath289\gamma^{(2)}\left(e^t p,\frac{e^t s}{v},g,\mu\right ) = 0\ ] ] the solution of this equation is standard@xcite @xmath290 with @xmath291 as @xmath292 i.e , the momentum and frequency are scaled towards the infrared we see from the rg @xmath103 function ( [ betag1lup ] ) that coupling is driven to its fixed point @xmath293 which in turn implies that @xmath294 where we introduced the new _ dynamical _ critical exponent @xmath295 therefore , in the asymptotic infrared limit we find that @xmath296 it is convenient to re - define @xmath297 to find @xmath298 and finally writing @xmath299 and using the property ( [ dimlessgamma ] ) we find the scaling form in the infrared limit @xmath300^{2-\eta } \phi\left(\vartheta \right)\quad ; \quad \vartheta= \left(\frac{s}{v(\mu)\mu^{1-z}p^{z}}\right)^2\ ] ] the solution of the rg equation clearly shows that the two point function in the infrared limit is a _ scaling _ function of the ratio @xmath301 highlighting the role of the _ new dynamical exponents _ @xmath11 given by eqn .( [ zdyn ] ) with @xmath302 to lowest order given by eqn .( [ etatemporal ] ) .the emergence of the new dynamical exponent @xmath11 is a consequence of the anisotropic renormalization between momentum and frequency , or space and time manifest in the renormalization of the speed of light .this novel phenomenon can be traced back to the different role played by time ( compactified ) and space in the euclidean formulation at finite temperature . a similar anisotropic rescaling emerges in a different context : a heisenberg ferromagnet with correlated impurities@xcite with similar renormalization group results . while the formal solution does not yield the function @xmath303, we can find it by matching to the lowest order perturbative expansion ( [ 2ptren ] ) when the coupling is at the non - trivial fixed point . from the form of the of the perturbative renormalized two point function given by eqn .( [ 2ptren ] ) and assuming the exponentiation of the leading logarithms via the renormalization group near the non - trivial fixed point @xmath304+\frac{s^2}{v_r^2}\left[1- \eta_t \ln\left ( \frac{s}{v_r\mu}\right)\right ] \thickapprox p^{2-\eta } \mu^{\eta } + \left(\frac{s}{v_r}\right)^{2-\eta_t}\mu^{\eta_t}\ ] ] which can be immediately written in the scaling form @xmath305 \nonumber \\ z & = & \frac{2-\eta}{2-\eta_t } \simeq 1+\frac{1}{2}(\eta_t-\eta)\end{aligned}\ ] ] clearly this form coincides with the scaling solution of the renormalization group and the perturbative expansion in the regime in which it is valid .we note , however , that in the computation of @xmath302 we have neglected contributions to the renormalization of the velocity @xmath280 of @xmath306 which would appear at next order and would lead to an @xmath307 contribution to @xmath302 , which is of the same order as @xmath308 in @xmath11 .thus consistently we must neglect the contribution of @xmath308 to the dynamical exponent @xmath11 which to lowest order is therefore @xmath309 the quasiparticle structure of the theory is obtained from the green s function @xmath310 .in particular the dispersion relation and the width of the quasiparticle are obtained from the real and imaginary parts , respectively . while the general solution of the rg equation does not determine the scaling function @xmath303 in eqn .( [ scalefunc ] ) the fact that it is only a function of the scaling ratio @xmath311 , allows to extract the quasiparticle structure .the analytic continuation @xmath245 leads to the analytic continuation of the scaling variable @xmath312 with @xmath313 writing the scaling function @xmath303 analytically continued in terms of the real and imaginary parts @xmath314 , the position of the quasiparticle pole corresponds to the value of @xmath315 for which the real part vanishes .call this dimensionless real number @xmath316 , hence it is clear that the dispersion relation for the quasiparticles obeys @xmath317 furthermore , _ assuming _ that @xmath318 vanishes linearly at @xmath316 we can write the green s function near the position of the pole in the form @xmath319^{2-\eta } } ~ \frac{1 } { ( \varpi-\varpi^*)\phi^{\prime}_r(\varpi^*)+i \phi_i ( \varpi^ * ) } \ ] ] alternatively , we can write the rg improved propagator , near the quasiparticle pole in the breit - wigner form @xmath320 with the dispersion relation , residue at the quasiparticle pole and quasiparticle width given by @xmath321^{2-\eta } \phi^{\prime}_r(\varpi^*)}\label{zpole }\\ \gamma_p & = & \frac{\phi_i ( \varpi^ * ) \ ; v(\mu ) \;\mu^{1-z } \;p^{z}}{\phi^{\prime}_r(\varpi^ * ) } \equiv \frac{\omega_p \phi_i ( \varpi^*)}{\varpi^ * \phi^{\prime}_r(\varpi^ * ) } \label{widthrg}\end{aligned}\ ] ] the imaginary part @xmath322 must be proportional to the anomalous dimensions , hence perturbatively small in the @xmath0 expansion ( this will be seen explicitly below to lowest order ) .the definite values for @xmath316 , @xmath323 must be found by an explicit calculation. however , the above quasiparticle properties , such as the position of the pole , group velocity , residue and width are _ universal _ in the sense that they only depend on the fixed point theory . for a positive dynamical exponent @xmath11the above analysis reveals a vanishing group velocity and width for long - wavelength quasiparticles at the critical point .furthermore the expression for the width given by ( [ widthrg ] ) not only displays the phenomenon of critical slowing down , i.e , the width of the quasiparticle vanishes in the long - wavelength limit , but also the validity of the quasiparticle picture , since @xmath53 in the @xmath0 expansion .* threshold singularities : * while we have assumed above that the real part of the scaling function vanishes linearly at the quasiparticle pole , this need not be the general situation .it is possible that the real part vanishes with an anomalous power law , i.e , @xmath324 in this case a quasiparticle width can not be defined as the residue will either vanish or diverge depending on the sign of @xmath325 .it is also possible that @xmath326 also vanishes with an anomalous power at @xmath327 .we refer to these cases as threshold singularities and we will find below an example of this case .another example of this situation has been found in dense qcd as a result of the breakdown of the fermi liquid theory in the normal phase@xcite. clearly only a detailed calculation of the scaling functions can reveal whether it vanishes linearly or with an anomalous power law at @xmath327 .the set of quasiparticle properties given above ( [ quuasipole]-[widthrg ] ) are only valid provided the real part vanishes _linearly_. we can go further and find the explicit form of the scaling function by focusing on the renormalization group improved propagator obtained in lowest order in the @xmath0 expansion given by eqn .( [ scalingform ] ) .the analytic continuation to real frequencies of the rg improved two point function to lowest order given by ( [ scalingform ] ) leads to @xmath328\ ] ] where we have approximated @xmath329 to lowest order . from this expressionwe see that the dispersion relation @xmath330 , group velocity @xmath331 width @xmath332 quasiparticles residue @xmath333 are of the form given by equations ( [ quuasipole]-[widthrg ] ) with @xmath334 , and with the following explicit expressions to leading order in the @xmath0 expansion @xmath335 with @xmath11 given by ( [ dynexpz ] ) above . several important features of these expressions must be highlighted * the dispersion relation ( [ disprel ] ) features an _ anomalous dimension _ given by the dynamical exponent @xmath336 .the product @xmath337 is a renormalization group invariant as can be seen from equations ( [ betav ] ) and ( [ betav2lup ] ) evaluated at the fixed point .thus , all of the above quantities that describe the physical quasiparticle properties are manifestly renormalization group invariant . * the group velocity ( [ groupvelo ] ) _ vanishes _ in the long wavelength limit as a power law completely determined by the dynamical anomalous dimension @xmath11 .this feature highlights the collective aspects of the long - wavelength excitations . ** critical slowing down : * is explicitly manifest in the width @xmath332 since @xmath338 as @xmath339 .furthermore we also emphasize the validity of the quasiparticle picture , the ratio @xmath340 .thus , the quasiparticles are narrow in the sense that their width is much smaller than the position of the pole . even considering @xmath341 corresponding to dynamical critical phenomena in three spatial dimensions @xmath342thus , we see that the renormalization group resummation has led to a consistent quasiparticle picture , but in terms of a dispersion relation that features an anomalous dimension and a group velocity that vanishes in the long - wavelength limit .obviously these features of the quasiparticles can not be extracted from a naive perturbative expansion .having studied the quasiparticle aspects at the critical point , we now turn our attention to their study slightly away from the critical point .the critical region of interest is @xmath344 .critical behavior in the broken symmetry phase near the critical point with @xmath345 will be studied elsewhere with particular attention on the critical dynamics of goldstone bosons . in this articlewe restrict our attention to the normal phase near the critical point . in the lagrangian density ( [ lagrangianeuclidean ] )the term @xmath264 is the _ exact _ mass ( rather the exact inverse susceptibility ) determined by the condition ( [ massoft ] ) and the counterterm @xmath115 is adjusted consistently in perturbation theory to fulfil this condition .however , to relate the mass to the departure away from the critical temperature it is more convenient to re - arrange the perturbative expansion in a manner that displays explicitly the departure from @xmath346 .this is achieved as follows .consider the one loop contribution to the self - energy in the _ massless _ theory in @xmath213 -spatial dimensions : @xmath347 with @xmath348 only depends on the spatial dimensionality and @xmath349 stands for zero temperature terms .it is convenient to group this contribution with the bare mass term in the lagrangian in the form @xmath350^\frac{1}{d-1}\ ] ] where the zero temperature contributions ( denoted by @xmath349 in eqn .( [ masslup ] ) ) had been absorbed in @xmath351 .we re - organize the perturbative expansion by re - writing the euclidean lagrangian in the form @xmath352 where now the counterterm @xmath115 is simply the one loop tadpole diagram evaluated for zero mass given by ( [ masslup ] ) , it is of @xmath353 and is included in the perturbative expansion consistently . to one - loop order the two point functionis now given by @xmath354 the integral above is of the typical form as those studied in the previous sections ( see ( [ epsihit ] ) ) . separating the @xmath121 matsubara contribution from the @xmath355 for which we can set @xmath356 for @xmath357, we obtain @xmath358 } = m^2(t)\left [ \frac{\gamma(\frac{\epsilon}{2})\gamma(1- \frac{\epsilon}{2})}{(4\pi)^{\frac{d}{2}}\gamma(\frac{d}{2 } ) } m^{-\epsilon}(t ) + \mathcal{c}(d ) \ ;t^{-\epsilon}\right]\ ] ] again , for @xmath203 we can neglect the second term in the brackets in the limit @xmath359 . expanding in @xmath0 to obtain the lowest order contribution consistently in the @xmath0 expansion ,we obtain the two point function to one loop order @xmath360\ ] ] the renormalized mass parameter @xmath150 is defined by @xmath361 and @xmath362 is fixed by the renormalization condition @xmath363 since @xmath364 receives corrections at @xmath365 we choose @xmath362 to lowest order in the @xmath0 expansion to be @xmath366 before we embark on a full discussion of the dynamics away from the critical point , it proves convenient and illuminating to discuss the static aspects first . in particular since we will study the @xmath124 case but @xmath367 a relevant quantity is the inverse susceptibility @xmath368 , which is defined as @xmath369 which near the critical point and the non - trivial fixed point @xmath370 given by eqn .( [ fp ] ) is given by @xmath371\approx \left [ m^2_r(t ) \right]^{1+\frac{g^*}{2}}\ ] ] where we anticipated an exponentiation of the leading logarithms via the renormalization group , which will be borne out by the renormalization group analysis below .recalling that @xmath372 by eqn .( [ mast ] ) , we find latexmath:[\[\label{corrfin } \chi^{-1}(t\sim t_c ) \propto critical exponent @xmath30 is seen to be the correct one @xcite . just as in the case of the theory at the critical pointstudied above , we now study the dynamics of the theory in an @xmath0 expansion and implement a resummation of the leading infrared divergences via the renormalization group . as argued above the leading infrared behavioris obtained by setting the internal matsubara frequencies to zero in the two loops self energy . in @xmath258 spatial dimensions ,the self - energy at two loops is @xmath374[k^2+m^2(t)][(q+k+p)^2+\frac{s^2}{v^2_0}+m^2(t ) ] } \label{sig2ltg}\ ] ] the loop integrals are evaluated by introducing two feynman parameters leading to @xmath375^{\frac{\epsilon}{2}-1}y^{\frac{\epsilon}{2}-1 } \left[x(1-x)y(1-y)p^2+\right .\nonumber \\ & & \left .+ ( 1-y)x(1-x)m^2(t)+y m^2(t)+yx\frac{s^2}{v^2_0 } \right]^{1-\epsilon } \end{aligned}\ ] ] it is convenient to separate the static contribution from the dynamical part by writing @xmath376 the static contribution @xmath377 leads to wave function renormalization a renormalization of the mass and to @xmath307 corrections to the static anomalous dimensions , which will be neglected to leading order in the @xmath0 expansion .the second , dynamical contribution is obtained consistently in an @xmath0 expansion : the regions of the integrals in the feynman parameters that lead to inverse powers of @xmath0 in an @xmath0 expansion are @xmath378 and @xmath379 .the contributions of these regions can be isolated by partial integration , and after some straightforward algebra we find @xmath380 + \mathcal{o}(\epsilon^0,\epsilon)\ ] ] thus , putting together the one loop contribution found previously and the two loop contribution found above , the two point function at zero spatial momentum but away from the critical point is found to be @xmath381 + \frac{s^2}{v^2_0}+m^2(t)\left[1-\frac{g(\mu)}{\epsilon}+ \frac{g(\mu)}{2}\ln\left(\frac{m^2(t)}{\mu^2}\right ) \right]+ \nonumber \\ & & \frac{g^2(\mu)}{6\epsilon}\left[\frac{2}{\epsilon } \frac{s^2}{v^2_0}-2\frac{s^2}{v^2_0 } \ln\left(\frac{m^2(t)}{\mu^2}\right)- 2\left(m^2(t)+\frac{s^2}{v^2_0}\right)\ln\left(1 + \frac{s^2}{v^2_0 m^2(t ) } \right ) \right ] \nonumber \\ & & + \mathcal{o}(g^2,g^2\epsilon ) \end{aligned}\ ] ] where we have neglected logarithmic corrections that will exponentiate to anomalous dimensions of @xmath307 for momentum dependent terms .we have only displayed the contribution that will be cancelled by wave function renormalization just as in the critical case .obviously the @xmath382 limit coincides with the two point function at the critical point ( [ gammaeps ] ) to leading order @xmath173 . in the above expression we have not included the two loop contribution to the static @xmath383 part , since it will lead to an @xmath307 correction to the critical exponent for the correlation length ( inverse susceptibility ) .the renormalization conditions for the two point function away from the critical point are now summarized as follows : @xmath384 along with the renormalization conditions on the four point function ( [ couprenor ] ) . to leading order in the @xmath0 expansionthe renormalization constants @xmath385 are given by ( [ zphi ] ) , ( [ zvel ] ) and ( [ zmass ] ) respectively .thus , we find the renormalized two point function at two loop order and to leading order in the @xmath0 expansion ( since @xmath386 ) @xmath387 + \frac{s^2}{v^2_r}\left[1- \frac{g^2(\mu)}{3\epsilon } \ln\left(\frac{m^2_r(t)}{\mu^2}\right)\right ] \nonumber \\ & & -\frac{g^2(\mu)}{3\epsilon}\left(m^2_r(t)+ \frac{s^2}{v^2_0}\right)\ln\left(1 + \frac{s^2}{v^2_0 m^2_r(t ) } \right ) + \mathcal{o}(g^2,g^2\epsilon ) \end{aligned}\ ] ] since @xmath388 has dimension two it is convenient to introduce the dimensionless quantity @xmath389 and the corresponding renormalization group beta function @xmath390 where we have used ( [ renmas ] ) and ( [ zmass ] ) . the renormalization group equation for the @xmath56-point function away from the critical point is now given by @xmath391\gamma^{(n)}\left(p_1,\frac{s_1}{v } ; p_2,\frac{s_2}{v}\ldots , p_n;\frac{s_n}{v } ; g,\bar{m},\mu\right ) = 0\ ] ] the new ingredient as compared to the critical case ( [ egr ] ) is the dependence on @xmath392 . following the same steps as for the critical case, we now find that the solution of the renormalization group equation for the two point function obeys @xmath393 with @xmath394 the solution of the differential equation @xmath395 with the initial condition @xmath396 and @xmath397 just as in the solution of the renormalization group equation at criticality near the fixed point ( [ asirlim],[asirlimred ] ) , introducing @xmath398 we now find @xmath399 following the analysis of the critical case , and the scaling property ( [ dimlessgamma ] ) and writing @xmath299 we find the following scaling form @xmath400^{2-\eta } \phi\left(\frac{s}{v(\mu)\mu^{1-z } p^{z}};p\xi\right ) \quad;\quad \xi= \frac{1}{\mu}\left[\frac { m^2_r(t,\mu)}{\mu^2}\right]^{\frac{1}{\gamma^*_{\bar m}-2}}\ ] ] @xmath9 is therefore identified with the _ _ correlation length__@xcite @xmath401 it is important to note at this stage that the correlation length @xmath402 is a renormalization group invariant , as can be easily checked by using ( [ dimensionlessmass ] ) with the renormalization group beta function ( [ betamass ] ) .to study the limit of zero spatial momentum it is more convenient to rewrite the above scaling solution in the following form @xmath403^{-(2-\eta ) } \psi\left(\frac{s~\xi^{z}}{v(\mu)\mu^{1-z}};p\xi \right)\ ] ] from the definition of the inverse susceptibility @xmath404 we find the known result@xcite @xmath405 furthermore the two - point function is a function of two renormalization group invariant , dimensionless scaling variables @xmath406^{\frac{2-\eta}{2-\gamma^*_{\bar m}}}\psi\left(\varphi , \delta \right)\ ] ] with @xmath407 ^ 2\left[\frac{m^2_r(t)}{\mu^2}\right]^{\frac{2z}{\gamma^*_{\bar m}-2}}\equiv \left[\frac{s\,\xi^{z}}{v(\mu)\mu^{1-z } } \right]^2\label{varphi } \\ \delta & = & \frac{p^2}{\mu^2}\left[\frac { m^2_r(t,\mu)}{\mu^2}\right]^{\frac{2}{\gamma^*_{\bar m}-2 } } \label{delta } \equiv ( p\ , \xi)^2 \end{aligned}\ ] ] the renormalization condition ( [ newcond ] ) determines that @xmath408 .we can now follow the arguments provided in the previous subsection for the critical case . under the analytic continuation@xmath409 @xmath410 @xmath411 the position of the quasiparticle pole in the two point green s function corresponds to the value of @xmath412 for which @xmath413 .this condition determines the dispersion relation of the quasiparticle and is given by @xmath414 this expression emphasizes that the dispersion relation depends on @xmath2 through the scaling variable @xmath415 . _ assuming _ that near the quasiparticle pole @xmath416 vanishes _ linearly _ the green s function can be approximated by @xmath417 where @xmath418 . near the quasiparticle polewe can further write the above expression in the breit - wigner form @xmath419 with @xmath420 where we have suppressed the dependence on the scaling variable @xmath421 in the arguments of the real and imaginary parts to avoid cluttering of notation .furthermore , we have made explicit the combination of _ static and dynamic _critical exponents using the expression given in eqn .( [ corre ] ) for the static critical exponent @xmath422 and the dependence on the momentum is implicit through the dependence on the scaling variable @xmath421 of @xmath327 as well as the explicit dependence of the real and imaginary parts .again , the imaginary part must be proportional to the anomalous dimensions , hence perturbatively small in the @xmath0 expansion .therefore the expression for the width ( [ widthaway ] ) reveals both critical slowing down , since @xmath423 vanishing at @xmath46 , _ and _ the validity of the quasiparticle picture since @xmath53 in the @xmath0 expansion . at this pointwe recognize a fundamental difference with the wilsonian results of reference@xcite . while in ref .@xcite the width was found to be proportional to @xmath424 up to logarithms , we see from ( [ widthaway ] ) that the quasiparticle width actually involves the new dynamical anomalous exponent @xmath11 .the difference can be traced to the fact that the wilsonian approach advocated in ref.@xcite does _ not _ include two loop diagrams which are necessary to reveal the anisotropic renormalization through the renormalization of the speed of light and are directly responsible for the new dynamical anomalous exponent @xmath11 .we emphasize that the above breit - wigner form as well as the quasiparticle properties rely on the assumption that the real part of the scaling function vanishes linearly near the quasiparticle pole .as emphasized before in the critical case this need not be the general situation , and anomalous power laws can lead to threshold singularities as discussed above .while the solution of the renormalization group leads to a scaling form of the two point correlation function , it does _ not _ explicitly specify the scaling function @xmath425 .however , we can obtain the function @xmath425 by matching the leading logarithms to those of the perturbative expression ( [ rentwopointawaytc ] ) evaluated at the fixed point @xmath426 to lowest order in the @xmath0 expansion . matching the leading logarithms and assuming their exponentiation via the renormalization group it is straightforward to see that the two point function is given by @xmath427^{\frac{2-\eta}{2-\gamma^*_{\bar m}}}\left\{\delta+\left[1+\varphi \right]^{2-z}\right\}\ ] ] where we have used the lowest order results in the @xmath0 expansion @xmath428 and kept consistently the lowest @xmath173 in the exponentiation of the leading logarithms leading to ( [ resumed2pt ] ) .thus , we obtain the lowest order result for the scaling function @xmath429^{2-z}\ ] ] we can now obtain an explicit form of the real and imaginary parts of the scaling function that enter in the quasiparticle parameters .this is achieved by performing the analytic continuation ( [ contaway ] ) which leads to @xmath430^{2-z } \nonumber \\ & = & \delta-|\omega^2 - 1|^{2-z}\left[1+i~\pi~(z-1)~ \mathrm{sign}(\omega)\theta(\omega^2 - 1)\right ] \label{anaconti } \end{aligned}\ ] ] for @xmath124 i.e , @xmath431 we see that both the real and imaginary part of the scaling function vanish at @xmath432 _ with an anomalous power law _ providing an explicit example of the case of threshold singularities mentioned above . for @xmath433 and @xmath367we find a quasiparticle pole at @xmath434 where we have approximated the anomalous dimension by its leading order in @xmath0 using @xmath435 . from this expression for @xmath327we obtain the dispersion relation for quasiparticles @xmath436^{2z\nu}+\left[\frac{p^2}{\mu^2 } \right]^{z } \right\}\ ] ] with @xmath422 given by eqn .( [ corre ] ) , in particular we find that the frequency of zero momentum quasiparticles @xmath437 . obviously at @xmath438 ( @xmath439 )the dispersion relation coincides with that of the critical case given by eqn .( [ disprel ] ) . for @xmath433 , i.e ,@xmath440 the real part of the scaling function vanishes linearly and the breit - wigner approximation ( [ bwaway ] ) near the quasiparticle pole is valid and the relations ( [ omegapoletc]-[widthaway ] ) describe the properties of the quasiparticles .to lowest order in the @xmath0 expansion we find , using ( [ dynexpz ] ) that @xmath441 for @xmath124 , i.e , @xmath431 this ratio vanishes and @xmath442 diverges displaying the phenomenon of threshold singularity with a divergent residue @xmath333. for @xmath433 but @xmath443 ( @xmath444 ) this ratio equals that of the critical case ( see eqn .( [ width ] ) ) . for @xmath445we finally find the width of the long wavelength quasiparticles to be given to lowest order in the @xmath446 expansion by @xmath447^{\frac{1}{2}}\ ] ] with the following behavior to lowest order in the @xmath0 expansion @xmath448 thus , critical slowing down emerges in both limits , furthermore the validity of the quasiparticle picture is warranted in the @xmath0 expansion , since @xmath449 .critical phenomena , both static and dynamic in quantum field theory at finite temperature results in dimensional reduction since momenta and frequencies are @xmath158 and the correlation length is @xmath450 .the infrared physics is dominated by the contribution of the zero matsubara frequency in internal loops , which in turn results in an effective coupling @xmath34 in the perturbative expansion .naive perturbation theory at high temperature breaks down in four space - time dimensions because of the strong infrared behavior of loop diagrams near the critical point for long - wavelength phenomena .we propose an implementation of the renormalization group to study _ dynamical _ critical phenomena which hinges upon two main ingredients : * the leading infrared behavior near the critical point is determined by keeping only the zero matsubara internal frequency in the loops . to control the infrared consistently we implement an expansion in @xmath0 in @xmath38 space - time dimensions .dimensional reduction for long - wavelength phenomena near the critical point results in that the perturbative expansion is in terms of @xmath451 where @xmath37 is the scale of external momenta and frequencies in the diagram .the renormalized effective coupling is driven to a fixed point in the infrared which is of @xmath173 .therefore long - wavelength phenomena can be studied in perturbation theory around this fixed point for @xmath179 .the perturbative expansion is improved by implementing a renormalization group resummation which reveals dynamical scaling phenomena with anomalous dimensions .eventually the limit of physical interest @xmath196 must be studied by further borel and/or pad resummations . * the second important ingredient is the _ anisotropic _ scaling between space and time .while space is infinite , at finite temperature in the euclidean formulation the time direction is compactified to the interval @xmath452 $ ] .we introduce a new parameter , the effective speed of light in the medium , which is renormalized and runs with the renormalization transformations .the infrared renormalization of the speed of light results in a new _ dynamical _ anomalous exponent which determines the dispersion relation and all the quasiparticle properties .the @xmath0 expansion combined with the renormalization group leads to a consistent quasiparticle description of long - wavelength excitations near the critical point .the critical exponents , both static and dynamic are summarized below for the critical case @xmath46 as well as for @xmath453 but in the symmetric phase with @xmath454 .@xmath455 @xmath456^{2z\nu } + \left[\frac{p^2}{\mu^2 } \right]^{z } \\\hline \gamma_p ~ \propto & \eta_t~\omega_p~ \frac{(p\,\xi)^{2z}}{1+(p\,\xi)^{2z } } \\\hline \xi \propto & |t - t_c|^{-\nu } \\ \hline m^2_r(t ) ~\propto & |t - t_c| \\\hline \nu = & \frac{1}{2}+\frac{\epsilon}{12}+\mathcal{o}(\epsilon^2)~(\mathrm{static } ) \\\hline \eta_t ~ = & \frac{2\epsilon}{27}+\mathcal{o}(\epsilon^2 ) ~(\mathrm{dynamic } ) \\\hline z ~ = & \simeq 1+\frac{1}{2}(\eta_t-\eta ) = 1+\frac{\epsilon}{27}+\mathcal{o}(\epsilon^2 ) ~(\mathrm{dynamic } ) \\ \hline \end{array}\ ] ] the new dynamical exponent @xmath11 is missed by the wilsonian approach advocated in reference@xcite since two loops diagrams are completely neglected in that approach and anisotropic rescaling of frequency and momenta becomes manifest at two loop order and beyond . at this stagewe can generalize our results to the case of a scalar theory with @xmath12 symmetry at or slightly above the critical point .while the static critical exponents for the @xmath12 case are available in the literature@xcite , the dynamical critical exponent to lowest order in @xmath0 can be obtained simply by recognizing that the symmetry factors corresponding to the @xmath12 theory multiply the two loop expression for the self - energy by an overall factor . from the expression ( [ gammaeps ] ) we see that the coefficient of @xmath457 is a factor @xmath458 times the coefficient of @xmath459 , which immediately leads to the result @xmath460 since for the @xmath12 theory @xmath461 + \mathcal{o}(\epsilon^3)$ ] we find to lowest order in @xmath0 @xmath462 in summary the static and dynamic critical exponents to lowest order in the @xmath0 expansion for the @xmath12 theory are given by @xmath463we have studied the dynamical aspects of long - wavelength ( collective ) excitations at and near the critical point in scalar quantum field theories at high temperatures . after recognizing that naive perturbation theory breaks down at high temperature in the long - wavelength limit, we introduced an @xmath0 expansion around @xmath38 space time dimensions combined with the renormalization group at high temperature to resum the perturbative series .the effective long - wavelength theory at high temperature is described by a non - trivial fixed point at which the correlation functions feature scaling behavior .the anisotropy between spatial and time coordinates in euclidean space - time at finite temperature leads to consider the renormalization of the speed of light , which , in turn leads to a new _ dynamical _ exponent @xmath11 .all dynamical quantities , such as the dispersion relation and widths of long - wavelength quasiparticle ( collective ) excitations depend on this new dynamical exponent , as well as the static exponents .our results are summarized in the tables in the previous section .two very important aspects emerge from this treatment : i ) critical slowing down , i.e , the relaxation rate of the quasiparticle vanishes in the long - wavelength limit or at the critical point with definite anomalous dimensions determined by the new _ dynamical _ exponent @xmath11 and ii ) the quasiparticle picture , i.e , narrow widths @xmath464 is valid .the group velocity of quasiparticles vanishes at the critical point in the long - wavelength limit revealing the collective aspects of these excitations .the dynamical exponent @xmath52 describes a _ new universality class _ for _ dynamical _ critical phenomena in quantum field theory .as mentioned in the introduction these phenomena have phenomenological implications for the chiral phase transition in the quark gluon plasma with potential observational consequences if long - wavelength pion fluctuations freeze out at the chiral phase transition . an important aspect revealed by this programis that the effective coupling @xmath465 is driven to the wilson - fischer fixed point in the infrared , this in turn means that in this limit @xmath466 .this may be important in the linear sigma model description of low energy qcd near the critical point and may give rise to interesting phenomenological consequences . in this articlewe focused our attention on the approach to the critical temperature from above , therefore our results regarding the dynamical exponent @xmath11 are valid in the symmetric phase . an important question that we are currently addressing@xcite is the relaxation of pions slightly below @xmath346 .since the scattering amplitude of pions ( at zero temperature ) vanishes in the long - wavelength limit we expect novel behavior of critical slowing down for pion fluctuations below the critical temperature .we expect to report on our findings on these and other related issues soon@xcite . while we have provided a quantitative implementation of the program of the @xmath0 expansion _ with _the resummation via the renormalization group , the physical limit @xmath467 requires higher order calculations with borel or pade resummations much in the same way as in static critical phenomena .we have studied the dynamical aspects to lowest order in the @xmath0 expansion but clearly a formal proof of the consistency of the @xmath0 expansion to higher orders , just as in usual critical phenomena , must be explored .while clearly such programs are beyond the scope and goals of this article , we here provided the first steps of the program whose potential phenomenological implications as well as intrinsic interest in finite temperature quantum field theory warrant further study .* acknowledgements . * the work of d.b . was supported in part by the us nsf under grants phy-9988720 and nsf - int-9815064 , he thanks the lpthe at the universit de paris vi for hospitality during this work . h. j. de v. thanks the department of physics at the university of pittsburgh for hospitality .the authors thank m. simionato , s. b. liao , e. mottola and d. t. son for useful conversations .we derive in this appendix the behaviour of the one - loop diagram @xmath468 contributing to the four points function for high temperatures @xmath469 . setting @xmath470 to avoid cluttering of notation (the velocity of light is not relevant for the discussion in this section ) , we have from eq.([unlazo ] ) @xmath471 \left [ ( { \vec q}+{\vec p})^2 + ( 2 \pi \ ; t)^2 \ , ( n+l)^2 \right ] } \ ; , \ ] ] where @xmath472 is the number of spatial dimensions . the denominators in eq .( [ unlazoa ] ) can be combined using feynman parameters with the result @xmath473 ^ 2 } \cr \cr \cr & = & \frac{t}{2(4\pi)^{d\over 2 } } \gamma\left(2 - { d \over 2}\right ) \ ; \sum_{l \in { \cal z } } \int_0 ^ 1 dx \ ; \left[a_l(x , p , s)\right]^{{d \over 2}-2}\end{aligned}\ ] ] where we integrated over the spatial momenta and @xmath474 we single out now the contribution from the @xmath475 mode and study the behaviour of the sum over @xmath476 for large @xmath477 .let us first evaluate the @xmath478 term in the sum ( [ hps2 ] ) .@xmath479^{-{\epsilon \over 2 } } = \cr \cr & & = \frac{t}{2 ( 4\pi)^{2-{\epsilon\over 2 } } } \ ; \gamma\left({\epsilon \over 2}\right ) \ ; \int_0 ^ 1 x^{-{\epsilon\over 2 } } \ ; dx \ ; \left [ ( 1-x)p^2 + s^2 \right]^{-{\epsilon \over 2}}=\cr \cr & & = -\frac{t \mu^{-\epsilon}}{2 } { \gamma\left ( { \epsilon \over 2}-1\right ) \over \left ( { 4 \pi } \right)^{2-{\epsilon \over 2 } } } \left(\frac{s^2 + p^2 } { \mu^2 } \right)^ { - { \epsilon \over 2 } } \ ; f\left({\epsilon \over 2},1-{\epsilon \over 2};2-{\epsilon\over 2};{p^2 \over p^2 + s^2}\right ) \ ; , \end{aligned}\ ] ] where @xmath254 stands for the hypergeometric function@xcite .we have for the @xmath181 terms in the high temperature limit , @xmath480^{-{\epsilon \over 2 } } \buildrel{t \gg p , s}\over = \left ( 2 \pi t |l|\right)^{-\epsilon } - \mbox{sign}(l ) \ ; { \epsilon \ , x \ , \over |2\pi t l|^{1+\epsilon } } \cr \cr & & - { \epsilon \ , x \ , \over 2 \ , |2\pi t l|^{2+\epsilon } } \left\ { ( 1-x ) p^2 + s^2[1 - ( 2+\epsilon ) x ] \right\ } + { \cal o}\left(\left|t \ , l \right|^{-3 -\epsilon}\right ) \ ; . \end{aligned}\ ] ] the sum over @xmath181 then yields in the high temperature limit , @xmath481^{-{\epsilon \over 2 } } \buildrel{t \gg p , s}\over = \cr \cr & & = \frac{\left ( 2 \pi t \right)^{1-\epsilon } \gamma\left(\frac{\epsilon}{2}\right ) } { 2 \ , ( 4\pi)^{2-{\epsilon\over 2 } } } \ ; \ ; \left\ { \frac{\zeta(\epsilon ) } { \pi } \ ; -{\epsilon \ ; \zeta(2+\epsilon ) \over 12 \ , \pi \left ( 2 \pi t\right)^{2 } } \left [ p^2 - s^2(1 + 2 \epsilon ) \right ] + { \cal o}\left(t^{-3 -\epsilon}\right ) \right\}\end{aligned}\ ] ] we see that only the first term in the r.h.s . is important for @xmath482 and high temperature .this term is the dominant high temperature limit of the sum of nonzero matsubara modes .we then have , @xmath483 \left [ 1 + { \cal o}\left({p^2 , s^2 \over t^2}\right)\right ] \ ; , \ ] ] where @xmath484the renormalized two points function is given by ( all quantities are renormalized below ) @xmath485 where @xmath486 is the two - loops self - energy .using the renormalization conditions ( [ renconds ] ) we find for the wave function and velocity of light renormalizations , @xmath487 the two loops contribution to the self energy @xmath488 is given by @xmath489 \ ; [ k^2 + \frac{\omega^2_j}{v^2 } ] \ ; \left [ ( { \vec p}+{\vec k}+{\vec q})^2 + \frac{(\omega_l+\omega_j+s)^2}{v^2 } \right ] } \ ; .\ ] ] where @xmath490 .we combine the propagators in eq.([sigmata ] ) using feynman parameters and integrate over the momenta with the result @xmath491^{2 - { \epsilon \over 2 } } } \\ \cr & & \left\ { { x(1-x ) \xi(1-\xi ) \over 1 -x + x \xi(1-\xi)}\ ; p^2 + \left(\frac{2 \pi t}{v}\right)^2\left [ j^2 \ ; x\xi + l^2 \ ; x(1-\xi ) + ( j+l+n)^2 ( 1-x)\right ] \right\}^{1-\epsilon } \nonumber\end{aligned}\ ] ] using the definition ( [ dimenlessbare ] ) for the dimensionless coupling , to order @xmath492 we find from eqs . ( [ zetex ] ) and ( [ sigma2 ] ) @xmath493}{\left [ 1 -x + x\xi(1-\xi ) \right]^{3 - \frac{\epsilon}{2 } } } \cr \cr & & \left [ { x(1-x ) \xi(1-\xi ) \over 1 -x + x \xi(1-\xi)}\ ; \left({v\mu \over 2 \pi \,t}\right)^2 + j^2 \ ; x\xi + l^2 \ ; x(1-\xi ) + \left(j+l+{\mu \over 2 \pi \,t } \right)^2 ( 1-x ) \right]^{-\epsilon } \ ; , \cr \cr z_v & = & 1 - \frac{g^2}{3 } \ ; \left({2\pi t \over v\mu}\right)^{1 - 2\epsilon } \ ; \gamma(\epsilon ) \sum_{l , j \in { \cal z } } \int_0 ^ 1 dx \int_0 ^ 1 d \xi \ ; { x^{{\epsilon \over 2}-1 } ( 1 -x ) \ , \left(j+l+{v\mu \over 2 \pi \,t}\right ) \over \left [ 1 -x + x\xi(1-\xi ) \right]^{2 - { \epsilon \over 2 } } } \cr \cr & & \left [ { x(1-x ) \xi(1-\xi ) \over 1 -x + x \xi(1-\xi)}\ ; \left({v\mu \over 2 \pi \,t}\right)^2 + j^2 \ ; x\xi + l^2 \ ; x(1-\xi ) + \left(j+l+{v\mu \over 2 \pi \,t}\right)^2 ( 1-x ) \right]^{-\epsilon } \ ; . \end{aligned}\ ] ] we find from the definition of the anomalous dimension ( [ gammaexp ] ) and ( [ zetas ] ) for @xmath494 , @xmath495^{4 - { \epsilon \over 2 } } } \times \cr \cr & & \left ( 1-x + 2 \ , x \ , \xi(1-\xi ) \left [ 1 + { \pi\ , t\over v\mu}(j+l ) \right ] \right ) \times \\ \cr & & \left\ { { x(1-x ) \xi(1-\xi ) \over 1 -x + x \xi(1-\xi ) } + \left({2\pi\ , t\over v\mu } \right)^2\left [ j^2 \ ; x\xi + l^2 \ ; x(1-\xi ) + \left(j+l+{v \mu \over 2 \pi \,t}\right)^2 ( 1-x ) \right]\right\}^{-1-\epsilon } + { \cal o}(g^3 ) \ ; . \nonumber\end{aligned}\ ] ] we split the expression for @xmath496 as follows , @xmath497 where @xmath498 is the contribution from the zero matsubara mode in eq.([bega2 ] ) @xmath499^{3 - { 3\epsilon \over 2 } } \left [ 1 -x + 2\ , x\xi(1-\xi ) \right]^{\epsilon } } + { \cal o}(g^3)\ ] ] and @xmath500 stands for the contribution of the non - zero matsubara modes . for @xmath501, we see from eq.([bega2 ] ) that @xmath502 decreases as @xmath503 .[ notice that the coefficient of @xmath504 vanishes by symmetry when summing over @xmath505 .therefore , @xmath498 dominates for @xmath506 can be easily computed for small @xmath42 with the result , @xmath507 ^ 3 } + { \cal o}(\epsilon \ , g^2,g^3 ) \cr \cr & & = { g^2 \over 6 } + { \cal o}(\epsilon \ , g^2,g^3 ) \end{aligned}\ ] ] therefore , @xmath508 to the lowest non - trivial order in @xmath155 , that is @xmath509 ( two - loops ) , we find for the function @xmath510 @xmath511 where the derivatives are now at constant ( bare ) @xmath155 .using eq.([zetas ] ) for @xmath512 yields @xmath513^{2 - { \epsilon \over 2 } } } \ ; q_{j , l}(x,\xi)^{-\epsilon } \cr \cr & & \left\{\left [ { x \xi(1-\xi ) \over 1 -x + x \xi(1-\xi ) } + 1 \right ] { ( 1-x)\epsilon \over \pi \ ; q_{j , l}(x,\xi ) } + \left [ 1 + { 2\ , \epsilon \ , ( 1-x)\over q_{j , l}(x,\xi ) } \right](j+l ) \ ; { t \over v\mu } \right\}\end{aligned}\ ] ] where , @xmath514 + ( 1-x ) \left [ { 2 \pi \,t\over v\mu}(j+l ) + 1 \right]^2 \ ; .\ ] ] we find in the high temperature limit @xmath515 that this expression is dominated by its zero mode contribution @xmath516 ( corresponding to @xmath517 ) , @xmath518^{2 - { \epsilon \over 2 } } } \left [ { x \xi(1-\xi ) \over 1 -x + x \xi(1-\xi ) } + 1 \right]^{-\epsilon}\ ] ] which turns out to be @xmath33-independent . the sum of non - zero terms gives a subdominant contribution for @xmath519 and @xmath520 * strictly positive*. we find from eq.([wtot ] ) after calculation , @xmath521^{2 - { \epsilon \over 2 } } } \times \cr \cr & & \sum_{l , j \in { \cal z } } { j^2 \over \left[j^2(1-x+x\xi ) + l^2 \ , x -2jlx\xi\right]^{1+\epsilon } } \left[1 + { \cal o}({\mu^2 \over t^2 } , \ ; g)\right ] \ ; .\end{aligned}\ ] ] for @xmath522 and for @xmath515 , @xmath516 and therefore @xmath523 are dominated by the pole of @xmath516 at @xmath524 .that is @xmath525 \ ; , \ ] ] where we used that @xmath526 .therefore , we find for @xmath527 from eqs.([betava ] ) , ( [ wtot ] ) and ( [ wfin ] ) , @xmath528 + { \cal o}\left[\left({\mu \over t}\right)^{2 \epsilon } \right]\ ; .\ ] ]the formal proof to one loop order begins with the expression ( [ unlazoa ] ) from appendix [ sec : appendix1 ] above . with @xmath530 and the contour @xmath531is displayed in figure ( [ fig : contour ] ) below .the function @xmath532^{-\frac{\epsilon}{2}}$ ] has a cut running parallel to the real axis which for @xmath533 or in the massive case for arbitrary @xmath2 begins away from the imaginary axis and the contour @xmath531 .the contour can now be deformed to wrap around the cut and the analytic continuation @xmath534 can be performed . for @xmath158 the infrared behavior is dominated by @xmath535 for which @xmath536 \sim 2t / k_0 $ ] and the resulting expression features a pole at @xmath537 while the cut begins away from the origin .the cut can be deformed again to circle the origin and the integral is simply the residue at the pole @xmath538 .therefore the infrared dominant term is given by @xmath539^{-\frac{\epsilon}{2}}$ ] , result which coincides with the analysis in terms of the matsubara sums provided in appendix [ sec : appendix1 ] .see for example : f. karsch , _ lattice qcd at high temperature and density _ lectures given at 40th internationale universitatswochen fuer theoretische physik : dense matter ( iukt 40 ) , schladming , styria , austria , 3 - 10 mar 2001 , hep - lat/0106019 , and references therein .h. meyer - ortmanns , rev . of mod68 * , 473 ( 1996 ) ; b. muller , _ the physics of the quark - gluon plasma _ , lecture notes in physics , vol. 225 ( springer - verlag , berlin , 1985 ) ; l.p .csernai , _ introduction to relativistic heavy ion collisions _( john wiley and sons , england , 1994 ) ; c.y .wong , _ introduction to high - energy heavy ion collisions _( world scientific , singapore , 1994 ) .d. boyanovsky , d .- s .lee and a. singh , phys . rev . *d 48 * , 800 ( 1993 ) ; d. boyanovsky , h. j. de vega and r. holman , in : ` topological defects and the non - equilibrium dynamics of symmetry breaking phase transitions ' , eds .y. m. bunkov and h. godfrin , nato science series , kluwer , 2000 ) ( hep - ph/9903534 ) ; d. boyanovsky , h. j. de vega and r. holman , phys . rev .* d51 * , 734 ( 1995 ) .bowick , m. , momen , a. , phys.rev . *d58 * ( 1998 ) 085014 ; m. bowick , a. cacciuto and a. travesset _ the formation of vortex loops ( strings ) in continuous phase transitions _ , cond - mat/0107188 ( 2001 ) ( to appear in phys .e ) .a. v. chubukov , s. sachdev and j. ye , phys . rev . *b 49 * 11919 ( 1994 ) ; s. sachdev , phys . rev . *b 59 * 14054 ( 1999 ) ; s. sachdev , _ ` quantum phase transitions ' _ ( cambridge university press , cambridge , 1999 ) .

<S> the perturbative approach to the description of long wavelength excitations at high temperature breaks down near the critical point of a second order phase transition . </S><S> we study the _ dynamics _ of these excitations in a relativistic scalar field theory at and near the critical point via a renormalization group approach at high temperature and an @xmath0 expansion in @xmath1 space - time dimensions . </S><S> the long wavelength physics is determined by a non - trivial fixed point of the renormalization group . at the critical point </S><S> we find that the dispersion relation and width of quasiparticles of momentum @xmath2 is @xmath3 and @xmath4 respectively , the group velocity of quasiparticles @xmath5 vanishes in the long wavelength limit at the critical point . </S><S> away from the critical point for @xmath6 we find @xmath7^{\frac{1}{2}}$ ] and @xmath8 with @xmath9 the finite temperature correlation length @xmath10 . </S><S> the new _ dynamical _ exponent @xmath11 results from anisotropic renormalization in the spatial and time directions . for a theory with @xmath12 symmetry </S><S> we find @xmath13.this dynamical critical exponent describes a new universality class for dynamical critical phenomena in quantum field theory . </S><S> critical slowing down , i.e , a vanishing width in the long - wavelength limit , and the validity of the quasiparticle picture emerge naturally from this analysis . </S>

1005.2554

half - metallic ( hm ) ferromagnetism@xcite is an ideal property of materials from the viewpoint of their application in spintronic devices@xcite .thus a considerable research has been devoted in recent years to the designing and measurement of the physical properties of such materials .one group of such materials is known as the dilute magnetic semiconductors ( dms)@xcite , where the magnetic properties are mediated by the percolation effects connecting a very small concentration of transition metal ( tm ) atoms such as mn or cr in a semiconductor host .another group , that has been studied extensively , consists of tm - semiconductor binary / ternary alloys , where the tm concentration can be as high as the other components@xcite .in particular , some zb structure transition metal chalcogenides and pnictides have been predicted to be hm ferromagnets on the basis of theoretical calculations@xcite . herethe magnetic properties of the transition metal matrix is modified via the hybridization between the transition metal 3-@xmath2 and the @xmath3 and @xmath4 orbitals of the pinctogen or the chalcogen .half - metallicity arises as a result of interplay between the exchange splitting and the hybridization . some cr - based chalcogenides and pnictides@xcite ,have been predicted to be not only hm but also potential candidates for possessing high curie temperatures @xmath5 .akinaga @xcite have reported growing zb thin films of cras on gaas ( 001 ) substrates via molecular beam epitaxy .these showed ferromagnetic behavior at temperatures in excess of 400 k and magnetic moments of 3@xmath6 per cras unit .thin films of crsb grown similarly on gaas , ( al , ga)sb , and gasb substrates have been found to be of zb structure and ferromagnetic with @xmath5 higher than 400 k@xcite .deng @xcite were successful in increasing the thickness of zb - crsb films to @xmath7 3 nm using ( in , ga)as buffer layers , and li @xcite were able to grow @xmath7 4 nm thick zb - crsb films on nacl ( 100 ) substrates .two of the present authors have recently presented an extensive theoretical study of the exchange interactions and curie temperatures in a series of cr - based pnictides and chalcogenides in zb structure as a function of the lattice parameter@xcite .one important issue is that although the zb phase of these compounds shows half - metallicity and often ( the promise of ) high curie temperature , the ground state usually has a different structure . for many of these compoundsthe ground state is nias - type hexagonal ( na ) .the semiconductor substrates on which thin films of these compounds are grown have cubic structures , usually rs or zb . the other point is that often the equilibrium volume does not show half - metallicity , which appears only at an expanded volume . from this stand - pointit would be useful to explore the possibility of the half - metallic phase in structures other than zb , and in particular the rs structure . in this workwe have examined this problem for crte .zhao and zunger@xcite argued that zb crte should be epitaxially unstable against the na structure .however , recently sreenivasan @xcite and bi @xcite have reported growing @xmath75 nm thick thin films of zb crte by low temperature molecular beam epitaxy .their measurements of temperature - dependent remanent - magnetization indicate a curie temperature of @xmath7100 k in these films .one attractive feature of bulk crte , at least from a theoretical view point , is that the equilibrium volume in the zb structure is already half - metallic , whereas most other cr - based chalcogenides and pnictides show half - metallicity only at expanded volumes@xcite . besides , the half - metallic gap is larger in the chalcogenides like crte or crse than in the pnictides such as cras or crsbthis is due to larger cr - moment in the half - metallic phase for the chalcogenides ( @xmath74@xmath6 ) than for the pnictides ( @xmath73@xmath6 ) , which results in larger exchange splitting .our previous study@xcite has revealed that chalcogenides such as crs and crse have a high likelihood of exhibiting antiferromagnetic or complex magnetic behavior for a range of the lattice parameter , whereas such tendencies are less pronounced in crte .hence , the focus of this work is on crte .we study the appearance of half - metallicity in three competing structures for crte : na , rs and zb .the ground state structure ( na ) shows half - metallicity not at the equilibrium volume but at some expanded volume , while both rs and zb structures show half - metallicity at their respective equilibrium ( lowest energy ) volumes .the rs structure is found to be more stable than the zb , with an energy that is lower by 0.25 ev / atom .some band calculations for crte in the na structure have already appeared@xcite in the literature , and the same goes for zb crte@xcite .ahmadian @xcite have recently discussed half - metallicity at a zb crte(001 ) surface and its interface with znte(001 ) .so far crte in the rs structure has not been studied from the viewpoint of its electronic and magnetic properties .although both rs and zb structures are fcc - based , their magnetic properties should differ somewhat for the same lattice parameter due to different levels of hybridization between the cr 3-@xmath2 and te @xmath8-orbitals .this should influence the magnetic moments , half - metallic gaps and the exchange interactions .we find that the exchange interactions are predominantly ferromagnetic and strong in the rs structure , while in the zb structure cr - cr interactions can have non - negligible antiferromagnetic components for some lattice parameters .this results in a lower curie temperature @xmath5 for the zb structure over a wide range of the lattice parameter .the promise of lower energy and higher @xmath5 should prompt the experimentalists to try to fabricate crte thin films in the rs structure .the two fcc - based structures zb ( @xmath9 , no.@xmath10 ) and rs ( @xmath11 , no.@xmath12 ) are well - known .the na structure(p6@xmath13/mmc , no.@xmath14 ) is based on a distorted hcp array of the arsenide anions .by contrast with the wurtzite structure , which shares the same formula type ( ab ) with the na structure , the cations occupy all the octahedral sites rather than half the tetrahedral holes .there is one octahedral hole for each hcp lattice site , and so the ab stoichiometry is preserved . the local co - ordinations of the anions and cations are different in the na and wurtzite structures .bulk samples of crte are known to crystallize in this structure@xcite with cell parameters @xmath15 , @xmath16@xmath17 at room temperature .the cr - te bond length is 2.78@xmath17 and the @xmath18 ratio is 1.557 , different from the ideal hcp value of 1.633 .thus the six te atoms form a trigonally distorted octahedron around cr .the cr - cr nearest neighbor distance in the @xmath19 plane is @xmath15 , while the out of plane distance is @xmath20 .readers may consult fig . 1 of ref .[ ] to view the structure . for the electronic structure studywe have used the wien@xmath21k code @xcite , which employs the full - potential linear augmented plane waves ( fp - lapw ) plus local orbitals method . the generalized gradient approximation ( gga ) proposed by perdewwas used for exchange and correlation potentials@xcite . according to a study by continenza @xcite , gga is essential for obtaining accurate equilibrium volumes and magnetic moments for mn - based pnictides of zb and na structures .we consider full relativistic effects for the core states and use the scalar approximation for the valence states .spin - orbit coupling is neglected , as it should have negligible effect on the results for crte .convergence with respect to basis set and k - point sampling was carefully verified . about @xmath22 k pointswere used for the brillouin zone ( bz ) integrations , using the monkhorst - pack scheme@xcite and @xmath23 divisions of the bz .we used @xmath24 and angular momentum expansion up to @xmath25 in the muffin tins , and @xmath26 for the charge density .all core and valence states were treated self - consistently .for the calculations involving the different structures the muffin - tin ( mt ) radii were taken to be equal to @xmath27 and @xmath28 bohrs for cr and te atoms , respectively .the same mt radii were used in the calculations of different structures at their equilibrium volumes in order to compare accurately the total energies .self - consistency in charge density was achieved to an accuracy higher than @xmath29 .we carried out a systematic structural optimization of na , zb , and rs crte in fm , nm and afm configurations .for the na structure the c / a ratio was optimized and the result @xmath30 for the fm phase is in good agreement with previously obtained result@xcite .this optimized c / a ratio was used for the final na structure volume optimization .bulk modulus was calculated by fitting the volume vs. energy results to the empirical murnaghan equation of state@xcite .we find that for all three structures the fm phase has lower energy than the nm and afm phases . to assess the relative stability of crte in different structures we consider the total energy per formula unit in the zb and rs structures relative to the na structure .the calculated optimized equilibrium lattice constant a , bulk modulus b , the total energy difference e@xmath31 between the nm and the fm states : @xmath32 , and e@xmath33 between the afm and the fm states : @xmath34 , and the metastability energy @xmath35 , which is the total energy per formula unit relative to the na structure , are given in table[table1 ] .note that for rs and zb structures , in calculating @xmath36 we have considered the energies of three afm configurations : [ 001],[110 ] and [ 111 ] and considered the lowest of these energies .for the hexagonal na structure , the afm state was created by simply ascribing equal but opposite magnetic moments to the two cr atoms in the unit cell .cohesive properties are dependent on the treatment of the exchange - correlation effects in a non - trivial manner , as indicated by the lda values of the lattice parameter and bulk modulus shown in parentheses in table[table1 ] for the rs and zb equilibrium structures , where all other values are from gga . increased correlation effects in gga increases the equilibrium lattice parameter and lowers the bulk modulus .[ table1 ] [ cols="<,^,^,^,^,^,^,^",options="header " , ] in lmto - asa method the interstitial region is described by empty spheres and the electronic properties of this region is primarily dictated by the tails of the orbitals centered at the real atoms such as cr and te . as in the fp - lapw calculation , we find that the magnetic moments of the empty spheres are parallel to those of the cr atoms , while those of the te atoms are antiparallel .for the rs structure , the moment on the te atom is small ( around 1 - 2% of that of the cr atom ) for low values of the lattice parameter , while the moment on the two empty spheres are each about 3.5 - 4% of that of the cr atom . with increasing lattice parameter , the te - moment increases with respect to the cr - moment to about 7 - 8% at the highest lattice parameter studied .at these high lattice parameters the moments on the empty spheres , originating from the tails of the nearby cr-3@xmath2 orbitals , drop to about 2% of the cr - moment . in the zb structure , the two empty spheres are not equivalent and the induced moments on the empty spheres and the te atoms in this case has been described in detail in our recent publication@xcite . for the zb structurethe induced moment on te is negligible for small lattice parameters , increases steadily with increasing lattice parameter , achieving a maximum value of 4% of the cr - moment at the highest value of the lattice parameter studied .most often it is less than 1% .the magnitudes of the induced moments on the empty spheres decrease in the following order : sublattice es-1 ( the sublattice of empty spheres that is at the same distance with respect to the cr - sublattice as the te - sublattice ) , sublattice es-2 ( sublattice of empty spheres further away from the cr - sublattice ) .the induced moment on es-2 is always small , decreases with increasing lattice parameter from about 2% to about 0.4% of the cr - moment .the induced moment on es-1 is higher than the other two induced moments for smaller lattice parameters . for larger lattice parametersit is comparable to the te - moment , but opposite in sign . above a critical value of the lattice parameter , the moment per formula unit ( f.u . )saturates at a value of 4.0 @xmath6 , as the hm state is achieved , while the local moments on the cr- and te - sublattices increase in magnitude , remaining opposite in sign .sandratskii @xcite have discussed the problem associated with such induced moments in calculating the curie temperature .lmto - asa results using lda for the zb and rs structures are presented in fig.[fig8 ] .magnetism is strongly dependent on correlation effects . even in the metallic states , where magnetism is believed to be described reasonably well by density functional methods and lda , different treatment of the exchange - correlation effectslead to different values of magnetic moments .we thus compare the magnetic moments of our lmto - asa calculations for lda and lda+u@xcite .the enhanced correlation effects in the lda+u calculations ( using u=0.075 ry , j=0 ) increase the values magnetic moments and bring the results in general close agreement with those obtained via using gga in the wie2@xmath37 package .this is shown in fig.[fig9 ] .the agreement between gga and lda+u results is due to the fact that the gga goes beyond the lda by including electron correlations in some way .exchange interactions are usually calculated by mapping @xcite the system energy onto a classical heisenberg model : @xmath38 where @xmath39 are site indices , @xmath40 is the unit vector pointing along the direction of the local magnetic moment at site @xmath41 , and @xmath42 is the exchange interaction between the moments at sites @xmath41 and @xmath43 . in this work ,we use the method of liechtenstein @xcite , based on multiple - scattering formalism , which was later implemented for random magnetic systems by turek , using cpa and the tb - lmto method@xcite . for a discussion of the various methods used in the mapping procedurethe reader may consult our previous publication@xcite .the exchange integral in eq.([e1 ] ) is given by @xmath44 dz \ ; , \ ] ] where @xmath45 represents the complex energy variable , @xmath46 , and @xmath47 , representing the difference in the potential functions for the up and down spin electrons at site @xmath41. in the present work @xmath48 represents the matrix elements of the green s function of the medium for the up and down spin electrons .the negative sign in eq.([e1 ] ) implies that positive and negative values of @xmath42 are to be interpreted as representing ferromagnetic and antiferromagnetic interactions , respectively . for the rs structure , obtained in the fp - lapw ( wien2k ) calculation.,width=336 ] in the mean - field approximation ( mfa ), the curie temperature @xmath5 , based on the exchange interaction between the strong moments only , i.e. the cr atoms , can be written as .@xmath49 where the sum extends over all the neighboring shells centered around the cr - site o. an improved description of finite - temperature magnetism is provided by the random phase approximation ( rpa)@xcite , with @xmath50 given by @xmath51^{-1 } \ ; .\ ] ] here @xmath52 denotes the order of the translational group applied and @xmath53 is the lattice fourier transform of the real - space exchange integrals @xmath54 .@xmath55 can be shown to be smaller than @xmath56 @xcite , and has been found to be in close agreement with results given by monte - carlo simulations @xcite . unless otherwise stated , the @xmath5 estimates given in this paper are based on rpa , as given by eq.([e3 ] ) .sandratskii@xcite have discussed the case when , in addition to the interaction between the strong moments , there is one secondary , but much weaker , interaction between the strong and one induced moment . in this case , the curie temperature , calculated under the mean - field approximation ( mfa ) , is enhanced due to this secondary interaction , irrespective of the sign of the secondary interaction . in other words , the curie temperature would be somewhat higher than that calculated by considering only the interaction between the strong moments .the corresponding results under the random phase approximation ( rpa ) have to be obtained by solving two equations simultaneously .one can assume that the rpa results for the curie temperature follow the trends represented by the mfa results , being only somewhat smaller , as observed in the absence of induced moments . in fig.[fig10] we compare the exchange interactions between the cr atoms obtained by using eq.([eq - jij ] ) for the equilibrium lattice parameter of 5.727 @xmath17(fp - lapw gga result ) of rs crte with those in zb crte for two values of the lattice parameter , one above and the other below the value 5.727 . for the rs structurethe interactions are all ferromagnetic , while for the zb structure they are predominantly antiferromagnetic in this range of the lattice parameter . in our previous publicationwe have shown that due to these antiferromagnetic interactions , the curie temperature in zb crte is drastically reduced for a certain range of the lattice parameter .the ground state for this structure may even have a more complex magnetic phase than simple ferromagnetic . thus for lattice parameters around 5.74 - 5.75 @xmath17 rs crte is not only guaranteed to be a half - metallic ferromagnet , but also should have a higher @xmath5 than zb crte of similar volume per atom . in fig.[fig11] we compare the @xmath55 values for rs and zb crte obtained by using lmto - asa and lda .the reference state used for the calculation is the fm state . for the zb phase , the use of eq.(4 ) neglecting all the induced moments , results in unphysical negative @xmath5 in the lattice parameter range @xmath75.7 - 6.0 @xmath17 because of dominant antiferromagnetic cr - cr exchange at the nearest and next nearest neighbor sites .the zb phase has been discussed in detail in our previous publication@xcite , where we showed that in spite of these nearest and next nearest neighbor afm interactions , the fm state is lower in energy than the afm[001 ] and afm[111 ] states .the cumulative effect of the fm interactions involving the more distant pairs of cr atoms , as well as that due to the induced moments , may stabilize the fm phase .either the cumulative effects of distant pair interactions make the system fm or the magnetic state is other than afm[001 ] and afm[111 ] , possibly complex .if the magnetic state is indeed fm , then according to the prescription of sandratskii@xcite @xmath5 should be enhanced with respect to the value given by eq.(4 ) and may be positive but small .this explains why sreenivasan @xcite and bi @xcite observe @xmath5s only as high as 100 k in their thin film sample of zb crte . for rs crte , on the other hand ,all dominant interactions are fm , and as a result the @xmath5 calculated by using eq.(4 ) and based on only cr - cr interactions give positive values .the actual @xmath5s are most probably higher than what is shown in fig.[fig11 ] .the effect of enhanced correlation for rs crte incorporated via lda+u method , using u=0.075 ry , and j=0 , is shown in fig.[fig12 ] . within the uncertainties of our resultsthere seems to be no appreciable difference between the predictions of lda and lda+u .it would transpire from fig.[fig10 ] that a j@xmath57-j@xmath58 model may suffice to describe the magnetism of the rs phase .however , for quantitative accuracy of @xmath5 the more distant interactions , although individually small compared to the first two , can not be neglected , i.e. their cumulative effect is non - negligible .note that for the zb phase a j@xmath57-j@xmath58 model would clearly be inadequate .for the rs phase , if one is prepared to neglect interactions beyond the next nearest neighbor pairs , then the nearest neighbor interaction j@xmath57 and the next nearest interaction j@xmath58 can be obtained from the energies of the fm and afm[001 ] and afm[111 ] states@xcite .we have calculated j@xmath57 and j@xmath58 using this procedure for the fp - lapw - gga equilibrium lattice parameter of 5.727 , and the results are : j@xmath57=3.76 mev , j@xmath58=27.63 mev ( fp - lapw - gga ) , j@xmath57=3.29 mev , j@xmath58=18.32 mev ( fp - lapw - lda ) , j@xmath57=4.61 mev , j@xmath58=20.0 mev ( lmto - asa - lda ) . to compare these results with those given by the procedure based on the multiple scattering formalism and eq.([eq - jij ] ) we need to consider that all the distant neighbor interactions between the cr atoms ( and the interactions involving the induced moments ) get mapped effectively on to just two interactions in the j@xmath57-j@xmath58 model .this is tantamount to reproducing the band structure of a long - ranged hamiltonian via a tb - model with only nearest and next - nearest neighbor interactions .success is possible , but not always guaranteed ( unless further interactions are included ) . a meaningful way to compare the two approaches is to compute the dispersion @xmath59 . in fig.[fig13 ] we show these dispersion curves for rs crte along the various symmetry directions and all the different cases considered .the values of the exchange interactions quoted above for the j@xmath57-j@xmath58 model would put the @xmath5 for the lattice parameter 5.727 @xmath17 around @xmath71000 k. one reason that the magnitudes of the @xmath60 values based on eq.([eq - jij ] ) come out smaller than those given by the j@xmath57-j@xmath58 model is that in computing @xmath60 from eq.([eq - jij ] ) we have neglected all interactions involving the induced moments . ) ) as a function of the lattice parameter for the zb and rs structures of crte , calculated by considering the cr - cr interaction only and by using the fm reference state .the arrows indicate the equilibrium lattice parameter values according to various calculations .the results for @xmath5 are somewhat affected by the induced moments on the te atoms as well as the empty spheres used in the lmto - asa calculation . for zb crte , this error , combined with strong antiferromagnetic spin fluctuation for the lattice parameters in the range @xmath75.7 - 6.0 , result in unphysical negative values of @xmath5 in this range ( see text for details ) .away from this range @xmath5 values are reliable , although somewhat underestimated.,width=360 ] and the various cases discussed in the text . ,width=384 ] in our previous publication@xcite we have argued that estimates of @xmath5 from above the transition can be obtained by considering the reference state for the exchange calculation in the method of liechtenstein @xcite to be the paramagnetic state described by the disordered local moments ( dlm)@xcite model . in this descriptionmagnetic moments appear only on the magnetic , i.e. cr , atoms , but are oppositely directed for half of the randomly chosen atoms , giving a zero moment in total .the randomness is treated using the cpa@xcite .@xmath5 given by the dlm reference states are usually higher than their fm counterparts .the dlm estimates of @xmath5 are free from errors originating from the neglect of the exchange interactions involving induced moments in eqs.([e2 ] ) and ( [ e3 ] ) .for this reason we have computed the @xmath5 using the dlm model as well and compare these with the @xmath5 obtained for the fm reference states in fig.[fig14 ] . as expected ,the dlm values are usually higher for most of the lattice parameters .since the ground state of rs crte is known to be fm for the entire range of the lattice parameters studied , we expect the fm results to be valid , although somewhat underestimated due to the neglect of the induced moments .thus the theoretical estimates of @xmath5 should lie between those given by the fm and dlm results .experimental results on cr - based chalcogenides and pnictides show that these estimates are on the high side .one should note , however , that second order phase transition in a bulk material is different from that in thin films grown on substrates and @xmath5s may differ even though the two transitions belong to the same universality class .a point to note is that for larger volumes per atom the dlm energy is lower than the nm energy , with a cr - moment that is slightly less than that for the fm state of the same volume .dominating @xmath42s for the dlm reference states at these volumes are afm - like . for smaller volumesthe cr - moment disappears , indicating strong hybridization between the orbitals on cr- and te - sublattices . as a function of lattice parameter for fm anddlm reference states in rs crte , width=384 ]we have provided a comparison of the magnetic properties of crte in three different structures , na , rs and zb , at their respective equilibrium volumes .a detailed comparison of the half - metallic properties of the zb and rs phases is provided , as thin films grown on most substrates are likely to have either of these two structures .the rs phase is more stable than the zb phase , in that the lowest energy state in the rs phase is about 0.25 ev / atom lower than that for the zb phase . for some range of the lattice parameter ,namely between 5.6 - 6.1 @xmath17 , in the zb phase there are significant antiferromagnetic interactions between the cr atoms .the rs phase appears to be free from such strong antiferromagnetic interactions .ferromagnetism appears to be more robust in the rs phase than in the zb phase .increased antiferromagnetic interactions in the zb phase result from the smaller distance between the cr and the te atoms and consequently increased hybridization between the cr-@xmath61 and te-@xmath62 orbitals in the zb phase than in the rs phase for the same lattice parameter . increased stability with respect to the zb phase , stronger ferromagnetism and the promise of a higher @xmath5 should make the growth and measurements of the physical properties of rs crte an attractive project for experimentalists in the field .this work was supported by a grant from the natural sciences and engineering research council of canada . j.k .acknowledges financial support from av0z 10100520 and the grant agency of the czech republic ( 202/09/0775 ) .computational facility for this work was provided by sharcnet , canada .99 w. e. pickett and j. s. moodera , phys . today * 54*(issue 5 ) , 39 ( 2001 ) .i. @xmath63utic , j. fabian , and s. das sarma , rev .76 * , 323 ( 2004 ) .g. a. prinz , science * 282 * , 1660 ( 2004 ) .s. a. wolf , d. d. awschalom , r. a. buhrman , j. m. daughton , s. von molnar , m. l. roukes , a. y. chtchelkanova , d. m. treger , science * 294 * , 1488 ( 2001 ) . h. saito , v. zayets , s. yamagata , and k. ando , phys . rev. lett . * 90 * , 207202 ( 2003 ) .see k. sato , t. fukushima and h. katayama - yoshida , j. phys . :matter * 19 * , 365212 ( 2007 ) , and references therein .see b. belhadji , l. bergqvist , r. zeller , p.h .dederichs , k. sato and h. katayama - yoshida , j. phys . :matter * 19 * , 436227 ( 2007 ) , and references therein .r. q. wu , j. appl .* 97 * , 063911 ( 2005 ) . l. kronik , m. jain , and j. r. chelikowsky , , 041203(r ) ( 2002 ) .yong liu and bang - gui liu , j. magn .magn . mater . *307 * , 245 ( 2006 ) .n. e. brener , j. m. tyler , j. callaway , d. bagayoko , and g. l. zhao , , 16582 ( 2000 ) .g. m. zhao , h. keller , w. prellier , and d. j. kang , , 172411 ( 2001 ) .xie , y - q .xu , b - g .liu , and d.g .pettifor , 037204 ( 2003 ) .liu , lecture notes in physics vol .* 676 * ( springer , berlin , 2005 ) , pages 267 - 291 , and references therein .i. galanakis , p. h. dederichs , and n. papanikolaou , , 134428 ( 2002 ) ; _ ibid . _* 66 * , 174429 ( 2002 ) .i. galanakis , , 012406 ( 2002 ) .i. galanakis and p. mavropoulos , , 104417 ( 2003 ) .b. sanyal , l. bergqvist , and o. eriksson , , 054417 ( 2003 ) .e. z. kurmaev , a. moewes , s. m. butorin , m. i. katsnelson , l. d. finkelstein , j. nordgren , and p. m. tedrow , , 155105 ( 2003 ) . g. a. dewijs and r. a. degroot , , 020402(r ) ( 2001 ) .m. nakao , , 214429 ( 2004 ) .j. c. zheng and j. w. davenport , , 144415 ( 2004 ) .m. i. katsnelson , v. yu .irkhin , l. chioncel , a. i. lichtenstein , and r. a. de groot , rev .* 80 * , 315 ( 2008 ) . h. akinaga , t. manago , and m. shirai , jpn .phys.*39 * , l1118 ( 2000 ) .s. li , j - g duh , f. bao , k - x liu , c - l kuo , x. wu , liya l , z. huang , and y du , j. phys .* 41 * 175004 ( 2008 ) .m. shirai , j. appl . phys . * 93 * , 6844 ( 2003 ) .h. akinaga , m. mizuguchi , k. nagao , y. miura , and m. shirai in _springer lecture notes in physics _ * 676 * , 293 - 311 ( springer - verlag , berlin 2005 ) .k. yamana , m. geshi , h. tsukamoto , i. uchida , m. shirai , k. kusakabe , and n. suzuki , j. phys . : condens .matter * 16 * , s5815 ( 2004 ) .l. kahal , a. zaoul , m. ferhat , j. appl . phys .* 101 * , 093912 ( 2007 ) .pask , l.h .yang , c.y .fong , w.e .pickett , and s. dag , , 224420 ( 2003 ) .t. ito , h. ido , and k. motizuki , j. mag . mag . mat . * 310 * , e558 ( 2007 ) .l - j shi and b - g liu , j. phys .: condens .matter * 17 * , 1209 ( 2005 ) .m. zhang , j. phys . :matter * 15 * , 5017 ( 2003 ) .j. kbler , , 220403(r ) ( 2003 ) .zhao , f. matsukura , k. takamura , e. abe , d. chiba , and h. ohno , appl .. lett . * 79 * , 2776 ( 2001 ) .deng , j.h .zhao , j.f .niu , f.h .yang , x.g .wu , and h.z .zheng , j. appl .* 99 * , 093902 ( 2006 ) .bose and j. kudrnovsk , , 054446 ( 2010 ) y - j .zhao and a. zunger , , 132403 ( 2005 ) .sreenivasan , j.f .teo , and t. liew , j. appl . phys . *103 * , 043908 ( 2008 ) .sreenivasan , k.l .teo , and t. liew , j. phys .phys . * 41 * , 045002 ( 2008 ) .j. dijkstra , h. h. weitering . c. f. van bruggen ,c. haas and r. a. de groot , j. phys . :matter * 1 * , 9141 ( 1989 ) .f. ahmadian , m.r .abolhassani , s.j .hashemifar , and m. elahi , j. magn .magn . mater .* 322 * 1004 ( 2010 ) .m. chevreton , e.f .bertaut , and f. jellinek , acta crystallogr .* 16 * , 431 ( 1963 ) .p. blaha , k. schwarz , p. sorantin and s. b. trickey , comput . phys . commum . * 59 * 399 , ( 1990 ) .p. blaha , k. schwarz , g. k. h. madsen , d. kvasnicka and j. luitz , wien2k , an augmented - plane - wave+ local orbitals program for calculating crystal properties , ( karlheinz schwarz , tech .wien , austria ) isbn 3 - 9501031 - 1 - 2 , 2001 .j. p. perdew , k. burke and m. ernzerhof , , 3865 ( 1996 ) .a. continenza , s. picozzi , w. t. geng and a. j. freeman , , 085204 ( 2001 ) .j. monkhorst , and j. d. pack , , 5188 ( 1976 ) .f. d. murnaghan , proc .natl . acad .sci . * 30 * , 244 ( 1944 ) .see , e.g. , o.k .andersen , o. jepsen , and d. gltzel , in _ highlights of condensed matter theory _ , edited by f. bassani , f. fumi , and m.p .tosi ( north - holland , amsterdam , 1985 ) , p.59 . j. kudrnovsk and v. drchal , , 7515 ( 1990 ) . i. turek , v. drchal , j. kudrnovsk , m. ob , and p. weinberger , _ electronic structure of disordered alloys , surfaces and interfaces _( kluwer , boston - london - dordrecht , 1997 ) .vosko , l. wilk , and m. nusair , can .* 58 * , 1200 ( 1980 ) .see e.g. , v.i .anisimov , f. aryasetiawan , and a.i .lichtenstein , j. phys .: condens .matter * 9 * , 767 ( 1997 ) .sandratskii , r. singer , and e. sasiolu , , 184406 ( 2007 ) .m. pajda , j. kudrnovsk , i. turek , v. drchal , and p. bruno , phys .b * 64 * , 174402 ( 2001 ) .liechtenstein , m.i .katsnelson and v.a .gubanov , j. phys.f : met.phys .* 14 * , l125 ( 1984 ) .a. i. liechtenstein , m. i. katsnelson , v. p. antropov , v. a. gubanov , j. magn .magn . mater . *67 * , 65 ( 1987 ) .liechtenstein , m.i .katsnelson and v.a .gubanov , solid.state.commun .* 54 * , 327 ( 1985 ) .gubanov , a.i .liechtenstein , a.v .postnikov _ magnetism and the electronic structure of crystals _ , edited by m. cardona , p. fulde , k. von klitzing , h .- j .queisser ( springer , berlin , 1992 ) .i. turek , j. kudrnovsk , v. drchal , and p. bruno , philos . mag .* 86 * , 1713 ( 2006 ) .wang , r.e .prange , and v. korenman , phys .b * 25 * , 5766 ( 1982 ) .j. rusz , l. bergqvist , j. kudrnovsk , and i. turek , phys .b * 73 * , 214412 ( 2006 ) .g. fischer , m. dne , a. ernst , p. bruno , m. lders , z. szotek , w. temmerman , and w. hergert , phys .b 80 , 014408 ( 2009 ) v. heine , j.h .samson , and c.m.m .nex , j. phys .phys . * 11 * , 2645 ( 1981 ) .v. heine and j.h .samson , j. phys .* 13 * , 2155 ( 1983 ) .h. hasegawa , j. phys .. jpn . * 46 * , 1504 ( 1979 ) .pettifor , j. magn .mater * 15 - 18 * , 847 ( 1980 ) .staunton , b.l .gyorffy , a.j .pindor , g.m .stocks , and h. winter , j. phys .f * 15 * , 1387 ( 1985 ) .pindor , j. staunton , g.m .stocks , h. winter , j. phys .f * 13 * , 979 ( 1983 ) .

<S> using full - potential linear augmented plane wave method ( fp - lapw ) and the density functional theory , we have carried out a systematic investigation of the electronic , magnetic , and cohesive properties of the chalcogenide crte in three competing structures : rock - salt ( rs ) , zinc blende ( zb ) and the nias - type ( na ) hexagonal . </S><S> although the ground state is of na structure , rs and zb are interesting in that these fcc - based structures , which can possibly be grown on many semiconductor substrates , exhibit half - metallic phases above some critical values of the lattice parameter . </S><S> we find that the na structure is not half - metallic at its equilibrium volume , while both zb and rs structures are . </S><S> the rs structure is more stable than the zb , with an energy that is lower by 0.25 ev / atom . while confirming previous results on the half - metallic phase in zb structure , we provide hitherto unreported results on the half - metallic rs phase , with a gap in the minority channel and a magnetic moment of @xmath0 @xmath1 per formula unit . </S><S> a comparison of total energies for the ferromagnetic ( fm ) , non - magnetic ( nm ) , and antiferromagnetic ( afm ) configurations shows the lowest energy configuration to be fm for crte in all the three structures . </S><S> the fp - lapw calculations are supplemented by linear muffin - tin orbital ( lmto ) calculations using both local density approximation ( lda ) and lda+u method . </S><S> the exchange interactions and the curie temperatures calculated via the linear response method in zb and rs crte are compared over a wide range of the lattice parameter . </S><S> the calculated curie temperatures for the rs phase are consistently higher than those for the zb phase . </S>

0709.3854

neutral hydrogen ( hi ) absorption measurements of pulsar signals at @xmath9 cm are important probes of various properties of the interstellar medium such as the small - scale structure of of cold hi @xcite and calibrators of the pulsar distance scale and electron density models at large galactic distances @xcite .these measurements are complementary to interferometrically determined parallaxes , which can be utilized on nearer sources @xcite .we report new arecibo hi absorption observations of eight pulsars , and we use these measurements to study the electron density of the interstellar medium and the small - scale structure of neutral hydrogen clouds .sensitivity limitations indicate that few if any additional pulsar hi absorption measurements of electron density in the galactic plane at the intermediate first - quadrant galactic longitudes accessible to the arecibo telescope will be made in the next decade or so .hence it is an appropriate time to combine the new results with all previous pulsar hi measurements at these longitudes to globally assess the density in the galactic plane in this region .we will use this information to estimate the distance to the relativistic binary pulsar b1913 + 16 .in addition , hi absorption measurements on some of our high - latitude pulsars were originally measured over twenty years ago , during which interval the pulsars have moved many au through the interstellar medium . hence comparison of the old and new absorption spectra yields information on the small scale structure of the absorbing neutral hydrogen .the plan of the paper is as follows : the pulsar hi absorption observational technique is sketched in the next section ( 2 ) .we present hi absorption spectra and kinematic distance results for low - latitude pulsars in 3 . in 4, we use all measured pulsar distances in the inner galactic plane visible from arecibo observatory to analyze the electron density in that region . in 4.1 , we review the latest work on the galactocentric distance of the sun and the orbital velocity of the local standard of rest in order to select an optimum model to use in refining pulsar kinematic distances and galactic electron densities .then in 4.2 , we apply the results of 4 and 4.1 to determine the distance of the relativistic binary pulsar psr b1913 + 16 .we provide absorption spectra of high - latitude pulsars in 5 , along with analyses of time - variability of absorption in those cases where earlier epoch data are available .finally , we discuss our conclusions in 6 .all observations were made with the 305-m arecibo telescope from 1998 to 2000 .the radio frequency signals near 1420 mhz were mixed to baseband , sampled , and recorded with the caltech baseband recorder @xcite , a 10 mhz bandwidth , fast - sampling receiver backend .every 100 ns , the cbr sampled complex voltages with four - level digitization from the two orthogonally polarized feed channels , and recorded the samples to tape for later processing .( for additional details of the observing techniques and equipment , see stanimirovi ' c et al [ 2003a , b ] ) .the data were then corrected for quantization @xcite , fourier transformed and folded modulo the apparent pulsar period with the supercomputers of caltech s center for advanced computing research , resulting in data cubes consisting of intensity as a function of pulsar rotational phase ( 128 phase bins ) and radio frequency ( 2048 frequency channels , each having 1 km / s width ) .subsequent processing at carleton college collapsed the data cubes into two spectra : the pulsar-_on _ spectrum is a sum of the spectra gathered during the pulsar pulse weighted by @xmath10 , while the sum of those gathered in the interval between pulses is called the pulsar-_off _ spectrum . here@xmath11 is the broadband pulsar intensity in a given pulse phase bin @xmath12 .two final spectra are formed and displayed for each pulsar .the pulsar _ absorption _ spectrum , which represents the spectrum of the pulsar alone less any absorption caused by intervening neutral hydrogen , is the normalized difference of the pulsar-_on _ and -_off _ spectra . in order to maximize sensitivity ,multiple integrations are summed with a weight depending on the square of the pulsar signal strength @xmath13 during each integration . in some cases ,additional sensitivity was achieved by hanning smoothing the final absorption spectrum , yielding a resolution of 2 km / s .( any absorption spectrum that has been hanning smoothed is labelled as such when displayed below . ) the hi _ emission _ spectrum is the time - integrated , pulsar-_off _ spectrum , calibrated in brightness temperature units by matching its peak with the leiden / argentine / bonn hi survey @xcite .all spectra were frequency switched and low - order polynomials were fitted to and removed from them in some cases in order to flatten intrinsic or scintillation - induced bandpass ripples .the basic observing parameters are tabulated for low- and high - latitude pulsars in the early columns of tables 1 and 3 , respectively .the quantities @xmath14 and @xmath15 , the system temperature and measured 1-@xmath16 noise in optical depth away from the hi line , are both given .the former value is the sum of an estimated 40 k receiver contribution , plus a sky background determined by extrapolating the 408 mhz sky temperature @xcite at the pulsar position to 1420 mhz with a spectral index of -2.6 .the hi emission line itself contributes significantly to the system temperature ( and hence to the noise ) at those velocities where it is present . to determine the expected optical depth noise at any velocity @xmath17, one may use @xmath18/t_{sys,\rm off - line}$ ] , where @xmath19 is the measured brightness temperature of hi at that velocity .a low - latitude pulsar hi absorption spectrum can be used to set limits on the pulsar s distance , using the `` kinematic '' technique .a pulsar is farther than an hi cloud that absorbs its signal , and closer than one that does _not_. the latter `` no absorption '' criterion is impossible to ascertain in real spectra because of the inevitable presence of noise which could mask weak absorption .however it was shown by @xcite that emission features with @xmath20 k almost always exhibit significant absorption of radiation from background objects .therefore subsequent investigators have assigned an upper distance limit only at the velocity where one finds both no significant absorption , _ and _ an emission feature of @xmath20 k. we use a flat @xcite galactic rotation model and the iau galactic constants of @xmath21 kpc and @xmath22 km / s @xcite to convert the radial velocities to distance . the resulting model radial velocities are shown in a panel under each absorption / emission spectrum pair . we also add and subtract velocities of 7 km / s to our nominal velocities , in order to derive estimates of the uncertainties in distance limits due to streaming and random gas motions in the galaxy .these procedures are identical to those used in the critical evaluation of all such measurements then extant by @xcite , and by all subsequent pulsar hi absorption experimenters .hence our results are directly intercomparable with these earlier ones .our derived distance limits are discussed below for each pulsar , and are summarized in table 1 .see 4.1 for a discussion of modifications to our derived distances for different values of the galactic constants .the farthest observed absorption feature is at @xmath23 km / s , well before the tangent point .hence the pulsar lies beyond a lower distance limit of @xmath24 kpc .unfortunately the hi emission has @xmath25 k at all velocities where one might test for an upper distance limit via lack of absorption , because the @xmath26 line of sight rapidly leaves the hydrogen layer .since significant absorption could not in any case be guaranteed at the velocity of such weak emission @xcite , no upper distance limit can be set .@xcite observed this pulsar .their hi spectrum was contaminated by interstellar scintillation and was rather noisy , prompting us to reobserve it with greater sensitivity . the highest velocity absorption that @xcite could confidently detect was at @xmath27 km / s , leading to their limit of @xmath28 kpc ) @xmath29 .our new observations clearly exhibit absorption at @xmath30 km / s , near the tangent point .hence we revise the lower distance limit significantly upwards : @xmath31 kpc ) @xmath29 .the @xcite spectrum also showed an absorption feature at this velocity but it was not sufficiently above the noise to serve as a reliable kinematic distance indicator .the dip in the current absorption spectrum at @xmath32 km / s is probably noise and will not be used to set a distance limit .( it appears much more prominently in one of the two circular polarization channels than in the other , and is not visible in the @xcite absorption spectrum .furthermore , the nearby source g55.6 + 2.3=b1923 + 210 does not exhibit reliable far side absorption until @xmath33 km / s [ dickey et al 1983 ; colgan et al 1988 ] . )conversely , the _ lack _ of absorption in the @xmath34 k emission feature at @xmath35 km / s sets the upper distance limit of @xmath36 pc . while the fich et al .model predicts a tangent point velocity of 48 km / s in this direction , we detect significant emission and absorption well beyond this velocity , with the last major feature centered near @xmath37 km / s .these features are due to the sagittarius arm @xcite . @xcite and @xcite also observe hi absorption out to these velocities in the nearby sources g50.625 - 0.031 and g51.4 - 0.0 , respectively .we choose the tangent point as our lower distance limit , so @xmath38 kpc @xmath29 .the lack of absorption in the @xmath39 k emission feature at @xmath40 km / s yields the upper distance limit : @xmath41 kpcthe currently reported hi absorption distance measurements are probably among the last to be determined at arecibo in the foreseeable future , as all known pulsars having sufficient flux density to achieve adequate @xmath42 in a reasonable time have now been measured as well as is possible with this procedure .unfortunately galactic hi emission is now the dominant noise source , so that future receiver improvements will not significantly decrease the overall noise .therefore , new low - latitude pulsars discovered in the future in this longitude range will probably be too faint for the hi absorption technique , even at arecibo .consequently now is a good time to review all such measurements and their implications for the electron density in the inner galactic plane visible from arecibo .the dispersion measure @xmath43 of a pulsar , derived from multiple - frequency timing measurements and reported in @xcite , directly yields the path - integrated electron density along the line of sight : @xmath44 .hence our distance limits , @xmath45 , can be combined with the published dispersion measures to yield mean electron densities @xmath46 along the line of sight : @xmath47 table 2 lists the electron density limits derived in this fashion from our measurements and from all earlier ones in similar directions ; i.e. , toward the inner galactic plane accessible to the arecibo telescope .note that the kinematic distances listed in table 2 were all derived with the standard procedures discussed above in 3 , namely the uniform criteria established by @xcite and the iau standard galactic constants @xcite .procedures to modify these values for other choices of galactic constants are given below in 4.1 . in fig .[ fig : density]a , we display measured limits on the mean electron density as a function of galactic longitude @xmath48 in the inner galactic plane visible from arecibo . only those pulsars from table 2 with @xmath49 and a lower distance limit @xmath50 kpcare included in fig .[ fig : density]a in order to concentrate on kpc - scale averages in the galactic plane .almost all of these measurements were derived from arecibo pulsar hi absorption spectra . while there is significant scatter, it is apparent that the measured densities tend to decline as the line of sight rotates away from the inner galaxy , with a notable drop in the region between the sagittarius arm ( the first arm interior to our location ) and the local spiral arm ( see fig .[ fig : galplane ] ) . @xcitewere the first to have adequate @xmath46 measurements to infer that interarm densities are lower than those in spiral arms .as illustrated in fig .[ fig : density]a , the numerous measurements made since that epoch serve to confirm and refine their suggestion , at least in this particular interarm region .indeed , the drop in @xmath46 as @xmath48 exceeds @xmath51 corresponds with the line of sight finally reaching a longitude where it no longer intercepts the sagittarius arm ( see fig . [fig : galplane ] ) .meanwhile , with only one exception , _ every _ pulsar measurement in the interarm range @xmath2 is consistent with a relatively low @xmath52 @xmath4 ( see below for further analysis ) . the electron density at lower latitudes sampled in fig .[ fig : density]a is clearly significantly higher , with an average in the vicinity of 0.05 @xmath4 ( plus superposed variations ) in the @xmath53 longitude range .it is also useful to compare our updated set of electron density measurements with the most widely used galactic density model ( ne2001 , cordes & lazio 2002 ) . fig .[ fig : density]b exhibits the measured limits on @xmath46 , normalized by the mean electron densities _ predicted _ by the ne2001 model .the measured - to - model electron density ratio hovers near 1 ( with some inevitable scatter ) for @xmath54 , indicating that the model fits the data adequately at those longitudes .however , at the higher longitudes @xmath2 discussed above , there is a tendency for the typical measured - to - model density ratio to lie below 1 , suggesting that the ne2001 model densities are somewhat high in this interarm region . in order to further explore the electron density in this interarm region, we performed monte carlo simulations to assess the best value and uncertainty on @xmath46 and on the measured - to - model @xmath46 ratio for the pulsars in this region .of the ten pulsars with measured distance limits and limits on @xmath46 in this range ( see table 2 ) , we discarded the data from psr j1935 + 1616 since it lacks a measured upper distance and lower density limit .for each of the remaining nine pulsars , all possible distances between the measured upper and lower limits were rendered equally probable in our simulations by choosing distances randomly from a uniform distribution between the measured distance limits .note that the assumption of uniform probability is the simplest reasonable hypothesis for measurements such as these which have only a lower and upper limit .we find that @xmath55 ) ] @xmath4 , and that the ratio of measured - to - ne2001 electron density is @xmath5 for @xmath2 .future models should take this discrepancy into account . for simplicity , boththe measured and model densities discussed above are based on the standard iau galactic constants . as described below in 4.1 , the _ ratio _ plotted in fig . [ fig : density]b remains virtually identical if both constituents are rescaled to incorporate other choices of galactic constants .much progress has been made on galactic structure and kinematics in the @xmath56 years since the iau constants were defined .therefore it is useful to discuss improved measurements of @xmath0 and and @xmath1 , and to investigate the influence of these better values on the kinematic distances and electron densities discussed in 3 and 4 . from remarkable proper motion and radial velocity measurements of the orbit of star s2 about sgr a * through much of its 15-yr period , @xcite find that the distance to the galactic center @xmath57 kpc . @xcite have made equally stunning interferometric measurements of the proper motion of sgr a * , which yield an angular velocity of the local standard of rest ( lsr ) about the galactic center , @xmath58 km s@xmath59 kpc@xmath59 .both of these measurements are notable in that they rest on far fewer assumptions than earlier determinations , and hence should be freer of systematic errors . consequently we adopt them in what follows .an even newer estimate of @xmath0 from infrared measurements of red clump giants in the galactic bulge @xcite yields a value of @xmath60 ( statistical ) @xmath61 ( systematic ) kpc , which is consistent with our above adopted value .consider material traversing a circular orbit in the plane at galactocentric radius @xmath62 .given a flat rotation curve with circular velocity @xmath1 , its radial velocity with respect to the lsr is @xmath63 . with @xmath64 , the distance as a function of @xmath65is given by @xmath66 hence our new adopted galactic parameters , giving a small change in @xmath1 but a relatively large change in @xmath0 , result in a recalibration of the distance in a particular direction , accomplished by rescaling it with an approximately constant factor of @xmath67 .kinematic distance limits and galactic distance models that use the old iau galactic constants should be adjusted by this factor to reflect the improved constants .[ fig : rotcurve ] displays distance - radial velocity curves toward @xmath68 derived with the old and new flat galactic rotation curves . fig . [fig : rotcurve]a illustrates that the old and new model distances differ primarily by a constant multiplicative factor , as indicated in the previous analysis .it follows from the above , that when distance is _ normalized _ by @xmath0 , the old and new radial velocity - distance curves are almost identical ( fig .[ fig : rotcurve]b ) . such normalized distances are useful in the analysis of the acceleration of pulsars in the galactic gravitational field , as will be discussed in the next section ( 4.2 ) .as described above , the new galactic constants would result primarily in the measured distances being rescaled by the ratio of new to old @xmath0 .the electron densities , including model densities that have been calibrated via pulsar kinematic distance measurements , would then be rescaled by the inverse of this ratio ( see eq .hence fig .[ fig : density]a would be thus rescaled ; while fig .[ fig : density]b would remain unchanged since it displays the ratio of two densities , each of which should be rescaled by the same factor . the relativistic binary pulsar b1913 + 16 lies in the heart of the region of the galaxy that we are studying , at @xmath69 .the orbital decay of this system due to gravitational wave emission provides a strong test of relativistic gravitation @xcite .currently , the precision of the relativistic test is limited by uncertainties in the galactic acceleration of the pulsar and the solar system @xcite , which are currently dominated by uncertainties in the pulsar distance @xmath45 [ or more specifically , uncertainties in @xmath70 , the distance normalized by @xmath0 @xcite ] .hence it is very important to assess the current results in an effort to improve our knowledge of b1913 + 16 s distance .[ fig : density]a shows that the measured electron density appears to have a local minimum near the pulsar s longitude .the relatively low value of @xmath46 at @xmath71 is not surprising , since we noted in 4 that the line of sight in this direction traverses a long path _ between _ spiral arms ( see fig .[ fig : galplane ] ) .the tightest limits on @xmath46 in the whole inner plane visible from arecibo are for psr b1915 + 13 at @xmath72 , which happens also to be the pulsar closest to psr b1913 + 16 .hence we adopt the measured limits from b1915 + 13 , including an allowance for up to 7 km s@xmath59 non - circular velocities ( not shown in table 2 or fig .[ fig : density ] ) : @xmath73 note that this value is congruent with and has modestly tighter limits than the electon density derived in this interarm region in 4 .armed with our new estimate of @xmath46 , we use equation 1 to find the distance to psr b1913 + 16 , which has @xmath74 pc @xmath4 : @xmath75 kpc .our result can also be compared with the @xcite ne2001 model distance of @xmath76 kpc , which is lower for the reasons discussed above .@xcite used techniques similar to ours to estimate that @xmath77 . while the error bars of their and our estimates overlap significantly , we are not able to justify tightening them to the degree that @xcite did . finally , adopting what we judged in 4.1 to be the current best measurement of @xmath78 kpc ; eisenhauer et al .2005 ) leads to our best estimate of the distance to psr b1913 + 16 : @xmath79 kpc .as noted above , the uncertainty in the general relativistic orbital decay rate of this pulsar is currently dominated by the uncertainty in the quantity @xmath80 .for these purposes , it is useful to give @xmath81relatively nearby pulsars are excellent targets for studies of the tiny - scale atomic structure ( tsas , @xcite ) .comparison of their hi absorption spectra at multiple epochs as they move across the sky permits us to study structure down to au scales .we discuss here our measurements of five such pulsars .all but two have been previously observed , so we will search for changes between the previous and current epochs .the pulsars observing parameters , distances , and transverse velocities are listed in table 3 .the pulsars discussed in this section are all located at high galactic latitudes ( in all cases , @xmath82 ) . since their lines of sightleave the galactic hydrogen layer fairly quickly , they are not amenable to the kinematic distance technique discussed in 3 .conversely , we can not search for temporal absorption variations in the low - latitude pulsars of 3 because two have never before been observed , and the third s absorption spectrum was too noisy at the earlier epoch to yield meaningful differences .( indeed we reobserved it in order to more confidently measure its kinematic distance because the noise in the earlier spectrum precluded a secure result . )the hi absorption spectrum of this pulsar was first observed in 1976 - 77 by @xcite .that spectrum and our new one are plotted together in fig . [ fig : b0301 + 19 ] .both epochs spectra have a similar @xmath83 km / s resolution .both the _ integrated _ optical depth ( the equivalent width ) and the depth of most individual absorption components have changed significantly between the two epochs . although the early observations were gathered with a spectrometer that digitized the signal to only one bit , its response to rapidly varying pulsar signals was very well - characterized by @xcite .careful quantization corrections have also been made for the current epoch @xcite .hence we believe that both spectra are accurate and that the change is real .the channel spacing of the old and new spectra are 1.056 and 1.031 km / s , respectively . in order to study the absorption changes more carefully , we slightly coarsened the spacing of the new data via interpolation to match the channel spacing and center frequency of the old data .we also ensured that the old and new channel frequencies matched by verifying that the ( essentially noise - free ) _ emission _ spectra were consistent to less than a channel width .[ fig : b0301 + 19diff ] displays the old and ( resampled ) new absorption spectra ( top ) and their difference ( bottom ) .the difference spectrum also displays a @xmath84 noise envelope in order to assess the significance of variations .[ the noise envelope grows at the central velocities because the hi emission line itself contributes significantly to the system noise @xcite .] the difference spectrum clearly shows a general trend whereby the absorption line depth is greater at the later epoch in most of the central channels , with the largest single - channel difference being significant at the 2.6@xmath16 level .the @xmath85 year time baseline is unique in being one of the longest extant showing absorption variations , leading to a length scale of @xmath8 au which is second only to the psr 1557 - 50 scale of @xmath86 au @xcite .further implications of these results for small - scale structure in the interstellar medium will be analyzed in a separate paper ( stanimirovic et al , in preparation ) .@xcite also measured the hi absorption spectrum of this pulsar in 1976 - 77 .the original and current pulsar spectra show no absorption .these results are not surprising , given the galactic polar line of sight and the very weak hi emission in this direction .the pulsar moved @xmath87 au between the two observations .this pulsar is a member of a double neutron star binary system . its pulse timing parallax gives @xmath88 kpc @xcite . under the assumption that general relativity provides the correct description of gravitational wave emission , the excess observed orbital period change not attributable to gravitational wave damping yields @xmath89 kpc @xcite .there is no statistically significant absorption evident in our pulsar spectrum along this high galactic latitude line of sight .no previous hi absorption measurements have been made on this source .no statistically significant absorption is present in this high galactic latitude pulsar s spectrum .this is the first absorption spectrum from this pulsar .the broad , shallow dip in the displayed spectrum disappears in one of our two orthogonal polarizations , leading us to conclude that it is not real .this pulsar was also observed in 1976 - 77 by @xcite . in that case as well , no significant absorption was seen .the pulsar traversed @xmath90 au in the intervening time .we have determined the hi absorption spectra of eight pulsars .the three low - latitude pulsars yield kinematic distances and electron densities in the inner galactic plane visible from arecibo .these observations mark the completion of a two - decade effort to accurately measure the hi absorption spectrum of all such pulsars that are strong enough to be accessible to this technique .therefore , we have combined our new measurements with all others in this direction to study the electron density in this region .we find that the mean electron density in the plane in the interarm range @xmath2 is @xmath91 @xmath4 , which is @xmath5 of the @xcite model value currently used by most researchers . at the lower longitudes accessible to arecibo @xmath92, the @xcite model appears to conform generally to the measurements , aside from expected local variations . as part of the process, we show how to modify kinematic distances and electron densities as a function of the galactic constants @xmath0 and @xmath1 .we review recent efforts to determine the values of these constants and select the best current answers . applying all of these results to the relativistic binary pulsar b1913 + 16, we find a dispersion measure distance of @xmath93 kpc , or @xmath7 ) kpc if we adopt our current best choice for @xmath0 @xcite .the five high - latitude pulsars are most useful in multiepoch studies of small - scale structure in the interstellar medium .two were observed for the first time by us in this experiment . of the other three, two showed no measurable absorption at either of two epochs separated by twenty - two years ; while one , psr b0301 + 19 , exhibited a significant change in absorption profile over that timespan , indicating hi structure on a @xmath8 au scale .cccccccccccc psr j &psr b & @xmath94 & @xmath95&@xmath96 & _ l _ & _ b _ & dm & @xmath97 & @xmath98 & @xmath99 & @xmath100 + & & ( hr ) & ( k ) & & ( deg ) & ( deg ) & ( pc @xmath101 & ( km / s ) & ( km / s ) & ( kpc ) & ( kpc ) + j1909 + 0254 & b1907 + 02 & 3.0 & 45&0.11 & 37.6 & -2.7 & 171.7 & 60 & & @xmath24 & + j1922 + 2110 & b1920 + 21 & 2.5 & 43&0.09 & 55.3 & 2.9 & 217.1 & tp & -65 & @xmath102&@xmath103 + j1926 + 1648 & b1924 + 16 & 7.4 & 45&0.06 & 51.9 & 0.1 & 176.9 &tp & -47 & @xmath104&@xmath105 + llcrccccccc psr j & psr b & _ l _ & _ b _ & dm & @xmath99 & @xmath100 & @xmath106&@xmath107 & distance&assoc .+ & & ( deg ) & ( deg ) & ( pc @xmath101 & ( kpc ) & ( kpc ) & ( @xmath101 & ( @xmath101 & ref .+ j1848 - 0123 & b1845 - 01 & 31.34 & 0.04 & 159.5 & 4.2 & 4.8 & 0.038 & 0.033 & fw90 & + j1852 + 0031 & b1849 + 00 & 33.52 & 0.02 & 787.0 & 7.1 & 16.6 & 0.111 & 0.047 & fw90 & + j1856 + 0113 & b1853 + 01 & 34.56 & -0.50 & 96.7 & 2.7 & 3.9 & 0.036 & 0.025 & c75 & snr : w44 + j1857 + 0212 & b1855 + 02 & 35.62 & -0.39 & 506.8 & 6.9 & & 0.073 & & fw90 & + j1901 + 0331 & b1859 + 03 & 37.21 & -0.64 & 402.1 & 6.8 & 15.1 & 0.059 & 0.027 & fw90 & + j1901 + 0716 & b1859 + 07 & 40.57 & 1.06 & 252.8 & 2.8 & 4.7 & 0.090 & 0.054 & fw90 & + j1902 + 0556 & b1900 + 05 & 39.50 & 0.21 & 177.5 & 3.1 & 4.3 & 0.057 & 0.041 & fw90 & + j1902 + 0615 & b1900 + 06 & 39.81 & 0.34 & 502.9 & 6.5 & 15.8 & 0.077 & 0.032 & fw90 & + j1903 + 0135 & b1900 + 01 & 35.73 & -1.96 & 245.2 & 2.8 & 4.0 & 0.088 & 0.061 & fw90 & + j1905 + 0154a & & 36.20 & -2.20 & 194.0 & 6.4 & 8.2 & 0.030 & 0.024 & r97 & gc : ngc6749 + j1906 + 0641 & b1904 + 06 & 40.60 & -0.30 & 472.8 & 6.5 & 14.0 & 0.073 & 0.034 & fw90 & + * j1909 + 0254 * & * b1907 + 02 * & * 37.60 * & * -2.71 * & * 171.7 * & * 3.8 * & & * 0.045 * & & * this work * & + j1909 + 1102 & b1907 + 10 & 44.83 & 0.99 & 150.0 & 4.3 & 6.0 & 0.035 & 0.025 & fw90 & + j1915 + 1009 & b1913 + 10 & 44.71 & -0.65 & 241.7 & 6.0 & 14.5 & 0.040 & 0.017 & fw90 & + j1916 + 1312 & b1914 + 13 & 47.58 & 0.45 & 237.0 & 4.0 & 5.7 & 0.059 & 0.042 & fw90 & + j1917 + 1353 & b1915 + 13 & 48.26 & 0.62 & 94.5 & 4.8 & 5.7 & 0.020 & 0.017 & fw90 & + * j1922 + 2110 * & * b1920 + 21 * & * 55.28 * & * 2.93 * & * 217.1 * & * 4.8 * & * 16.2 * & * 0.045 * & * 0.013 * & * this work * & + * j1926 + 1648 * & * b1924 + 16 * & * 51.86 * & * 0.06 * & * 176.9 * & * 5.2 * & * 14.9 * & * 0.034 * & * 0.012 * & * this work * & + j1930 + 1852 & & 54.10 & 0.27 & 308.0 & 3.2 & 10 & 0.096 & 0.031 & vb88 ; las01 & snr : g54.1 + 0.3 + j1932 + 2020 & b1929 + 20 & 55.57 & 0.64 & 211.2 & 4.8 & 14.9 & 0.044 & 0.014 & fw90 & + j1932 + 2220 & b1930 + 22 & 57.36 & 1.55 & 219.2 & 10.4 & 13.7 & 0.021 & 0.016 & fw90 & + j1935 + 1616 & b1933 + 16 & 52.44 & -2.09 & 158.5 & 5.2 & & 0.030 & & fw90 & + j1939 + 2134 & b1937 + 21 & 57.51 & -0.29 & 71.0 & 4.6 & 14.8 & 0.015 & 0.005 & fw90 & + j1952 + 3252 & b1951 + 32 & 68.77 & 2.82 & 45.0 & 1.0 & 4.0 & 0.045 & 0.011 & b84 & snr : ctb80 + j2004 + 3137 & b2002 + 31 & 69.01 & 0.02 & 234.8 & 7.0 & 12.0 & 0.034 & 0.020 & fw90 & + ccccccccccc psr j & psr b & @xmath94 & @xmath95&@xmath96 & dm & @xmath45 & @xmath108 & prop .& @xmath109 & @xmath109 + & & ( hr ) & ( k ) & & ( pc @xmath101 & ( kpc ) & ( mas / yr ) & ref . &( au / yr ) & ( km / s ) + j0304 + 1932 & b0301 + 19 & 1.3 & 41&0.015&15.737 & 0.62 & 37 ( 5 ) & las82 & 23 & 110 + j1239 + 2453 & b1237 + 25 & 1.4 & 41&0.009 & 9.242 & 0.85 & 115.6 ( 12 ) & b03 & 98 & 470 + j1537 + 1155 & b1534 + 12 & 1.7 & 42&0.13 & 11.61436 & 1.02 & 25.086 ( 20 ) & kws03 & 26 & 121 + j1543 + 0929 & b1541 + 09 & 2.9 & 42&0.03 & 35.24 & 3.49 & 8.3 ( 10 ) & b03 & 29 & 140 + j2305 + 3100 & b2303 + 30 & 1.2 & 41&0.07 & 49.544 & 3.66 & 20 ( 2 ) & b03 & 73 & 350 +

<S> we have used the arecibo telescope to measure the hi absorption spectra of eight pulsars . </S><S> we show how kinematic distance measurements depend upon the values of the galactic constants @xmath0 and @xmath1 , and we select our preferred current values from the literature . </S><S> we then derive kinematic distances for the low - latitude pulsars in our sample and electron densities along their lines of sight . </S><S> we combine these measurements with all others in the inner galactic plane visible from arecibo to study the electron density in this region . </S><S> the electron density in the interarm range @xmath2 is @xmath3 ) ] @xmath4 . </S><S> this is @xmath5 of the value calculated by the @xcite galactic electron density model . </S><S> the model agrees more closely with electron density measurements toward arecibo pulsars lying closer to the galactic center , at @xmath6 . </S><S> our analysis leads to the best current estimate of the distance of the relativistic binary pulsar b1913 + 16 : @xmath7 ) kpc . </S><S> we use the high - latitude pulsars to search for small - scale structure in the interstellar hydrogen observed in absorption over multiple epochs . </S><S> psr b0301 + 19 exhibited significant changes in its absorption spectrum over 22 yr , indicating hi structure on a @xmath8 au scale . </S>

hep-th0205062

[ 3intro ] recently a lot of work on quantum einstein gravity ( qeg ) went into constructing an appropriate exact renormalization group ( rg ) equation @xcite , finding approximate solutions to it @xcite , and exploring their implications for black hole physics @xcite and cosmology @xcite .in particular , strong indications were found that qeg might be nonperturbatively renormalizable .if so , it could have the status of a fundamental , microscopic quantum theory of gravity .the basic tool used in these investigations is the effective average action and its exact rg equation @xcite .it is a continuum analog of wilson s lattice renormalization group of iterated block spin transformations @xcite .both in quantum field theory and statistical mechanics the idea is to integrate out all fluctuation modes which have momenta larger than a certain infrared ( ir ) cutoff @xmath2 ( `` fast degrees of freedom '' ) , and to take account of those modes in an implicit way by the modified dynamics which they induce for the remaining fluctuations with momenta smaller than @xmath2 ( the `` slow degrees of freedom '' ) . in field theory this `` renormalized '' dynamics is encoded in a scale dependent effective action , @xmath3 , whose dependence on the cutoff scale @xmath2 is governed by a functional differential equation referred to as the `` exact rg equation '' @xcite .this equation gives rise to a flow on the space of all actions ( `` theory space '' ) .the functional @xmath3 defines an effective field theory valid near the scale @xmath2 ; evaluated at tree level , it describes all loop effects due to the high - momentum modes .the effective average action can be thought of as a kind of microscope with a variable resolution . at large @xmath2 , the physics at short distances@xmath4 can be read off directly from @xmath3 ; at small @xmath2 we see a coarse - grained picture suitable for a simple description of structures with a large characteristic length scale @xmath4 @xcite .the effective average action @xmath3 , regarded as a function of @xmath2 , interpolates between the ordinary effective action @xmath5 and the bare ( classical ) action @xmath6 which is approached for @xmath7 .the construction of @xmath3 begins by adding a ir cutoff term @xmath8 to the classical action entering the standard euclidean functional integral for the generating functional @xmath9 of the connected green s functions .the new piece @xmath8 introduces a momentum dependent @xmath10-term @xmath11 for each mode of the quantum field with momentum @xmath12 . for @xmath13 , the cutoff function @xmath11is assumed to vanish so that the high - momentum modes get integrated out unsuppressed . for @xmath14, it behaves as @xmath15 ; hence the small - momentum modes are suppressed in the path integral by a mass term @xmath16 .apart from a correction term which is known explicitly @xcite , the effective average action @xmath3 is given by the legendre transform of the modified generating functional @xmath17 . from this definitionone can derive the exact rg equation obeyed by @xmath3 . in a slightly symbolic notation it is of the form @xmath18\;.\end{aligned}\ ] ]the rhs of this equation is a kind of `` beta - functional '' which summarizes the beta - functions for infinitely many running couplings .geometrically , it defines a vector field on theory space , the corresponding flow lines being the rg trajectories @xmath19 .the functional @xmath3 enters this vector field via its hessian @xmath20 , i.e. the infinite - dimensional matrix of all second functional derivatives of @xmath3 with respect to the dynamical , i.e. , non - background fields . in eq .( [ 3in1 ] ) the c - number argument of @xmath21 is replaced with the operator @xmath22 .the discrimination of high-``momentum '' vs. low-``momentum '' modes is performed according to the spectrum of this operator , i.e. @xmath23 is an eigenvalue of @xmath22 . in simple theories where no gauge or diffeomorphism invariance needs to be respected , @xmath24 is the free laplacian , @xmath25 , whose eigenmodes are momentum eigenstates in the usual sense of the word . in yang - millstheory @xcite it has proven convenient to use the background field formalism @xcite and to set @xmath26 where @xmath27 is the covariant derivative in the background field . the background field technique plays a dual role in this context .using a background gauge fixing term makes @xmath3 a gauge invariant functional of its argument , and using @xmath27 in the cutoff leads to a flow equation of the relatively simple type ( [ 3in1 ] ) , similar to non - gauge theories .( the rg equation resulting from @xmath28 with @xmath29 constructed from the _ dynamical _ gauge field is quite unwieldy . ) along similar lines an effective average action for @xmath30-dimensional euclidean qeg has been constructed in ref .@xcite , and the corresponding flow equation has been derived . leaving the faddeev - popov ghosts aside for a moment ,the gravitational effective average action , @xmath31 $ ] , is a functional of two different metrics , the `` ordinary '' dynamical metric @xmath32 and the background metric @xmath33 .the usual effective action @xmath34 $ ] is recovered by taking the limit @xmath35 of the functional @xmath36\equiv\gamma_k[g_{\mu\nu},\bar{g}_{\mu\nu}=g_{\mu\nu}]$ ] in which the two metrics are taken equal .thanks to the background gauge fixing condition this construction leads to a functional @xmath36 $ ] which is invariant under general coordinate transformations .one of the many advantages which the exact rg approach has in comparison to the standard canonical or path integral quantization is that it offers a very natural and intuitive nonperturbative approximation scheme . by truncating the theory spaceone can obtain approximate solutions to the rg equation which do not need a small expansion parameter .the idea is to project the rg flow from the `` huge '' infinite - dimensional space of all actions onto some smaller , typically finite dimensional subspace which is easier to handle . in this waythe functional rg equation for @xmath3 becomes a system of ordinary differential equations for a ( finite ) set of coupling constants which have the geometrical interpretation of coordinates on the subspace .it is clear that in applying this strategy the key problem is finding the `` relevant '' subspace which contains the essential physics . in a first attempt at solving the gravitational rg equation @xcitethe flow has been projected onto the 2-dimensional subspace of theory space which is spanned by the invariants @xmath37 and @xmath38 .this is the so - called einstein - hilbert truncation defined by the ansatz @xmath39=\left(16\pig_k\right)^{-1}\int d^dx\,\sqrt{g}\left\ { -r(g)+2\bar{\lambda}_k\right\}+\mbox{classical gauge fixing}.\end{aligned}\ ] ] the two running couplings involved are the running newton constant @xmath40 and the running cosmological constant @xmath41 .the similarity of ( [ 3in2 ] ) to the action of classical general relativity is accidental in a sense ; improved truncations would include both higher powers of the curvature and nonlocal terms @xcite . in ( [ 3in2 ] )only the gauge fixing term depends on @xmath33 ; it vanishes when we set @xmath42 .the scale dependence of @xmath40 and @xmath41 is most conveniently visualized as a flow in the @xmath43-@xmath44plane where @xmath45 and @xmath46 are the dimensionless newton constant and cosmological constant , respectively . using the original cutoff of `` type a '' @xcite the system of equations for @xmath47 and @xmath48was derived in @xcite and solved numerically in @xcite . in @xcite a new cutoff of `` type b '' was introduced and the corresponding flow equations in the einstein - hilbert truncation were derived .( the `` type b '' cutoff is convenient if one uses the tt - decomposition of the metric @xcite . )the fixed point properties of these equations were first discussed in @xcite and @xcite , and analyzed in detail in @xcite and @xcite .one finds that the rg flow in the @xmath43-@xmath44plane is governed by two fixed points @xmath49 : a trivial or `` gaussian '' fixed point at @xmath50 , and a non - gaussian fixed point with @xmath51 and @xmath52 . in order to appreciate the importance of the non - gaussian fixed point we recall what it means to `` quantize '' a theory in the average action approach .one picks a bare action @xmath6 and imposes the initial condition @xmath53 at the ultraviolet ( uv ) cutoff scale @xmath54 , uses the rg equation to find @xmath3 at all lower scales @xmath55 , and finally sends @xmath35 and @xmath56 .a _ fundamental _ theory has the property that the `` continuum '' limit @xmath56 actually exists after redefining only finitely many parameters in the action .this is the case in perturbatively renormalizable theories @xcite , but there are also examples of perturbatively nonrenormalizable theories which possess a limit @xmath56 @xcite .the continuum limit of those `` nonperturbatively renormalizable '' theories is taken at a non - gaussian fixed point , i.e. the theory is defined by the set of rg trajectories which leave the fixed point when we lower @xmath2 .these trajectories span the uv critical hypersurface of the fixed point , @xmath57 .if it is finite dimensional , the quantum theory thus constructed has only finitely many free parameters and therefore keeps its predictive power even at arbitrarily large momentum scales .this behavior is to be contrasted with an _ effective _ field theory which , at high energies , typically contains an increasing number of free parameters which must be taken from the experiment . in his `` asymptotic safety '' scenario weinberg@xcite conjectured that a fundamental quantum field theory of gravity could perhaps be constructed nonperturbatively by taking the continuum limit at a non - gaussian fixed point @xcite .while originally this idea could be implemented in @xmath58 dimensions only , the recent results coming from the effective average action strongly support the hypothesis that this fixed point exists also in 4 dimensions . within the einstein - hilbert truncation ,the existence of a suitable non - gaussian fixed point is definitely established by now ; the crucial question is whether it is the projection of a fixed point present in the exact theory or merely an artifact of the approximation .let us assume for a moment that the fixed point indeed exists in the exact 4-dimensional theory and that we define qeg by taking the @xmath56 limit there .then , since @xmath59 approaches @xmath60 for @xmath7 , the dimensionful couplings behave as @xmath61 for large @xmath2 .obviously @xmath40 vanishes for @xmath7 .at least as far as this coupling is concerned qeg is asymptotically free similar to yang - mills theory .the @xmath62-dependence of the running newton constant will lead to a characteristic momentum dependence of the cross sections for graviton - graviton scattering and graviton mediated matter - matter scattering . because of this characteristic momentum dependence , qeg could be distinguished experimentally from alternative theories of quantum gravity such as string theory , at least in principle .let us generalize the standard definition of the planck mass , @xmath63 , and introduce the _ running planck mass _ @xmath64 at the laboratory scale @xmath65 reduces to @xmath66 , most probably , and its dependence on @xmath2 is negligible .however , in the fixed point regime @xmath7 , the asymptotic freedom of @xmath40 implies that @xmath65 is proportional to the scale @xmath2 itself : @xmath67 .this shows that the running planck mass is a rather elusive `` barrier '' which never can be jumped across in any experiment .if we analyze a system with a probe of increasing momentum @xmath2 we will always push the running planck mass ahead of us and never reach it . also the standard constant planck mass ,defined more precisely in terms of the ir value of @xmath40 , @xmath68^{-1/2}\;,\end{aligned}\ ] ] plays an important role in qeg , similar to that of @xmath69 in qcd . according to the numerical solutions of the @xmath43-@xmath70-system @xcite , @xmath66 marks the lower boundary of the asymptotic scaling region .near @xmath71 there is a crossover from the scaling laws ( [ 3in3 ] ) of the non - gaussian fixed point to those of the gaussian fixed point . according to the uv scaling laws ( [ 3in3 ] )the dimensionful cosmological constant diverges for @xmath7 proportional to @xmath72 .this has an interesting geometrical interpretation .let us consider the @xmath2-dependent , effective field equations implied by the truncation ansatz ( [ 3in2 ] ) with @xmath42 for @xmath73 .they happen to coincide with the familiar vacuum einstein equations with the cosmological constant replaced by the scale - dependent quantity @xmath41 : @xmath74 since @xmath41 is the only quantity which sets a scale , every solution to ( [ 3in6 ] ) has a typical radius of curvature @xmath75 .( for instance , the maximally symmetric @xmath76-solution has the radius @xmath77 . ) the @xmath2-dependence of the solutions and in particular of @xmath78 should be interpreted as follows .if we want to explore the spacetime structure at a fixed length scale @xmath79 it is most convenient to use the action @xmath36 $ ] at @xmath80 because for this , and only this , functional a _ tree level _analysis is sufficient to describe the essential physics at this scale , including all quantum effects .hence , when we observe spacetime with a `` microscope '' of resolution @xmath81 , we will see an average radius of curvature given by @xmath82 .once @xmath81 is smaller than the ( standard ) planck length @xmath83 we are in the fixed point regime ( [ 3in3 ] ) so that @xmath84 , or @xmath85 thus , when we look at the structure of spacetime with a microscope of resolution @xmath81 , the average radius of curvature which we measure is proportional to the resolution itself. if we want to probe finer details and decrease @xmath81 we automatically decrease @xmath78 and hence _ in_crease the average curvature .spacetime seems to be more strongly curved at small distances than at larger ones .the scale - free relation ( [ 3in7 ] ) suggests that at distances below the planck length quantum spacetime is a kind of fractal with a self - similar structure .it has no intrinsic scale . before we continuea remark might be in order on which theory precisely we refer to as `` quantum einstein gravity '' or `` qeg '' .while flow equations can also be used in the effective field theory approach to quantum gravity @xcite , in the present context `` qeg '' stands for the fundamental theory whose continuum limit @xmath56 is taken at the non - gaussian fixed point .this theory has @xmath86 free parameters , and fixing these parameters amounts to picking a specific trajectory @xmath19 in the full theory space .for @xmath7 this trajectory hits the fixed point action @xmath87 , regarded as the collection of its infinitely many dimensionless coordinates on theory space .the fixed point action @xmath87 corresponds to the `` bare '' or `` classical '' action in conventional field theory .however , unlike the latter @xmath87 is not put in by hand but is rather _ derived _ with the help of the rg equation .the usual canonical or path integral quantization is always based upon a `` prejudice '' about what the classical action is . in the asymptotic safety scenario the `` classical '' action is fixed instead by the condition of nonperturbative renormalizability ; it can not be guessed by simple power - counting , symmetry , or invariance arguments , but the effective average action provides a computational framework to determine it .the really crucial property which defines qeg is the requirement of _ diffeomorphism invariance_. before we can write down a flow equation we must declare what the theory space is on which the renormalization group is supposed to operate . in the case of qeg it is defined to be the space of functionals @xmath34 $ ] depending on a nondegenerate , symmetric rank-2 tensor field in a diffeomorphism invariant way . leaving technical details aside , the theory space then fixes the flow equation which in turn determines the rg trajectories and the fixed points .so far our discussion referred to the exact theory on the full theory space .if we project on the subspace spanned by the einstein - hilbert truncation it is clear that also the fixed point action @xmath87 must be of the einstein - hilbert type .however , we emphasize that this is a trivial consequence of the simple truncation we have chosen , and we have no reason to believe that the exact @xmath87 is of the einstein - hilbert type , too .in fact , within the more complicated truncation of the present paper @xmath87 receives corrections which go beyond the einstein - hilbert form .hence `` quantum einstein gravity '' does _ not _ mean that the einstein - hilbert action is the bare action to be quantized . in this respectour approach is different from canonical quantum gravity , along the lines of ashtekar s program @xcite , for instance .( it is intriguing that also in this context remarkable finiteness properties have been proven recently @xcite . )clearly it is a highly attractive idea that there could be a nonperturbatively renormalizable field theory of the metric field so that there is no longer any conceptual need for leaving the framework of quantum field theory in order to arrive at a consistent microscopic theory of quantum gravity .therefore every effort should be made to show that the non - gaussian fixed point found in the einstein - hilbert truncation is not just an artifact of this approximation . in ref .@xcite we therefore started an extensive analysis of the reliability of the einstein - hilbert truncation near the fixed point . therethe strategy was to use the scheme dependence of universal quantities in order to get a first idea about the precision which can be achieved with this truncation . here `` scheme dependence '' refers to the dependence on the details of the cutoff procedure , i.e. on the shape of the function @xmath11 . by definition ,universal quantities are exactly scheme independent in the exact theory , but they might acquire some scheme dependence once we make approximations . the level of this residual scheme dependence can serve as a measure for the quality of the approximation .typical universal quantities are the critical exponents of fixed points and , as we argued , the product @xmath88 .the upshot of our analysis was that the einstein - hilbert truncation seems to provide a description that is much more reliable and precise than originally hoped for , and that it would be very hard to understand the approximate scheme independence we found if the fixed point was just due to a misleading approximation .these results are certainly very encouraging , but it is clear that the ultimate justification of a truncation ansatz consists of adding further terms to it and verifying that its predictions do not change much . in the present paperwe take a first step in this direction and add one further invariant constructed from @xmath32 to the ansatz . which invariant should we take ?in standard renormalized perturbation theory where ( at least implicitly ) the @xmath56 limit is taken at the gaussian fixed point , the relative importance or `` relevance '' of the various field monomials is measured by their scaling dimensions at the gaussian fixed point , i.e. by their canonical dimensions simply . since at the non - gaussian fixed pointthe anomalous dimensions are large we have no similarly simple guide line at our disposal , and a priori all invariants are equally plausible . to get a first idea about what happens away from the einstein - hilbert subspace we shall include the higher - derivative invariant @xmath89 and study the rg flow in the `` @xmath1-truncation '' , @xmath90 , @xmath91 . ]@xmath92=\int d^dx\,\sqrt{g}\left\{\left(16\pi g_k\right)^{-1 } \left[-r(g)+2\bar{\lambda}_k\right]+\bar{\beta}_k\,r^2(g)\right\ } + \mbox{classical gauge fixing}\end{aligned}\ ] ] its truncation subspace is 3-dimensional , with coordinates @xmath93 , @xmath94 and the new coupling @xmath95 .it is well known that beyond @xmath89 there exist two where one invariant can be eliminated by virtue of the gauss - bonnet identity . ]more @xmath96-invariants : @xmath97 and @xmath98 . in a standard perturbative calculation nearthe gaussian fixed point consistency would require to include them along with the @xmath1-term because they all have the same canonical dimension .as for the non - gaussian fixed point , we have no a priori information from general principles about the relative importance of the three terms . since anyhow the best we can do is to take a `` step into the dark '' , without knowing whether we walk in the most `` relevant '' direction , we shall omit the other two invariants here . including themwould go far beyond the present calculational possibilities , in particular since it would require a much more complicated projection technique @xcite .in this paper we shall derive the ( extremely complicated ) 3-dimensional rg equations of the @xmath1-truncation , and we shall use them in order to investigate the fixed points of the flow .our main results will be the following : a ) the gaussian fixed point of the einstein - hilbert truncation does _ not _ generalize to a fixed point of the @xmath1-truncation .b ) the non - gaussian fixed point does indeed generalize to a fixed point of the @xmath1-truncation , and the @xmath43-@xmath44projection of this fixed point is described almost perfectly by the einstein - hilbert truncation . within the ( weak ) residual scheme dependence ,the fixed point properties are almost insensitive to the inclusion of the @xmath1-invariant .this is further strong evidence against the theoretical possibility that the non - gaussian fixed point is a truncation artifact . in the second part of the paper we shall address a very important general problem which is of a more technical nature .it is related to a notorious disease of standard euclidean quantum gravity : the conformal factor problem . in setting up the truncated rg equation the cutoff function @xmath21 ( actually a matrix in field space )is adapted to the truncation in such a way that , for @xmath14 , the inverse propagator of every massless mode , @xmath23 , is replaced by @xmath99 .a problem arises if there are modes , such as those of the conformal factor in the einstein - hilbert truncation , which have a negative kinetic energy , i.e. their inverse propagator is @xmath100 . in @xciteit has been argued that for these modes also the sign of @xmath21 should be reversed so as to obtain the regularized inverse propagator @xmath101 .while there is little doubt that this procedure is correct for the einstein - hilbert truncation , it leads to the seemingly paradoxical situation that in the euclidean path integral the modes of the conformal factor are enhanced rather than suppressed in the ir @xcite .contrary to the einstein - hilbert truncation , the @xmath1-truncation yields a functional @xmath36 $ ] which is bounded below and , as we shall see , gives positive kinetic energy to _ all _ modes , provided one stays close to the uv fixed point .hence our investigation of the non - gaussian fixed point is not plagued by the conformal factor problem , and the construction of the cutoff becomes straightforward .this advantage is an independent motivation for studying the @xmath1-truncation . in the usual perturbative approach higher - derivative theories of ( lorentzian ) gravityare notoriously problematic as far as causality and unitarity are concerned @xcite . while in @xmath73 the most general @xmath96-theory , when expanded about flat space , is power - counting renormalizable it suffers from excitations with , classically , negative linearized energy and , quantum mechanically , a wrong - sign residue of the propagator leading to a state space containing negative - norm states @xcite .these `` ghosts '' have masses of the order of the planck mass .correspondingly , if a truncation ansatz is of the @xmath96-type the @xmath2-dependent effective propagator @xmath102 , evaluated for flat space , has similar ghosts with masses @xmath103 @xmath65 . by itselfthis does not indicate any real problem because generically flat space anyhow is not a solution of the effective equation of motion .compared to the perturbative quantization the potential problem of ghost excitations manifests itself in the euclidean rg approach in a conceptually different , more tractable manner . herethe linearization is performed about the backgrounds needed for the projection procedure , not about flat space . for a well - defined computation of the rg trajectories on a certain @xmath2-intervalit is sufficient that the ( truncated ) @xmath3 gives positive linearized action to all modes contributing to the rg running in this interval . in our calculationthis will indeed be the case for @xmath2 large enough where the truncation is believed to be reliable @xcite .the much more subtle issues related to a lorentzian interpretation of the theory and its causality properties can be addressed only once a complete trajectory , valid down to the ir , and in particular the precise form of the fixed point action is known . from allwhat we can tell now the exact qeg could very well be `` causal '' in an appropriate sense .a brief summary of some of the results derived in the present paper appeared in @xcite , and an informal introduction to the older work can be found in @xcite . in the present paper we focus on pure gravity .the gravitational average action with matter fields included was discussed in @xcite and @xcite .the gauge fixing dependence of the original formulation @xcite was investigated in @xcite and @xcite .an incomplete higher - derivative calculation was begun in @xcite where the running of the @xmath1-couplings was neglected , however , and no conclusions about the fixed point could be drawn .the remaining sections of this paper are organized as follows . in section [ 3s2 ]we review some general properties of the exact rg equation which will be needed later on .section [ 3s3 ] is devoted to the construction of cutoffs which are adapted to a specific truncation ; in particular the complications due to the conformal factor problem will be discussed there . in section [ @xmath1-trunc ]we derive the system of rg equations which results from the @xmath1-truncation , and in section [ 3s5 ] we analyze its fixed point structure .we discuss the fate of the gaussian fixed point which is present only in the einstein - hilbert truncation , and we reanalyze the non - gaussian fixed point in the more general setting . in section [ 3s6 ] the positivity properties of the truncated action functional , its hessian , andthe cutoff operator are investigated ; in particular we show that our analysis of the non - gaussian fixed point is not affected by the conformal factor problem .the conclusions are contained in section [ conclusio ] .many important technical results , including the coefficients occuring in the rather complicated beta - functions of the @xmath43-@xmath70-@xmath104system , are tabulated in various appendices .[ 3s2 ] in this section we briefly review the construction of the `` type b '' rg equation for quantum gravity performed in ref .@xcite to which we refer for the details .we start from a scale - dependent modification of the generating functional for the connected green s functions , @xmath17 .it is defined by the following euclidean functional integral : @xmath105\right\}=\int{\cal d}h_{\mu\nu}\ , { \cal d}c^\mu\,{\cal d}\bar{c}_\mu\,\exp\left[-s[\bar{g}+h]\right . }\nonumber\\ & & \left .- s_{\rm gf}[h;\bar{g}]-s_{\rm gh } [ h , c,\bar{c};\bar{g}]-\delta_k s[h , c,\bar{c};\bar{g}]-s_{\rm source}\right]\;.\end{aligned}\ ] ] in ( [ a ] ) we use the background gauge fixing technique which necessitates the decomposition of the full quantum metric @xmath106 into a fixed background metric @xmath33 and a fluctuation variable @xmath107 : @xmath108 .it allows us to replace the integration over @xmath106 by an integration over @xmath107 .furthermore , @xmath109 and @xmath110 are the faddeev - popov ghosts of the gravitational field .the first term of the action , @xmath111 $ ] , is the classical part , which is assumed to be invariant under general coordinate transformations . for the time being , we also assume that it is positive definite , @xmath112 . the gauge fixing term is given by @xmath113=\frac{1}{2\alpha}\int d^dx\,\sqrt{\bar{g}}\ , \bar{g}^{\mu\nu}\,f_\mu[\bar{g},h]\,f_\nu[\bar{g},h]\;.\end{aligned}\ ] ] in the present paper we use the linear gauge condition @xmath114=\sqrt{2}\kappa\,{\cal f}^{\alpha\beta}_\mu[\bar{g}]\ , h_{\alpha\beta}$ ] with @xmath115=\delta^\beta_\mu\bar{g}^{\alpha\gamma } \bar{d}_\gamma-\frac{1}{2}\bar{g}^{\alpha\beta}\bar{d}_\mu$ ] , which amounts to a background version of the harmonic coordinate condition .here we introduced the constant @xmath116 where @xmath117 is the bare newton constant .moreover , @xmath27 denotes the covariant derivative constructed from the background metric @xmath33 , while we shall write @xmath29 for the covariant derivative involving the complete metric @xmath106 .@xmath118 is the faddeev - popov ghost action resulting from the above gauge fixing .furthermore , @xmath8 and @xmath119 are the cutoff and the source action , respectively .@xmath8 provides an appropriate infrared cutoff for the integration variables and will be discussed in detail in a moment ; @xmath119 introduces sources for the fields @xmath107 , @xmath110 and @xmath109 .next we decompose the gravitational field @xmath107 according to ( see e.g. @xcite ) @xmath120 in order to obtain this `` tt - decomposition '' one starts by writing @xmath107 as a sum of its orthogonal parts : @xmath121 .here @xmath122 , @xmath123 and @xmath124 represent the transverse traceless , longitudinal traceless and pure trace part , respectively . introducing two scalar fields @xmath125 and@xmath126 , and a transverse vector field @xmath127 , the tensors @xmath124 and @xmath123 can be expressed by @xmath128 and @xmath129 with @xmath130 and @xmath131 .thereby we end up with eq .( [ tt ] ) . in the following the components of @xmath107thus introduced will be referred to as the `` component fields '' .they obey the relations @xmath132 obviously the complete field @xmath107 receives no contribution from those @xmath127- and @xmath126-modes which satisfy the killing equation @xmath133 and the scalar equation @xmath134 respectively .such `` unphysical '' @xmath127- and @xmath126-modes have to be excluded from the functional integral and all subsequent calculations @xcite .having a closer look at the scalar equation ( [ 010 ] ) , one recognizes that there is a one - to - one correspondence between the nonconstant solutions of ( [ 010 ] ) and the purely longitudinal , or proper , conformal killing vectors ( pckv ) @xmath135 .they are related via @xmath136 .likewise we decompose the ghost and the antighost into their orthogonal components : @xmath137 here @xmath138 and @xmath139 are the transverse components of @xmath109 and @xmath110 : @xmath140 , @xmath141 . furthermore , the scalars @xmath142 and @xmath143 parametrize the longitudinal part of @xmath109 and @xmath110 , respectively .the constant @xmath142- and @xmath143-modes represent unphysical modes which have to be excluded . for calculational conveniencewe now introduce new variables @xmath144 replacing @xmath145 , by means of the momentum dependent ( nonlocal ) redefinitions @xmath146 here the operator @xmath147 maps vectors onto vectors according to @xmath148 . in accordance with the decompositions ( [ tt ] ) , ( [ t ] ) and the redefinitions ( [ y ] ) we then perform the combined transformation of integration variables @xmath149 , @xmath150 , @xmath151 in the functional integral ( [ a ] ) .the jacobian induced by this change of variables is such that it boils down to an umimportant constant if einstein backgrounds , characterized by @xmath152 with @xmath153 a constant , are inserted into ( [ a ] ) .( for the general case see ref .@xcite . )let us now come to the `` type b '' cutoff term @xmath154 .at the component field level , it is a sum of inner products , @xmath155 = \frac{1}{2}\sum\limits_{\zeta_1,\zeta_2\in i_1 } \left<\zeta_1,\left({\cal r}_k\right)_{\zeta_1\zeta_2}\ , \zeta_2\right>+\frac{1}{2 } \sum\limits_{\psi_1,\psi_2\in i_2 } \left<\psi_1,\left({\cal r}_k\right)_{\psi_1\psi_2}\psi_2\right>\end{aligned}\ ] ] with the index sets @xmath156 , @xmath157 . at this stage of the discussion it is not necessary to specify the explicit structure of the cutoff operators @xmath21 acting on the component fields . in order to provide the desired suppression of low - momentum modes, these operators must vanish for @xmath158 ( in particular for @xmath35 ) and must behave as @xmath159 for @xmath160 .( the meaning of the constant @xmath161 will be explained later . )furthermore , they have to satisfy certain hermiticity conditions @xcite .now we are in a position to construct the effective average action @xmath3 .it is defined as the difference between the legendre transform of @xmath17 at fixed @xmath33 , denoted @xmath162 $ ] , and the cutoff action with the classical fields inserted @xcite : @xmath163 \equiv\widetilde{\gamma}_k\left[g-\bar{g},v,\bar{v};\bar{g}\right ] -\delta_k s[g-\bar{g},v,\bar{v};\bar{g}]\;.\end{aligned}\ ] ] here the classical fields represent the ( @xmath2-dependent ) expectation values of the quantum fluctuations : @xmath164 , @xmath165 , @xmath166 .they are obtained in the usual way as functional derivatives of @xmath17 with respect to the sources . in ( [ nn ] ) we expressed @xmath167 in terms of the classical counterpart @xmath32 of the quantum metric @xmath168 which , by definition , is given by @xmath169 .the classical analogs of the components @xmath170 will be denoted @xmath171 .the exact rg equation for the effective average action describes the change of @xmath3 induced by an infinitesimal change of the scale @xmath2 . introducing the rg `` time '' @xmath172 ,it can be derived from the @xmath173-derivative of the functional integral ( [ a ] ) .it takes the form @xmath174 & = & \frac{1}{2}{\rm tr}'\left[\sum\limits_{\zeta_1,\zeta_2\in \bar{i}_1 } \left(\gamma_k^{(2)}[g,\bar{g},v,\bar{v } ] + { \cal r}_k\right)^{-1}_{\zeta_1\zeta_2 } \partial_t\left({\cal r}_k\right)_{\zeta_2\zeta_1}\right ] \nonumber\\ & & + \frac{1}{2}{\rm tr}'\left[\sum\limits_{\psi_1,\psi_2\in \bar{i}_2 } \left(\gamma_k^{(2)}[g,\bar{g},v,\bar{v } ] + { \cal r}_k\right)^{-1}_{\psi_1\psi_2 } \partial_t\left({\cal r}_k\right)_{\psi_2\psi_1 } \right]\;.\end{aligned}\ ] ] here @xmath20 denotes the hessian of @xmath3 with respect to the component fields .furthermore , we wrote @xmath175 , @xmath176 and introduced the index sets @xmath177 .furthermore , the primes at the traces indicate that all unphysical @xmath178- and @xmath179-modes , characterized by eqs .( [ 09 ] ) and ( [ 010 ] ) , are to be excluded from the calculation of the traces .[ 3s3 ] in concrete applications of the exact rg equation one encounters the problem of dealing with an infinite system of coupled differential equations .usually it is impossible to find an exact solution so that we are forced to rely upon approximations .a powerful nonperturbative approximation scheme is the truncation of theory space , which means that only a finite number of couplings is considered and the rg flow is projected onto a finite - dimensional subspace of theory space . in practice one proceeds as follows .one makes an ansatz for @xmath3 that comprises only a few couplings and inserts it on both sides of eq .( [ oo ] ) . by projecting the rhs of this equation onto the space of operators appearing onthe lhs one obtains a finite set of coupled differential equations for the couplings taken into account .given an arbitrary truncation it is not clear a priori whether it is sensible and leads to at least approximately correct results . in this respect the modified brs ward identities satisfied by the exact @xmath3 @xcite are of special importance , since only those truncations which are ( approximately ) consistent with them can be reliable . in @xciteit was shown that under certain conditions truncations of the form @xmath180=\bar{\gamma}_k[g]+\widehat{\gamma}_k[g,\bar{g } ] + s_{\rm gf}[g-\bar{g};\bar{g}]+s_{\rm gh}[g-\bar{g},v,\bar{v};\bar{g}]\end{aligned}\ ] ] which neglect the rg running in the ghost sector are approximate solutions to the ward identities for the exact @xmath3 . here@xmath181 $ ] is defined as @xmath182\equiv\gamma_k[g , g,0,0]\end{aligned}\ ] ] and @xmath183 $ ] encodes the quantum corrections of the gauge fixing term .( for the details we refer to @xcite . ) inserting the ansatz ( [ truncclass ] ) into the exact evolution equation ( [ oo ] ) leads to a truncated rg equation which describes the rg flow of @xmath3 in the subspace of action functionals spanned by ( [ truncclass ] ) .the equation governing the evolution of the purely gravitational action @xmath184\equiv\gamma_k[g,\bar{g},0,0]=\bar{\gamma}_k[g]+s_{\rm gf } [ g-\bar{g};\bar{g}]+\widehat{\gamma}_k[g,\bar{g}]\end{aligned}\ ] ] takes the form @xmath185 & = & \frac{1}{2}{\rm tr}'\left[\sum\limits_{\zeta_1,\zeta_2\in\bar{i}_1 } \left(\gamma_k^{(2)}[g,\bar{g } ] + { \cal r}_k\right)^{-1}_{\zeta_1\zeta_2 } \partial_t\left({\cal r}_k\right)_{\zeta_2\zeta_1}\right]\nonumber\\ & & + \frac{1}{2}{\rm tr}'\left[\sum\limits_{\psi_1,\psi_2\in\bar{i}_2 } \left(s_{\rm gh}^{(2)}[g,\bar{g } ] + { \cal r}_k\right)^{-1}_{\psi_1\psi_2 } \partial_t\left({\cal r}_k\right)_{\psi_2\psi_1 } \right]\;.\end{aligned}\ ] ] here @xmath186 and @xmath187 are the hessians of @xmath188 $ ] and @xmath189 $ ] with respect to the gravitational and the ghost component fields , respectively .they are taken at fixed @xmath33 . in order to obtain a tractable evolution equation for a given truncationit is necessary to use a cutoff which is adapted to this truncation but still has the desired suppression properties for a class of backgrounds which is as large as possible .a convenient cutoff which is adapted to the truncation ansatz can be found in the following way @xcite . given a truncation, we assume that for @xmath190 the kinetic operators of all modes with a definite helicity can be brought to the form @xmath191 where @xmath192 is a set of c - number functions and the indices @xmath193 refer to the different types of fields .then we choose the cutoff operator @xmath21 in such a way that the structure @xmath194 is achieved . herethe function @xmath195 , @xmath196 , describes the details of the mode suppression ; it is required to satisfy the boundary conditions @xmath197 and @xmath198 , but is arbitrary otherwise . by virtue of eq .( [ rule ] ) , the inverse propagator of a massless field mode with covariant momentum square @xmath199 is proportional to @xmath200 which equals @xmath23 for @xmath13 and @xmath99 for @xmath14 .this means that the small-@xmath23 modes , and only those , have acquired a mass @xmath201 which leads to the desired suppression . in order to see the potential problems of the rule ( [ rule ] )let us be more specific and assume that the functions @xmath202 are linear in @xmath203 and contain no constant term . then , after diagonalizing @xmath202 with respect to the field indices , @xmath20 decomposes into a set of ( massless , by assumption ) inverse propagators @xmath204 with a running wave function normalization @xmath205. in this diagonal basis @xmath21 is diagonal , too .it is of the form @xmath206 a priori @xmath161 is a free constant , but when we apply the rule ( [ rule ] ) we are forced to set @xmath207 for each mode .only then the propagator and the cutoff combine in the right way , leading to the modified inverse propagator @xmath208 $ ] .the choice @xmath207 is certainly the correct one if @xmath205 is positive .this is indeed the case in the familiar unitary theories on flat spacetime .in qeg , however , there are truncations , the einstein - hilbert truncation , for instance , which give a negative kinetic energy to certain modes @xmath209 of the metric . in particular , in the einstein - hilbert truncation, the conformal factor @xmath210 has @xmath211 . the important question is how @xmath161 should be chosen when @xmath205 is negative .if we continue to use @xmath207 , the rg equation is still well - defined because the inverse propagator @xmath212 $ ] never vanishes so that the functional traces on the rhs of ( [ truncflow ] ) are not suffering from any ir problem .in fact , if we write down the perturbative expansion of the traces , for instance , we see that all propagators are correctly cut off in the ir , and that loop momenta smaller than @xmath2 are suppressed correctly .this would not have been the case if we had insisted on a positive @xmath161 , setting @xmath213 . in this casethe modified inverse propagator @xmath214 $ ] , because of the relative minus sign between @xmath23 and the @xmath21-term , fails to suppress the ir modes .even worse , it can introduce a spurious singularity at the value of @xmath23 for which @xmath215 . at first sightthe choice @xmath213 might have appeared to be the more natural one because only if @xmath216 the factor @xmath217 is a damped exponential which suppresses the low - momentum modes under the path integral .nevertheless it was argued in @xcite that the `` @xmath207-rule '' is the correct choice both for @xmath218 and @xmath211 .the calculations in @xcite and all subsequent papers @xcite were based upon this rule . on the one hand, this rule guarantees that the rg equation is well - defined and consistent . on the other hand ,it is difficult to give a meaning to the euclidean functional integral from which this rg equation was derived at the formal level . in the case @xmath219 the factor @xmath220 is a growing exponential which seems to enhance rather than suppress the low - momentum modes .however , as suggested by the perturbative argument above , this conclusion is too naive probably .let us now come to the case where the functions @xmath202 are not linear in @xmath203 but , say , of the form @xmath221 . in this casethe rule ( [ rule ] ) demands that we choose the corresponding operators @xmath222 in such a way that they contain cutoff terms adjusted to both the @xmath223- and the @xmath203-terms of the kinetic operator . for the above example , which is relevant to the @xmath1-truncation, @xmath222 assumes the general form @xmath224 + { \cal z}^{(2)}_k\,k^2\,r^{(0)}\left(\frac{p^2}{k^2}\right)\end{aligned}\ ] ] where we omitted the indices referring to the types of fields .obviously we have to set @xmath225 and @xmath226 in order to achieve that the propagator and the cutoff combine as prescribed by the rule ( [ rule ] ) .this leads to the modified inverse propagator @xmath227 ^ 2+z^{(2)}_k[p^2+k^2r^{(0)}(p^2/k^2)]$ ] . for brevitywe refer to this prescription , too , as the @xmath207-rule .we believe that the @xmath207-rule is correct also for @xmath211 , and that it is the relation between the manifestly well - defined flow equation and the formal path integral that needs to be understood better .various attitudes are possible here . for instance , one could postulate that the fundamental definition of the theory is in terms of the flow equation rather than the path integral .since the former is much better defined than the latter ( in particular also with respect to the usual uv and ir problems ) one would simply discard the path integral then .another way out is to adopt the usual , albeit rather ad hoc , prescription of wick - rotating the conformal factor @xmath228 which turns the growing exponential into a decaying one .a much more attractive and less radical possibility is the following .presumably it will be possible to construct an effective average action for _ lorentzian _ quantum gravity by invoking a kind of stationary phase argument for the mode suppression .then one deals with oscillating exponentials @xmath229 , and apart from the trivial substitutions @xmath230 , @xmath231 the flow equation remains the same as in euclidean gravity . for @xmath207 it has all the desired features , and @xmath211 poses no special problem for the path integral .it is interesting that there are also recent indications @xcite coming from the dynamical triangulation approach to quantum gravity @xcite which suggest that the lorentzian path integral might have a better chance of being well - defined than the euclidean one . as the last possibility we mention the best of all situations , namely that in an exact treatment there are simply no factorsif this is actually the case , the conformal factor problem which we encounter in the einstein - hilbert and similar truncations would have the status of an unphysical truncation artifact .if so , the `` @xmath207-rule '' could be interpreted as a device which helps in approximating as well as possible the exact rg flow by a truncated flow .it is one of the main results of the present paper that this scenario is indeed realized to some extent .we shall see that within the @xmath1-truncation those terms of @xmath202 which dominate at sufficiently large momenta have @xmath218 at least for large enough values of @xmath2 @xmath232 . for too low scales@xmath233 some of the @xmath205 s might turn negative , but at these scales the @xmath1-truncation becomes unreliable probably @xcite so that the negative @xmath205 s might be due to an insufficient truncation .it is not excluded that in the exact theory the dominating @xmath205 s are positive down to @xmath234 .as all those @xmath205 s which determine the sign of the dominating contributions to @xmath202 are positive for large values of @xmath2 , the cutoff has the standard suppression properties and is not plagued by any conformal factor problem for @xmath7 .this , then , allows for an unambiguous investigation of the uv fixed point and its properties .it is quite remarkable that , as we shall see , all results concerning the fixed point are basically the same for the @xmath1-truncation ( where @xmath235 at least for the dominating terms ) and the einstein - hilbert truncation with both positive and negative factors @xmath207 .this is certainly a quite impressive confirmation of the @xmath207-rule . in the next sectionwe shall see in detail that for the truncation studied in this paper we can comply with the @xmath207-rule by using the following cutoff operators for the component fields : @xmath236\nonumber\\ & & + \left({\cal y}_k^{\bar{\phi}\bar{\sigma}}\,\bar{r}+ { \cal z}_k^{\bar{\phi}\bar{\sigma}}\kappa^2\right)\left[\sqrt{\left(\bar{p}_k + \frac{d}{d-1}\bar{d}_\mu\bar{r}^{\mu\nu}\bar{d}_\nu(-\bar{d}^2)^{-1}\right ) \bar{p}_k}\right.\nonumber\\ & & -\left.\sqrt{\left(\bar{d}^2\right)^2+\frac{d}{d-1 } \bar{d}_\mu\bar{r}^{\mu\nu}\bar{d}_\nu}\right]\;,\nonumber\\ \left({\cal r}_k\right)_{\bar{\phi}\bar{\phi}}&=&{\cal x}_k^{\bar{\phi } \bar{\phi}}\left(-2d^2\,k^2r^{(0)}(-d^2/k^2 ) + k^4r^{(0)}(-d^2/k^2)^2\right)\nonumber\\ & & + \frac{1}{2}\left\{{\cal z}_k^{\bar{\phi}\bar{\phi}}\kappa^2 + { \cal y}_k^{\bar{\phi}\bar{\phi}}\,\bar{r},k^2r^{(0)}(-d^2/k^2)\right\}\ ; , \nonumber\\ \left({\cal r}_k\right)^{\mu\nu}_{\bar{v}^tv^t}&=&-\left({\cal r}_k \right)^{\mu\nu}_{v^t\bar{v}^t}={\cal z}_k^{\bar{v}^t v^t}\ , \bar{g}^{\mu\nu}\,k^2 r^{(0)}\left(-\bar{d}^2/k^2\right)\;,\nonumber\\ \left({\cal r}_k\right)_{\bar{\varrho}\varrho } & = & -\left({\cal r}_k\right)_{\varrho\bar{\varrho } } = { \cal z}_k^{\bar{\varrho}\varrho}\,k^2 r^{(0)}\left(-\bar{d}^2/k^2\right)\;.\end{aligned}\ ] ] here @xmath237 is defined as @xmath238 and the curly brackets denote the anticommutator , i.e. @xmath239 for arbitrary operators @xmath240 , @xmath241 . the remaining cutoff operators which appear in ( [ 013 ] ) but are not listed in eq .( [ cutoff ] ) are set to zero .the constants @xmath242 , @xmath243 , and @xmath161 will be adjusted later .it should be noted that the terms proportional to the @xmath242 s and @xmath243 s provide the cutoff for those contributions to @xmath20 which come from the higher - derivative terms . for @xmath244 ,( [ cutoff ] ) actually boils down to the cutoff of type b used in @xcite in the context of the einstein - hilbert truncation .( this cutoff type has to be distinguished from the cutoff of type a used in the original paper @xcite , which is formulated in terms of the complete fields and does not involve the component fields . )each cutoff contains some `` shape function '' @xmath245 .a particularly suitable choice is the exponential shape function @xmath246^{-1}\;.\end{aligned}\ ] ] in order to check the scheme independence of universal quantities we employ a one - parameter generalization of ( [ expcut ] ) , the class of exponential shape functions , @xmath247^{-1}\;,\end{aligned}\ ] ] with the `` shape parameter '' @xmath248 parametrizing the profile of @xmath245 @xcite .another admissible choice we are going to use is the following class of shape functions with compact support : @xmath249\right ] & \;\;\;b < y<1.5\\0 & \;\;\;y\ge 1.5 \end{array}\right.\;.\end{aligned}\ ] ] here it is the shape parameter @xmath250 which parametrizes the profile of @xmath245 @xcite .[ @xmath1-trunc ] in all previous papers @xcite the flow equation of qeg was used in the einstein - hilbert truncation . in this sectionwe generalize this truncation by taking also an @xmath1-term with associated running coupling @xmath251 into account and we derive the rg flow within this `` @xmath1-truncation '' .we assume that , at the uv scale @xmath56 , gravity in @xmath30 dimensions is described by the action @xmath252=s[g]=\int d^dx\,\sqrt{g } \left\{\frac{1}{16\pi\bar{g}}\left(-r(g)+2\bar{\lambda}\right ) + \bar{\beta}\,r^2(g)\right\}\;.\end{aligned}\ ] ] it consists of the conventional einstein - hilbert action and a higher - derivative term with bare coupling @xmath95 . in order to study the rg flow of @xmath253 $ ] towards smaller scales @xmath254we employ a truncated action functional of the following form : @xmath255&=&\int d^dx\,\sqrt{g}\left\{2\kappa^2 z_{nk}\left ( -r(g)+2\bar{\lambda}_k\right)+\bar{\beta}_k\,r^2(g)\right\}\nonumber\\ & & + \kappa^2 \frac{z_{nk}}{\alpha}\int d^dx\,\sqrt{\bar{g}}\,\bar{g}^{\mu\nu } \left({\cal f}_\mu^{\alpha\beta } g_{\alpha\beta}\right)\left({\cal f}_\nu^{\rho\sigma}g_{\rho\sigma}\right)\;.\end{aligned}\ ] ] the ansatz ( [ trunc ] ) is obtained from @xmath256 by replacing @xmath257 so that its form agrees with that of the gravitational sector in the ansatz ( [ truncclass ] ) with @xmath258=\kappa^2 \frac{z_{nk}-1}{\alpha}\int d^dx\ , \sqrt{\bar{g}}\,\bar{g}^{\mu\nu}\left({\cal f}_\mu^{\alpha\beta } g_{\alpha\beta}\right)\left({\cal f}_\nu^{\rho\sigma}g_{\rho\sigma}\right)\;.\end{aligned}\ ] ] in principle , also the gauge fixing parameter @xmath259 should be treated as a scale - dependent quantity : @xmath260 .its evolution is neglected here for simplicity . however , setting @xmath261 by hand mimics a dynamical treatment of the gauge fixing parameter since @xmath261 can be argued to be a rg fixed point @xcite .the ansatz ( [ trunc ] ) comprises three @xmath2-dependent couplings .they satisfy the initial conditions @xmath262 , @xmath263 which implies @xmath264 , and @xmath265 . herethe uv scale @xmath54 is taken to be large but finite . in order to determine the evolution of @xmath41 , @xmath266 and @xmath251 towards smaller scales we have to project the flow equation onto the space spanned by the operators @xmath37 , @xmath38 and @xmath89 .after having inserted the ansatz ( [ trunc ] ) into both sides of the flow equation and having performed the @xmath32-derivatives implicit in @xmath20 we may set @xmath267 .as a consequence , the gauge fixing term drops out from the lhs which then reads @xmath268=\int d^dx\,\sqrt{\bar{g } } \left\{2\kappa^2\left[-\bar{r}(\bar{g})\,\partial_t z_{nk}+2\partial_t\left(z_{nk}\bar{\lambda}_k\right)\right ] + \bar{r}^2(\bar{g})\,\partial_t\bar{\beta}_k\right\}\;.\end{aligned}\ ] ] obviously the lhs is spanned by the operators @xmath37 , @xmath269 and @xmath89 .this means that we have to perform a derivative expansion on the rhs in order to extract precisely those contributions from the traces which are proportional to these operators . by equating the result to ( [ lhs ] ) and comparing the coefficients we can read off the system of coupled differential equations for @xmath41 , @xmath266 and @xmath251 . in order to make these technically rather involved calculations feasiblewe may insert any metric @xmath33 that is general enough to admit a unique identification of the operators spanning the truncated theory space .we exploit this freedom by assuming that @xmath33 corresponds to a maximally symmetric space .such spaces form a special class of einstein spaces and are characterized by @xmath270 with the curvature scalar @xmath271 considered a constant number rather than a functional of the metric .it is sufficient to employ spaces with @xmath272 , i.e. @xmath30-spheres @xmath273 .they are parametrized by their radius @xmath274 which is related to the curvature scalar and the volume in the usual way , @xmath275 we emphasize that the beta - functions of @xmath41 , @xmath266 , and @xmath251 do not depend on this choice for @xmath33 ; it is simply a technical trick without any physical meaning . in principlethe beta - functions could be computed without any specification of @xmath33 .while this projection technique is capable of distinguishing @xmath276 from both @xmath277 and @xmath278 , it can not disentangle the three @xmath96-invariants @xmath89 , @xmath279 , and @xmath280 which are all proportional to @xmath281 .if one wants to project them out individually one has to insert non - maximally symmetric spaces , but then the evaluation of the functional traces on the rhs of ( [ truncflow ] ) is a rather formidable problem with the present technology . in fact, this is one of the reasons for omitting the other two @xmath96-invariants from our truncation ansatz . in ref .@xcite we discussed already the expansion of fields defined on spherical backgrounds .both the classical and the quantum tt - component fields can be expanded in terms of transverse - traceless tensor harmonics @xmath282 , transverse vector harmonics @xmath283 , and scalar harmonics @xmath284 .they form complete sets of orthogonal eigenfunctions with respect to the corresponding covariant laplacians .we summarize the main results of @xcite in appendix [ harm ] .in particular , the expansions of @xmath285 , @xmath125 , @xmath110 , @xmath109 and their classical counterparts can be read off from eq .( [ 067 ] ) , while the remaining component fields are expanded according to @xmath286 similar expansions hold for the associated classical fields ( expectation values ) .contrary to eq .( [ 067 ] ) , the summations in eq .( [ 092 ] ) do _ not _ start at @xmath287 for vectors and at @xmath288 for scalars , but at @xmath289 for @xmath290 and @xmath291 , and at @xmath287 for the scalar ghost fields .the modes omitted here are the killing vectors @xmath292 , the solutions of the scalar equation ( [ 010 ] ) @xmath293 , and the constants @xmath294 .as we mentioned in section [ 3s2 ] , these modes do not correspond to fluctuations of @xmath107 or the faddeev - popov ghosts . as in @xcite , we decompose the quantum field @xmath125 into a part @xmath295 spanned by the same set of eigenfunctions as @xmath291 , and a part @xmath296 containing the contributions from the remaining modes : @xmath297 due to the orthogonality of the spherical harmonics , @xmath296 is orthogonal to @xmath295 and @xmath291 : @xmath298 @xmath299 .this implies @xmath300 and @xmath301 . as a consequence ,splitting @xmath125 according to eq .( [ split ] ) ensures that any nonzero bilinear cross term of the scalar fields is such that the scalars involved can be expanded in the same set of eigenfunctions .of course , the same holds for the corresponding classical fields @xmath302 and @xmath303 .let us now start with the evaluation of the rhs of eq .( [ truncflow ] ) .after having inserted the truncation ansatz ( [ trunc ] ) we identify the two metrics @xmath32 and @xmath33 .therefore it is sufficient to calculate the operators @xmath304 + { \cal r}_k[\bar{g}])^{-1}$ ] and @xmath305 + { \cal r}_k[\bar{g}])^{-1}$ ] at @xmath267 . to this endwe first expand the ansatz ( [ trunc ] ) according to @xmath306=\gamma_k[\bar{g},\bar{g}]+{\cal o}(\bar{h } ) + \gamma_k^{\rm quad}[\bar{h};\bar{g}]+{\cal o}(\bar{h}^3)\end{aligned}\ ] ] and concentrate on the part quadratic in @xmath167 , i.e. @xmath307 $ ] .this leads to @xmath308 = \int d^dx\,\sqrt{\bar{g}}\,\bar{h}_{\mu\nu}\bigg\{\kappa^2 z_{nk}\bigg [ -\left(\frac{1}{2}\delta^\mu_\rho\delta_\sigma^\nu+\frac{1 - 2\alpha}{4\alpha } \bar{g}^{\mu\nu}\bar{g}_{\rho\sigma}\right)\bar{d}^2}\\ & & + \frac{1}{4}\left(2\delta^\mu_\rho\delta^\nu_\sigma-\bar{g}^{\mu\nu } \bar{g}_{\rho\sigma}\right)\left(\bar{r}-2\bar{\lambda}_k\right ) + \bar{g}^{\mu\nu}\bar{r}_{\rho\sigma}-\delta_\sigma^\mu\bar{r}^\nu_{\ ; \rho}-\bar{r}^{\nu\;\;\mu}_{\;\rho\;\;\;\sigma}\nonumber\\ & & + \frac{1-\alpha}{\alpha } \left(\bar{g}^{\mu\nu}\bar{d}_\rho\bar{d}_\sigma-\delta^\mu_\sigma \bar{d}^\nu\bar{d}_\rho\right)\bigg]\nonumber\\ & & + \frac{1}{2}\bar{\beta}_k\bigg [ \frac{1}{2}\left(\frac{1}{2}\bar{g}^{\mu\nu}\bar{g}_{\rho\sigma } -\delta^\mu_\rho\delta^\nu_\sigma\right)\bar{r}^2 + 2\bar{g}^{\mu\nu}\bar{r}\left(-\bar{r}_{\rho\sigma}+\bar{d}_{\rho } \bar{d}_\sigma-\bar{g}_{\rho\sigma}\bar{d}^2\right)\nonumber\\ & & + 2r\left(\delta_\sigma^\nu\bar{r}^\mu_{\;\,\rho } -\bar{r}^{\mu\;\;\;\;\nu}_{\;\;\rho\sigma } -3 \delta^\nu_\sigma\bar{d}^{\mu}\bar{d}_{\rho } + 2\bar{g}_{\rho\sigma}\bar{d}^\mu\bar{d}^\nu + 2\delta^\mu_\rho\delta^\nu_\sigma \bar{d}^2 -\delta^\nu_\sigma\bar{d}_\rho\bar{d}^\mu\right ) + 2\delta^\nu_\sigma \bar{d}_\rho\bar{r}\bar{d}^\mu\nonumber\\ & & -3\delta^\mu_\rho\delta^\nu_\sigma\bar{d}^{\lambda}\bar{r}\bar{d}_\lambda + 4\delta^\nu_\sigma \bar{d}^\mu\bar{r}\bar{d}_\rho -4\bar{g}_{\rho\sigma}\bar{d}^{\mu}\bar{r}\bar{d}^\nu + \bar{g}^{\mu\nu}\bar{g}_{\rho\sigma}\bar{d}^{\lambda}\bar{r}\bar{d}_\lambda + 2\bar{r}^{\mu\nu}\bar{r}_{\rho\sigma}\nonumber\\ & & -4\bar{r}^{\mu\nu } \left(\bar{d}_\rho\bar{d}_\sigma-\bar{g}_{\rho\sigma}\bar{d}^2\right ) + 2\bar{d}^\mu\bar{d}^\nu\bar{d}_\rho\bar{d}_\sigma -4\bar{g}^{\mu\nu}\bar{d}^2\bar{d}_\rho\bar{d}_\sigma + 2\bar{g}^{\mu\nu}\bar{g}_{\rho\sigma}\left(\bar{d}^2\right)^2 \bigg]\bigg\}\bar{h}^{\rho\sigma}\;.\nonumber\end{aligned}\ ] ] at this stage @xmath33 is still arbitrary .in order to ( partially ) diagonalize this quadratic form we insert the family of @xmath273 background metrics into eq .( [ quadform ] ) and decompose @xmath167 according to eq .( [ tt ] ) .then we apply eq .( [ split ] ) to the classical field @xmath309 to decompose it as well .this yields @xmath310 = \int d^dx\,\sqrt{\bar{g}}\,\frac{1}{2}\bigg\ { \bar{h}^t_{\mu\nu}\bigg [ z_{nk}\kappa^2\left(-\bar{d}^2+a_t(d)\,\bar{r}-2\bar{\lambda}_k\right ) } \nonumber\\ & & + \bar{\beta}_k\left(\bar{r}\,\bar{d}^2+g_t(d)\,\bar{r}^2\right ) \bigg]\bar{h}^{t\mu\nu}\nonumber\\ & & + \bar{\xi}_\mu\left[\frac{2}{\alpha } z_{nk}\kappa^2\left(-\bar{d}^2+a_v(d,\alpha)\,\bar{r}-2\alpha \bar{\lambda}_k\right)+g_v(d)\,\bar{\beta}_k\,\bar{r}^2\right ] \bar{\xi}^\mu\nonumber\\ & & + \bar{\sigma}\bigg [ c_{s2}(d,\alpha)\,z_{nk}\kappa^2\left(-\bar{d}^2+a_{s2}(d,\alpha)\,\bar{r } + b_{s2}(d,\alpha)\,\bar{\lambda}_k\right)\nonumber\\ & & + \bar{\beta}_k\left(h_s(d)\,\left(\bar{d}^2\right)^2 -g_{s1}(d)\,\left(2\bar{r}\,\bar{d}^2+\bar{r}^2\right)\right ) \bigg]\bar{\sigma}\nonumber\\ & & + 2\bar{\phi}_1\bigg[c_{s2}(d,\alpha)\,c_{s3}(d,\alpha)\,z_{nk}\kappa^2 + \bar{\beta}_k\left(-h_s(d)\,\bar{d}^2 + 2g_{s1}(d)\,\bar{r}\right)\bigg ] \nonumber\\ & & \times\sqrt{-\bar{d}^2}\sqrt{-\bar{d}^2-\frac{\bar{r}}{d-1}}\,\bar{\sigma } \nonumber\\ & & + \sum\limits_{\bar{\phi}\in\{\bar{\phi}_0,\bar{\phi}_1\ } } \bar{\phi}\bigg[c_{s2}(d,\alpha)\,c_{s1}(d,\alpha)\,z_{nk}\kappa^2 \left(-\bar{d}^2+a_{s1}(d,\alpha)\,\bar{r } + b_{s1}(d,\alpha)\,\bar{\lambda}_k\right)\nonumber\\ & & + \bar{\beta}_k\left(h_s(d)\,\left(\bar{d}^2\right)^2 -g_{s2}(d)\,\bar{r}\,\bar{d}^2+g_{s3}(d)\,\bar{r}^2\right ) \bigg]\bar{\phi}\bigg\}\;.\end{aligned}\ ] ] here the various @xmath240 s , @xmath241 s , @xmath153 s and @xmath93 s and @xmath311 are functions of the dimensionality @xmath30 and the gauge fixing parameter @xmath259 .the explicit expressions for these coefficients are given in appendix [ s_kcoeff ] .this partial diagonalization is performed in order to simplify the inversion of the operator @xmath312+{\cal r}_k[\bar{g}]$ ] .in fact , this is the main reason for using the tt - decomposition ( [ tt ] ) and specifying a concrete background .note that in the pure einstein - hilbert truncation it is only the term in ( [ quadform ] ) which is proportional to @xmath313 that gives rise to mixings between the traceless part of @xmath167 and @xmath309 and therefore necessitates the complete decomposition ( [ tt ] ) . for @xmath314 ,a complete diagonalization can be achieved by merely splitting off the trace part @xcite .this has to be contrasted with the @xmath1-truncation where the higher - derivative term introduces additional mixings between the traceless part and @xmath309 .these cross terms do _ not _ vanish for @xmath314 .hence the complete decomposition of @xmath167 is necessary for a partial diagonalization even in the case @xmath314 . at the component field levelthe cross terms boil down to a purely scalar @xmath179-@xmath309 mixing term that vanishes for the spherical harmonics @xmath315 and @xmath316 . since these modes contribute to @xmath309 , but not to @xmath179 , we can not directly invert the associated matrix differential operator @xmath317 . as a way out, we split @xmath309 according to eq .( [ split ] ) into @xmath302 and @xmath303 .this has the effect that only mixings between the scalars @xmath179 and @xmath303 survive , which can be expanded in the same set of eigenfunctions @xmath284 starting at @xmath289 . hence the resulting matrix differential operator @xmath318 is invertible .however , it should be noted that this additional split of @xmath309 leads to a slightly modified flow equation since it affects the matrix structure of this operator .in fact , the summation in the gravitational sector of eq .( [ truncflow ] ) now runs over the set of fields @xmath319 , with @xmath320 and @xmath321 . as a next stepwe calculate the contributions from the ghost fields appearing on the rhs of eq .( [ truncflow ] ) . for this purposewe insert the family of spherical background spaces @xmath273 into @xmath118 and set @xmath322 .then we use eq .( [ t ] ) to decompose the ghost fields .this yields @xmath323=\sqrt{2}\int d^dx\,\sqrt{g}\left\{\bar{v}_\mu^t\left[-d^2-\frac{r}{d}\right ] v^{t\mu}+\bar{\varrho}\left[-d^2 - 2\frac{r}{d}\right]\varrho\right\}\;.\end{aligned}\ ] ] from now on the bars are omitted from the metric , the curvature and the operators @xmath324 and @xmath325 .before we can continue with our evaluation we have to specify the precise form of the cutoff operators . adapting them to @xmath20 and @xmath187 of eqs .( [ gravquad ] ) , ( [ ghquad ] ) by applying the rule ( [ rule ] ) leads precisely to the structure ( [ cutoff ] ) with the following choices for the @xmath242 s , @xmath243 s and @xmath161 s : @xmath326 thus , for @xmath267 , the nonvanishing entries of the matrix differential operators @xmath327 and @xmath328 take the form @xmath329+{\cal r}_k\right)_{\bar{h}^t\bar{h}^t } & = & z_{nk}\kappa^2\left(p_k+a_t(d)\,r-2\bar{\lambda}_k\right ) + \bar{\beta}_k\left(-r\,p_k+g_t(d)\,r^2\right)\;,\nonumber\\ \left(\gamma^{(2)}_k[g , g]+{\cal r}_k\right)_{\bar{\xi}\bar{\xi } } & = & \frac{2}{\alpha } z_{nk}\kappa^2\left(p_k+a_v(d,\alpha)\,r-2\alpha \bar{\lambda}_k\right)+g_v(d)\,\bar{\beta}_k\,r^2\;,\nonumber\\ \left(\gamma^{(2)}_k[g , g]+{\cal r}_k\right)_{\bar{\sigma}\bar{\sigma } } & = & c_{s2}(d,\alpha)\,z_{nk}\kappa^2\left(p_k+a_{s2}(d,\alpha)\,r + b_{s2}(d,\alpha)\,\bar{\lambda}_k\right)\nonumber\\ & & + \bar{\beta}_k\left(h_s(d)\,p_k^2 + g_{s1}(d)\,\left(2r\,p_k - r^2\right)\right)\;,\nonumber\\ \left(\gamma^{(2)}_k[g , g]+{\cal r}_k\right)_{\bar{\phi}_1\bar{\sigma } } & = & \left(\gamma^{(2)}_k[g , g]+{\cal r}_k\right)_{\bar{\sigma}\bar{\phi}_1 } \nonumber\\ & = & \bigg[c_{s2}(d,\alpha)\,c_{s3}(d,\alpha)\,z_{nk}\kappa^2 + \bar{\beta}_k\left(h_s(d)\,p_k+2g_{s1}\,r\right)\bigg ] \nonumber\\ & & \times\sqrt{p_k}\sqrt{p_k-\frac{r}{d-1}}\;,\nonumber\\ \left(\gamma^{(2)}_k[g , g]+{\cal r}_k\right)_{\bar{\phi}_0\bar{\phi}_0 } & = & \left(\gamma^{(2)}_k[g , g]+{\cal r}_k\right)_{\bar{\phi}_1\bar{\phi}_1 } \nonumber\\ & = & c_{s2}(d,\alpha)\,c_{s1}(d,\alpha)\,z_{nk}\kappa^2 \left(p_k+a_{s1}(d,\alpha)\,r + b_{s1}(d,\alpha)\,\bar{\lambda}_k\right)\nonumber\\ & & + \bar{\beta}_k\left(h_s(d)\,p_k^2 + g_{s2}(d)\,r\,p_k+g_{s3}(d)\,r^2\right)\;,\nonumber\\ \left(s_{\rm gh}^{(2)}[g , g]+{\cal r}_k\right)_{\bar{v}^tv^t } & = & -\left(s_{\rm gh}^{(2)}[g , g]+{\cal r}_k\right)_{v^t\bar{v}^t } = \sqrt{2}\left[p_k-\frac{r}{d}\right]\;,\nonumber\\ \left(s_{\rm gh}^{(2)}[g , g]+{\cal r}_k\right)_{\bar{\varrho}\varrho } & = & -\left(s_{\rm gh}^{(2)}[g , g]+{\cal r}_k\right)_{\varrho\bar{\varrho } } = \sqrt{2}\left[p_k-2\frac{r}{d}\right]\;.\end{aligned}\ ] ] for notational simplicity we set @xmath330\right)_{\psi_1\psi_2}\equiv \left(s_{\rm gh}^{(2)}[g , g]\right)_{\psi_1\psi_2}$ ] with @xmath331 .now we are in a position to write down the rhs of the flow equation with @xmath267. we shall denote it @xmath332 in the following .obviously we need the inverse operators @xmath333 and @xmath334 .this inversion is carried out in appendix [ inv ] .inserting the inverse operators into @xmath332 leads to the somewhat complicated result @xmath335\nonumber\\ & & + { \rm tr}_{(1t)}'\left[\left(p_k+a_v(d,\alpha)r-2\alpha \bar{\lambda}_k+g_v(d)\,(z_{nk}\kappa^2)^{-1}\bar{\beta}_k\,r^2\right)^{-1 } { \cal n}\right]\nonumber\\ & & + { \rm tr}_{(0)}''\bigg [ \bigg\{\big(c_{s2}(d,\alpha)\left(p_k+a_{s2}(d,\alpha)\,r + b_{s2}(d,\alpha)\,\bar{\lambda}_k\right)\nonumber\\ & & + ( z_{nk}\kappa^2)^{-1}\bar{\beta}_k\left(h_s(d)\,p_k^2 + g_{s1}(d)\,\left(2r\,p_k - r^2\right)\right)\big)\nonumber\\ & & \times\big(c_{s2}(d,\alpha)\,c_{s1}(d,\alpha ) \left(p_k+a_{s1}(d,\alpha)\,r + b_{s1}(d,\alpha)\,\bar{\lambda}_k\right)\nonumber\\ & & + ( z_{nk}\kappa^2)^{-1}\bar{\beta}_k\left(h_s(d)\,p_k^2 + g_{s2}(d)\,r\,p_k+g_{s3}(d)\,r^2\right)\big)\nonumber\\ & & -\big(c_{s2}(d,\alpha)\,c_{s3}(d,\alpha ) + ( z_{nk}\kappa^2)^{-1}\bar{\beta}_k\left(h_s(d)\,p_k+2g_{s1}\,r\right)\big)^2\ , p_k\big(p_k-\frac{r}{d-1}\big)\bigg\}^{-1}\nonumber\\ & & \times\bigg\{\big(c_{s2}(d,\alpha)\left(p_k+a_{s2}(d,\alpha)\,r + b_{s2}(d,\alpha)\,\bar{\lambda}_k\right)\nonumber\\ & & + ( z_{nk}\kappa^2)^{-1}\bar{\beta}_k\left(h_s(d)\,p_k^2 + g_{s1}(d)\,\left(2r\,p_k - r^2\right)\right)\big)\nonumber\\ & & \times\big(c_{s2}(d,\alpha)\,c_{s1}(d,\alpha ) \,{\cal n } + ( z_{nk}\kappa^2)^{-1}\bar{\beta}_k\left(h_s(d)\,{\cal t}_2 + g_{s2}(d)\,r\,{\cal t}_1\right)\big)\nonumber\\ & & + \big(c_{s2}(d,\alpha)\,c_{s1}(d,\alpha ) \left(p_k+a_{s1}(d,\alpha)\,r + b_{s1}(d,\alpha)\,\bar{\lambda}_k\right)\nonumber\\ & & + ( z_{nk}\kappa^2)^{-1}\bar{\beta}_k\left(h_s(d)\,p_k^2 + g_{s2}(d)\,r\,p_k+g_{s3}(d)\,r^2\right)\big)\nonumber\\ & & \times\big(c_{s2}(d,\alpha)\,{\cal n } + ( z_{nk}\kappa^2)^{-1}\bar{\beta}_k\left(h_s(d)\,{\cal t}_2 + 2g_{s1}(d)\,r\,{\cal t}_1\right)\big)\nonumber\\ & & + 2\big(c_{s2}(d,\alpha)\,c_{s3}(d,\alpha ) + ( z_{nk}\kappa^2)^{-1}\bar{\beta}_k\left(h_s(d)\,p_k+2g_{s1}\,r\right)\big)\ , \sqrt{p_k}\sqrt{p_k-\frac{r}{d-1}}\nonumber\\ & & \times \frac{1}{2z_{nk}\kappa^2}\,\partial_t\bigg ( h_s(d)\,\bar{\beta}_k\bigg[p_k\,\sqrt{\left(p_k + \frac{d}{d-1}\,d_\mu\,r^{\mu\nu}\,d_\nu(-d^2)^{-1}\right ) p_k}\nonumber\\ & & + d^2\,\sqrt{\left(d^2\right)^2+\frac{d}{d-1 } \,d_\mu\,r^{\mu\nu}\,d_\nu}\bigg]\nonumber\\ & & + \big(2g_{s1}(d)\,\bar{\beta}_k\,r+ c_{s2}(d,\alpha)\,c_{s3}(d,\alpha)\,z_{nk}\kappa^2\big)\nonumber\\ & & \times \bigg[\sqrt{\left(p_k+\frac{d}{d-1}\,d_\mu\,r^{\mu\nu}\,d_\nu(-d^2)^{-1}\right ) p_k}-\sqrt{\left(d^2\right)^2+\frac{d}{d-1 } \,d_\mu\,r^{\mu\nu}\,d_\nu}\bigg]\bigg)\bigg\}\bigg]\nonumber\\ & & -2{\rm tr}_{(1t)}\left[\left(p_k-\frac{r}{d}\right)^{-1}{\cal n}_0\right ] -2{\rm tr'}_{(0)}\left[\left(p_k-2\frac{r}{d}\right)^{-1}{\cal n}_0\right ] \nonumber\\ & & + \frac{1}{2z_{nk}\kappa^2}\sum\limits_{l=0}^{1}\bigg [ d_l(d,0)\bigg\{c_{s2}(d,\alpha)\,c_{s1}(d,\alpha)\,\big(\lambda_l(d,0)+ k^2r^{(0)}(\lambda_l(d,0)/k^2)\nonumber\\ & & + a_{s1}(d,\alpha)\,r + b_{s1}(d,\alpha)\,\bar{\lambda}_k\big)\nonumber\\ & & + ( z_{nk}\kappa^2)^{-1}\bar{\beta}_k\big(h_s(d)\,\left(\lambda_l(d,0)+ k^2r^{(0)}(\lambda_l(d,0)/k^2)\right)^2\nonumber\\ & & + g_{s2}(d)\,r\left(\lambda_l(d,0)+ k^2r^{(0)}(\lambda_l(d,0)/k^2)\right)+g_{s3}(d)\,r^2\big)\bigg\}^{-1 } \nonumber\\ & & \times\partial_t\bigg\ { c_{s2}(d,\alpha)\,c_{s1}(d,\alpha)\,z_{nk}\kappa^2 \,k^2r^{(0)}(\lambda_l(d,0)/k^2)\nonumber\\ & & + \bar{\beta}_k\big(h_s(d)\ , \left(2\lambda_l(d,0)\,k^2r^{(0)}(\lambda_l(d,0)/k^2 ) + k^4r^{(0)}(\lambda_l(d,0)/k^2)^2\right)\nonumber\\ & & + g_{s2}(d)\,r\,k^2r^{(0)}(\lambda_l(d,0)/k^2)\big)\bigg\ } \bigg]\end{aligned}\ ] ] the new quantities @xmath336 , @xmath337 , @xmath338 and @xmath339 introduced in eq .( [ squareflow ] ) are defined as @xmath340\nonumber\\ & = & \left[1-\frac{1}{2}\eta_n(k)\right]k^2r^{(0)}(-d^2/k^2 ) + d^2r^{(0)'}(-d^2/k^2)\;,\nonumber\\ { \cal n}_0&\equiv&2^{-1 } \partial_t \left[k^2r^{(0)}(-d^2/k^2)\right ] = k^2r^{(0)}(-d^2/k^2)+d^2r^{(0)'}(-d^2/k^2)\;,\nonumber\\ { \cal t}_1&\equiv&\left(2\bar{\beta}_k\right)^{-1 } \partial_t \left[\bar{\beta}_k\,k^2r^{(0)}(-d^2/k^2)\right]\nonumber\\ & = & \left[1-\frac{1}{2}\eta_{\beta}(k)\right]k^2r^{(0)}(-d^2/k^2 ) + d^2r^{(0)'}(-d^2/k^2)\;,\nonumber\\ { \cal t}_2&\equiv&\left(2\bar{\beta}_k\right)^{-1 } \partial_t \left[\bar{\beta}_k\left(-2d^2\,k^2r^{(0)}(-d^2/k^2 ) + k^4r^{(0)}(-d^2/k^2)^2\right)\right]\nonumber\\ & = & 2 p_k\left[k^2r^{(0)}(-d^2/k^2 ) + d^2r^{(0)'}(-d^2/k^2)\right]\nonumber\\ & & -\frac{1}{2}\eta_{\beta}(k)\left(-d^2\,k^2r^{(0)}(-d^2/k^2 ) + k^4r^{(0)}(-d^2/k^2)^2\right)\;.\end{aligned}\ ] ] here @xmath341 and @xmath342 are the anomalous dimensions of the operators @xmath38 and @xmath89 , respectively .furthermore , the prime at @xmath245 denotes the derivative with respect to the argument . in eq .( [ squareflow ] ) we refined our notation concerning the primes at the traces . from now on oneprime indicates that the mode corresponding to the lowest eigenvalue has to be excluded , while two primes indicate the subtraction of the contributions from the lowest two eigenvalues .the subscripts at the traces describe on which kind of field the operators under the traces act .we use the subscripts @xmath343 , @xmath344 and @xmath345 for spin-0 fields , transverse spin-1 fields , and symmetric transverse traceless spin-2 fields , respectively .next we derive the flow equations for the couplings . in order to make the rather complicated calculations feasiblewe are forced to work from now on in the technically convenient gauge @xmath314 .here we merely summarize the main steps , the details of the calculation can be found in appendix [ trace ] . by expanding @xmath332 where @xmath347 with respect to @xmath274 and evaluating the traces by means of heat kernel techniques we extract those pieces from the rhs of the flow equation , eq .( [ squareflow ] ) , which are proportional to the appropriate powers of the radius , i.e. @xmath348 , @xmath349 , and @xmath350 .then we equate the result to the lhs , eq .( [ lhs ] ) , and compare the coefficients of the various powers of @xmath274 .this leads to a system of coupled differential equations for @xmath41 , @xmath266 and @xmath251 . in order to present it in a transparent mannerwe introduce the dimensionless running cosmological constant @xmath351 the dimensionless running newton constant @xmath352 and the dimensionless running @xmath1-coupling @xmath353 @xmath354 denotes the dimensionful running newton constant . in terms of the couplings@xmath48 , @xmath47 , and @xmath346 , our final result for the 3-dimensional flow equation reads @xmath355 @xmath356g_k\;,\end{aligned}\ ] ] @xmath357\beta_k\;.\end{aligned}\ ] ] the anomalous dimensions are explicitly given by @xmath358-c_1(\lambda_k , g_k,\beta_k;d)b_3(\lambda_k , g_k,\beta_k;d ) } { \left[1-g_k b_2(\lambda_k , g_k,\beta_k;d)\right]\left[\beta_k + c_3(\lambda_k , g_k,\beta_k;d)\right]+g_k c_2(\lambda_k , g_k,\beta_k;d ) b_3(\lambda_k , g_k,\beta_k;d)}\nonumber\end{aligned}\ ] ] and @xmath359+g_kb_1(\lambda_k , g_k,\beta_k;d ) c_2(\lambda_k , g_k,\beta_k;d)}{\left[1-g_k b_2(\lambda_k , g_k,\beta_k;d)\right ] \left[\beta_k+c_3(\lambda_k , g_k,\beta_k;d)\right]+g_k c_2(\lambda_k , g_k,\beta_k;d ) b_3(\lambda_k , g_k,\beta_k;d)}\;.\nonumber\end{aligned}\ ] ] the three @xmath360-functions @xmath361 , @xmath362 and @xmath363 contain the quantities @xmath364 , @xmath365 , @xmath366 , @xmath367 , which are extremely complicated functions of the couplings and the dimensionality @xmath30 .the explicit expressions for these coefficient functions can be found in appendix [ coefffnct ] .they contain the new threshold functions @xmath368 and @xmath369 of eqs .( [ psi ] ) and ( [ psitilde ] ) which functionally depend on @xmath245 .they generalize the familiar threshold functions @xmath370 and @xmath371 which occur in the einstein - hilbert truncation .the three @xmath360-functions are one of the main results of this paper .because of its complexity it is clearly impossible to solve the system of flow equations for @xmath48 , @xmath47 and @xmath346 , eqs .( [ del ] ) , ( [ deg ] ) and ( [ deb ] ) , exactly . even a numerical solution would be a formidable task .however , it is possible to gain important information about the general structure of the rg flow by looking at its fixed point structure . given a set of @xmath372-functions corresponding to an arbitrary set of dimensionless essential couplings @xmath373 , it is often possible to predict their scale dependence for very small and/or very large scales @xmath2 by investigating their fixed points .they are those points in the space spanned by the @xmath374 where all @xmath372-functions vanish .fixed points are characterized by their stability properties .a given eigendirection of the linearized flow is said to be uv or ir attractive ( or stable ) if , for @xmath7 or @xmath35 , respectively , the trajectories are attracted towards the fixed point along this particular direction .the uv ( ir ) critical hypersurface @xmath57 ( @xmath375 ) in the space of all couplings is defined to consist of all trajectories that run into a given fixed point for @xmath7 ( @xmath35 ) . in quantum field theory ,fixed points play an important role in the modern approach to renormalization theory @xcite . at a uv fixed pointthe infinite cutoff limit can be taken in a controlled way , the theory can be renormalized nonperturbatively there . as for gravity , weinberg @xcite argued that a theory described by a rg trajectory lying on a _ finite - dimensional _ uv critical hypersurface of some fixed point is presumably free from unphysical singularities .it is predictive since it depends only on a _ finite _ number of free ( essential ) parameters . in weinberg s words ,such a theory is _asymptotically safe_. asymptotic safety has to be regarded as a generalized , nonperturbative version of renormalizability .it covers the class of perturbatively renormalizable theories , whose infinite cutoff limit is taken at the gaussian fixed point @xmath376 , as well as those perturbatively nonrenormalizable theories which are described by a rg trajectory on a finite - dimensional uv critical hypersurface of a non - gaussian fixed point @xmath377 and are nonperturbatively renormalizable therefore @xcite .let us now consider the component form of the exact rg equation , i.e the system of differential equations @xmath378 for a set of dimensionless essential couplings @xmath379 .in an exact treatment the number @xmath380 is infinite ; in a specific truncation it might be finite .we assume that @xmath381 is a fixed point of eq .( [ fp0 ] ) , i.e. @xmath382 for all @xmath383 .we linearize the rg flow about @xmath381 which leads to @xmath384 where @xmath385 are the entries of the stability matrix @xmath386 . diagonalizing @xmath387 according to @xmath388 , @xmath389 , where @xmath390 is the right - eigenvector of @xmath387 with eigenvalue @xmath391 we have @xmath392 the general solution to eq .( [ gfp4 ] ) may be written as @xmath393 here @xmath394 are arbitrary real parameters and @xmath395 is a reference scale .obviously a fixed point @xmath396 is uv attractive for a given trajectory ( i.e. attractive for @xmath7 ) only if all its @xmath397 corresponding to negative @xmath398 are set to zero .therefore the dimensionality @xmath399 of the uv critical hypersurface belonging to this particular fixed point equals the number of positive @xmath400 s .conversely , for a trajectory where all @xmath397 corresponding to positive @xmath400 are set to zero , @xmath381 is an ir attractive fixed point ( approached in the limit @xmath35 ) . as a consequence ,the ir critical hypersurface @xmath375 of a fixed point has a dimensionality @xmath401 which equals the number of negative @xmath400 s . in a slight abuse of languagewe shall refer to the @xmath400 s as the _critical exponents_. strictly speaking , the solution ( [ gfp5 ] ) and its above interpretation is valid only in such cases where all eigenvalues @xmath391 are real , which is not guaranteed since the matrix @xmath402 is not symmetric in general . if complex eigenvalues occur one has to consider complex @xmath397 s and to take the real part of eq .( [ gfp5 ] ) , see below .then the real parts of the critical exponents determine which directions in coupling constant space are attractive or repulsive .at this point it is necessary to discuss the impact a change of the cutoff scheme has on the scaling behavior .since the path integral for @xmath3 depends on the cutoff scheme , i.e. on the @xmath8 chosen , it is clear that generically the @xmath2-dependent couplings and their fixed point values are scheme dependent .hence a variation of the cutoff scheme , i.e. of @xmath21 , induces a change in the corresponding @xmath387-matrix .so one might naively expect that also its eigenvalues , the critical exponents , are scheme dependent .in fact , this is not the case . according to the general theory of critical phenomena and a recent reanalysis in the framework of the exact rg equations @xcite any variation of the cutoff schemecan be generated by a specific coordinate transformation in the space of couplings , with the cutoff held fixed .such transformations leave the eigenvalues of the @xmath387-matrix invariant , so that the critical behavior near the corresponding fixed point is universal .the positions of fixed points are scheme dependent but their ( non)existence and the qualitative structure of the rg flow are @xmath21-independent features .quantities like the @xmath400 s which are @xmath21-independent are called _universal_. their residual scheme dependence present in an approximate treatment ( truncation , etc . ) can be used in order to judge the quality of the approximation .a truncation can be considered reliable only if it predicts the same fixed point structure for all admissible choices of @xmath21 . in the context of the @xmath1-truncationthe space of couplings is parametrized by @xmath403 , @xmath404 and @xmath405 .the @xmath406-functions occurring in the three flow equations @xmath407 are given in eqs .( [ del ] ) , ( [ deg ] ) and ( [ deb ] ) , respectively . in ref .@xcite the fixed point structure of the pure einstein - hilbert truncation was investigated . in this casethe @xmath360-functions were found to have both a trivial zero at @xmath408 , referred to as the * gaussian fixed point * , and a * non - gaussian fixed point * at @xmath51 , @xmath52 .as we will see in subsection [ 3s5b ] , the gaussian fixed point is _ not _ present any more in the generalized truncation .this has to be contrasted with the non - gaussian fixed point which is found with the @xmath1-truncation , too . in subsection [ 3s5c ]we study its cutoff dependence and the cutoff dependence of the associated critical exponents employing the above @xmath360-functions with the families of shape functions ( [ expshape ] ) or ( [ supp ] ) inserted . in this subsectionwe study the fate of the gaussian fixed point found in the context of the pure einstein - hilbert truncation . in @xcitewe investigated the 2-dimensional rg flow near this fixed point @xmath50 and discussed its stability properties .it is an important question how the situation changes by enlarging the parameter space .quite remarkably , we find that in the 3-dimensional @xmath43-@xmath70-@xmath104space of the @xmath1-truncation there is no gaussian fixed point , i.e. @xmath409 is not a simultaneous zero of all three @xmath360-functions .while @xmath361 and @xmath362 vanish at the origin @xmath409 , setting @xmath410 in @xmath363 leads to @xmath411 the nonzero constant @xmath412 is given by @xmath413 here the @xmath414 s are defined as in subsection [ betacoeff ] . in @xmath73 dimensions@xmath415 is a universal quantity since @xmath416 independently of the cutoff , see appendix [ threshprop ] .although there is no fixed point at the origin of the parameter space it is nevertheless very interesting to study the rg flow in the vicinity of @xmath409 . for simplicitywe restrict our considerations to the case @xmath417 . expanding @xmath418 ) about the originwe obtain instead of ( [ gfp4 ] ) the _ inhomogeneous _ system @xmath419 the linearized renormalization group flow is governed by the jacobi - matrix @xmath420 , @xmath421 , which takes the form @xmath422 its entries follow from the expanded @xmath360-functions ( [ expandbeta ] ) of appendix [ approxngfp ] . here@xmath423 , @xmath424 and @xmath425 are @xmath30-dependent parameters defined as @xmath426 @xmath427 @xmath428 } \nonumber\\ & & \times\left[\frac{1}{4}h_{34}(d)\,\widetilde{\phi}^1_{d/2 - 2}(0 ) + \frac{1}{8}h_{35}(d)\,\widetilde{\phi}^2_{d/2 - 1}(0 ) + \frac{1}{12}h_{36}(d)\,\widetilde{\phi}^3_{d/2}(0)\right]\nonumber\\ & & + \left[h_{31}(d)\,\phi^1_{d/2 - 2}(0 ) + h_{32}(d)\,\phi^2_{d/2 - 1}(0)+h_{33}(d)\,\phi^3_{d/2}(0)\right]\\ & & \times\left[h_{39}(d)\,\widetilde{\phi}^0_{d/2 - 2}(0 ) + h_{40}(d)\,\widetilde{\phi}^1_{d/2 - 1}(0 )+ h_{41}(d)\,\widetilde{\phi}^2_{d/2}(0 ) + h_{42}(d)\,\widetilde{\phi}^3_{d/2 + 1}(0)+\frac{3}{2}\right]\bigg\}\;.\nonumber\end{aligned}\ ] ] at this point it should be noted that the sub - matrix @xmath429 coincides precisely with the stability matrix of the gaussian fixed point which was calculated in @xcite in the einstein - hilbert truncation .diagonalizing the matrix ( [ gaussianstab ] ) yields the ( obviously universal ) eigenvalues @xmath430 , @xmath431 and @xmath432 which are associated with the eigenvectors @xmath433 eq .( [ ev ] ) is valid only for @xmath434 .in @xmath435 we obtain @xmath436 , @xmath437 , and in the @xmath438-dimensional case the eigenvectors are @xmath439 , @xmath440 . thus in both cases the space spanned by the eigenvectors in only 2-dimensional , i.e. they do not form a complete system .for all values of @xmath30 , including @xmath435 and @xmath441 , the solutions for @xmath48 and @xmath47 obtained from the linearized system ( [ lingauss ] ) assume the following form : @xmath442 since the expanded @xmath360-function @xmath362 of eq .( [ expandbeta ] ) does not depend on @xmath48 and @xmath346 up to terms of third order in the couplings we can easily calculate also the next - to - leading approximation for @xmath47 near the origin . in terms of the dimensionful quantity @xmath40 this improved solution reads @xmath443^{-1}\end{aligned}\ ] ] with @xmath444 a @xmath30-dependent parameter .it agrees with the @xmath445 defined in @xcite in the context of the pure einstein - hilbert truncation . for @xmath446 and with the reference scale @xmath447 ( which is admissible only for specific initial conditions of the cosmological constant ) eq .( [ g_kflow ] ) yields @xmath448 \;.\end{aligned}\ ] ] for the dimensionful cosmological constant we obtain from eq .( [ lambdag ] ) @xmath449 eqs .( [ g_kflow2 ] ) and ( [ gfp11 ] ) agree completely with the corresponding results from the einstein - hilbert truncation .let us now discuss the solution for @xmath346 . in order to derive it we start by picking @xmath450 in ( [ lingauss ] ) and rewrite the corresponding equation as @xmath451=k^{d-5 } \left[\gamma_d+\varsigma_d\,\lambda_k+\tau_d\,g_k\right]\;.\end{aligned}\ ] ] then we insert the solutions for @xmath47 and @xmath48 of eq .( [ lambdag ] ) into ( [ rewrite ] ) .the resulting differential equation may easily be solved . for @xmath452 ,the solution reads @xmath453 \,\left(\frac{k_0}{k}\right)^{d-4}\;.\end{aligned}\ ] ] the solutions in @xmath435 , @xmath73 , and @xmath441 can be obtained from ( [ betan4 ] ) by a careful evaluation of the limits @xmath454 , @xmath455 , and @xmath456 , respectively .in the most interesting case of @xmath73 dimensions this leads to the following solution : @xmath457 the parameters appearing in eq .( [ beta4 ] ) are @xmath458 \left[-\frac{559}{1728 } + \frac{71}{288}\,\widetilde{\phi}^2_{1}(0 ) + \frac{347}{288}\,\widetilde{\phi}^3_{2}(0)\right]\nonumber\\ & & + \frac{419}{1080 } \left[-\frac{299}{180}+\frac{13}{3}\,\widetilde{\phi}^1_1(0 ) + \frac{40}{3}\,\widetilde{\phi}^2_{2}(0)\right]\bigg\}\end{aligned}\ ] ] employing the exponential shape function with @xmath459 and inserting the corresponding values of the @xmath460- and @xmath461-integrals given in appendix [ threshprop ] we obtain @xmath462 let us now analyze the rg flow near the origin of the parameter space . strictly speakingour analysis even extends to all points of the @xmath43-@xmath70-@xmath104space which satisfy @xmath463 and @xmath464 .this is because in any of the three @xmath360-functions all terms of second and higher orders in @xmath346 appear as products @xmath465 with @xmath466 . since @xmath467 and , for @xmath417 , @xmath468 , @xmath48 of eq .( [ lambdag ] ) starts growing as soon as @xmath2 falls below @xmath395 and the linearization breaks down , unless the couplings run along a trajectory which satisfies @xmath469 for sufficiently small values of @xmath2 @xcite , with @xmath40 given by eq .( [ g_kflow ] ) . in this case both @xmath48 and @xmath47 approach @xmath470 in the limit @xmath35 as long as @xmath471 is satisfied as well .since @xmath472 depends on the shape function @xmath245 , @xmath423 is not a universal quantity .therefore the slope of the distinguished trajectories characterized by ( [ irstable ] ) is not fixed in a universal manner .( [ irstable ] ) is exactly the condition for the `` separatrix '' found in @xcite in the context of the @xmath473-dimensional einstein - hilbert truncation . in the terminology of @xcite ,the separatrix is the `` type iia trajectory '' that interpolates between the gaussian and the non - gaussian fixed point of the einstein - hilbert truncation , thereby separating the region of trajectories with @xmath474 ( type ia ) from those which hit the boundary of parameter space ( @xmath475 ) for some finite value of @xmath2 ( type iiia ) . in fig .[ plotx](a ) we depict the separatrix , a type ia and a type iiia trajectory in the vicinity of @xmath476 .the plot should be thought of as a projection from @xmath43-@xmath70-@xmath104space onto its @xmath477plane . for the separatrix ,the equation ( [ g_kflow2 ] ) for the running of @xmath40 is valid down to @xmath234 because @xmath47 stays near the origin .for @xmath478 the parameters @xmath479 and @xmath480 appearing in @xmath445 are scheme dependent , and @xmath445 is nonuniversal . in the most interesting case of @xmath73 , @xmath481\;.\end{aligned}\ ] ] since @xmath482 and @xmath483 are positive for any admissible shape function with @xmath459 , for instance , we have @xmath484 , @xmath485 so that @xmath486 .furthermore , we have @xmath487 where @xmath488 denotes the riemann zeta function , and thus @xmath489 . ]we can infer from eq .( [ omega4 ] ) that @xmath490 is positive .thus , if we define qeg with vanishing renormalized cosmological constant to be the theory described by a trajectory in @xmath43-@xmath70-@xmath104space , whose @xmath43- and @xmath70-coordinates follow the separatrix in the limit @xmath35 , eq .( [ g_kflow ] ) implies that qeg is antiscreening in the ir , i.e. @xmath40 decreases as @xmath2 increases . up to now we investigated the @xmath491-projection of the flow on @xmath43-@xmath70-@xmath104space .next we discuss the linearized rg flow of the @xmath492-component . in @xmath73 the solution for @xmath346 , eq .( [ beta4 ] ) , diverges in the limit @xmath35 . if @xmath493 as @xmath494 , the leading divergence is quadratic in @xmath2 .however , in this case the linearization can not be trusted down to arbitrarily small values of @xmath2 anyhow since trajectories with @xmath493 ultimately run away from @xmath476 for @xmath35 . for the distinguished trajectories which satisfy the separatrix condition ( [ irstable ] ) for @xmath35 , the coefficient of the term @xmath495 in ( [ beta4 ] ) vanishes and only a logarithmic running with a _universal _ coefficient remains : @xmath496 since higher orders in @xmath346 appear exclusively as products @xmath465 with @xmath466 , the vanishing of @xmath497 in the limit @xmath35 then implies that terms of order @xmath498 remain negligible as @xmath35 . as a consequence ,the linearization does not break down for @xmath35 although @xmath346 diverges in this limit . according to eqs .( [ lambdag ] ) , ( [ irstable ] ) , @xmath48 and @xmath47 quickly approach @xmath470 so that the corresponding trajectories run almost along the @xmath492-axis for @xmath35 , and the rg flow becomes essentially one - dimensional .this logarithmic running of @xmath346 was expected on the basis of conventional perturbation theory @xcite .we observe that @xmath499 _de_creases logarithmically with _ in_creasing @xmath2 .this is what is usually referred to as the `` asymptotic freedom '' of the @xmath96-coupling .we emphasize , however , that according to our results this logarithmic running occurs only close to @xmath500 and does not represent the true short - distance behavior of the theory .in fact , we shall find that @xmath346 runs towards a fixed point value @xmath501 for @xmath7 .= 7.9 cm = 5.2 cm ( a ) = 7.9 cm = 5.2 cm ( b ) fig .[ plotx](b ) shows three typical trajectories of the @xmath1-truncation close to @xmath409 ; all of them satisfy the separatrix condition ( [ irstable ] ) .their @xmath492-component diverges logarithmically towards @xmath502 as @xmath2 goes to zero , which is due to the positive coefficient in front of the @xmath503-term in ( [ beta4.2 ] ) . in this figurewe also depict the common projection of the trajectories onto the @xmath43-@xmath44plane .it coincides precisely with the separatrix of the einstein - hilbert truncation @xcite , i.e. the curve in fig .[ plotx](a ) that hits @xmath476 .conversely , all trajectories of the @xmath1-truncation satisfying ( [ irstable ] ) represent specific `` lifts '' of the separatrix with nonvanishing @xmath492-components ; they are distinguished by their @xmath504- and @xmath505-values .let us now discuss the @xmath492-evolution for @xmath507 .again , the linearization breaks down for trajectories which do not satisfy ( [ irstable ] ) for @xmath35 since @xmath508 in this case. therefore we restrict our considerations to the trajectories with @xmath509 for sufficiently small @xmath2 . in this case the second , quadratically divergent term of eq .( [ betan4 ] ) drops out , and the only powers of @xmath2 which occur in @xmath346 are @xmath510 and @xmath511 . in @xmath512 ,both @xmath431 and @xmath432 are positive which implies that the rg trajectories considered are attracted towards @xmath513 as @xmath2 is sent to zero . for @xmath514 ,@xmath515 is negative and thus @xmath346 contains a divergent term @xmath516 . as a consequence , the coefficient of this term must vanish , if a trajectory is to hit the point @xmath513 in the limit @xmath517 .the distinguished trajectory which runs into this point as @xmath35 satisfies , for sufficiently small values of @xmath2 , @xmath518 for all other trajectories the @xmath492-component diverges for @xmath35 .however , higher orders of @xmath346 are again suppressed by powers of @xmath47 and may therefore be neglected .( note that @xmath519 for @xmath466 . ) as a consequence , the linearization can be trusted down to arbitrarily small scales @xmath2 even in this case .the shape of the corresponding trajectories resembles the one found in @xmath73 . while @xmath520 , the @xmath43- and @xmath70-components approach @xmath470 in the limit @xmath35 .thus , for sufficiently small @xmath2 , the trajectories are almost straight lines which virtually coincide with the @xmath492-axis . having a closer look at the @xmath360-functions one recognizes that the ir scaling behavior in @xmath521 dimensions is actually governed by a * `` quasi - gaussian '' fixed point * at @xmath522 the quasi - gaussian fixed point is not present in @xmath73 . linearizing the rg flow about this fixed point yields essentially the same results as our expansion about @xmath523 above . the linearized @xmath360-functions with stability matrix @xmath387 , and the linear solutions associated to this fixed point ,may be obtained from eqs .( [ lingauss ] ) , ( [ gaussianstab ] ) , ( [ lambdag ] ) , ( [ betan4 ] ) and ( [ irstable2 ] ) simply by replacing @xmath425 with @xmath524 in particular , @xmath525 .the constants @xmath526 assume the meaning of critical exponents now , and their signs determine the dimensionality @xmath527 of the ( truncated ) ir critical hypersurface @xmath375 of the quasi - gaussian fixed point .in @xmath528 we have one positive critical exponent @xmath529 and two negative critical exponents @xmath530 . therefore , within the truncation , @xmath531 , as suggested by the corresponding solutions discussed above .in @xmath532 , @xmath533 and @xmath534 are positive and @xmath535 is negative .hence , in this case @xmath536 , i.e. @xmath375 consists of a single trajectory . for sufficiently small values of @xmath2this ir critical trajectory is given by eq .( [ irstable2 ] ) with @xmath425 replaced with @xmath537 .since the parameters @xmath423 , @xmath424 and @xmath537 contain @xmath245-dependent integrals @xmath538 , @xmath539 , they are not universal .therefore the slopes in both directions of the distinguished trajectory ( [ irstable2 ] ) are not fixed in a universal manner .this is in accordance with the general expectation that the eigenvalues of @xmath387 should be universal , but not its eigenvectors .= 7.9 cm = 5.2 cm ( a ) = 7.9 cm = 5.2 cm ( b ) we illustrate our results for @xmath521 in fig . [ ploty ] . in fig .[ ploty](a ) we consider @xmath435 and in fig .[ ploty](b ) the @xmath540-dimensional case .each figure shows three typical trajectories in the vicinity of the quasi - gaussian fixed point .all of them satisfy ( [ irstable ] ) , so that in both @xmath435 and @xmath541 the projections of the 3 trajectories onto the @xmath477plane coincide with the separatrix . in @xmath435 ,[ ploty](a ) , all 3 trajectories hit the fixed point , independently of their @xmath504-value . as we already pointed out above , the quasi - gaussian fixed point in @xmath542 is ir attractive for _ all _ trajectories satisfying ( [ irstable ] ) . in @xmath532this is no longer the case .this is confirmed by fig .[ ploty](b ) for @xmath541 . hereonly one of the trajectories hits the quasi - gaussian fixed point for @xmath35 , and this is precisely the one which satisfies the additional condition ( [ irstable2 ] ) .the other two trajectories shown in fig . [ ploty](b ) correspond to @xmath504-values which are different from the one in ( [ irstable2 ] ) and thus their @xmath492-component diverges in the limit @xmath35 .depending on the @xmath504-value , @xmath346 runs towards @xmath543 or @xmath502 .now we turn to the nontrivial simultaneous zeros of the set of @xmath360-functions @xmath544 given by eqs .( [ del ] ) , ( [ deg ] ) , ( [ deb ] ) .such non - gaussian fixed points with @xmath545 , @xmath396 , @xmath501 all different from zero have the anomalous dimensions @xmath546 which follow immediately from eqs .( [ deg ] ) and ( [ deb ] ) . in @xmath73 dimensions , and for the cutoff of the type a introduced in @xcite , the non - gaussian fixed point of the pure einstein - hilbert truncationwas first discussed in @xcite , and in ref .@xcite the @xmath259- and @xmath245-dependence of its projection @xmath547 onto the @xmath70-direction has been investigated .however , since for @xmath548 the cutoff of type a was defined in @xcite by an ad hoc modification of the standard one - loop determinants it is not clear whether it can be derived from an action @xmath8 , except for the case @xmath314 @xcite .since a specification of @xmath8 is indispensable for the actual construction of @xmath3 , the status of the results derived in @xcite is somewhat unclear . in ref .@xcite we performed a comprehensive analysis of the fixed point properties using different cutoffs of type b , for which a @xmath8 is known to exist .in particular , we investigated the cutoff scheme dependence of various universal quantities of interest , both by looking at their dependence on the shape function @xmath245 and by comparing the `` type a '' and `` type b '' results . in this respect universal quantities are of special importance because , by definition , they are strictly cutoff scheme independent in the exact theory .any truncation leads to a scheme dependence of these quantities whose magnitude is a measure for the reliability of the truncation @xcite . typical examples of universal quantities are the critical exponents @xmath400 .the existence or nonexistence of a fixed point is also a universal , scheme independent feature , but its precise location in parameter space is scheme dependent .nevertheless it can be argued that , in @xmath73 , the product @xmath88 is universal @xcite while @xmath396 and @xmath545 separately are not . for later comparison with the @xmath1-truncation , let us briefly list some of the results we obtained in @xcite with the pure einstein - hilbert truncation : @xmath549 : both for type a and type b cutoffs the non - gaussian fixed point exists for all shape functions @xmath245 we considered .this result is highly nontrivial since in higher dimensions @xmath550 the fixed point exists for some but does not exist for other cutoffs @xcite .@xmath551 : while the position of the fixed point is scheme dependent , all cutoffs yield _ positive _ values of @xmath396 and @xmath545. a negative @xmath396 might be problematic for stability reasons , but there is no mechanism in the flow equation which would exclude it on general grounds .@xmath552 : for any cutoff employed , the non - gaussian fixed point is found to be uv attractive in both directions of the @xmath43-@xmath44plane . linearizing the flow equation according to eq .( [ gfp4 ] ) we obtain a pair of complex conjugate critical exponents @xmath553 with positive real part @xmath554 and imaginary parts @xmath555 . due to the positivity of @xmath554 , all trajectories in its basin of attractionhit the fixed point as @xmath2 is sent to infinity . because of the nonvanishing imaginary part @xmath556 the trajectories spiral into the fixed point for @xmath7 . solving the full , nonlinear flow equations @xcite shows that the asymptotic scaling region where the linearization is valid extends from @xmath557 down to about @xmath558 with the planck mass defined as @xmath559it is the regime above the planck scale where the asymptotic freedom of @xmath40 sets in .@xmath560 : the critical exponents are reasonably constant within about a factor of 2 . for the gauges @xmath314 and @xmath261 , for instance , they assume values in the ranges @xmath561 , @xmath562 and @xmath563 , @xmath564 , respectively. the universality properties of the product @xmath88 are much more impressive though . despite the rather strong scheme dependence of @xmath396 and @xmath545 separately ,their product exhibits almost no visible @xmath245-dependence .its value is @xmath565 for @xmath314 and @xmath566 for @xmath261 .the differences between the `` physical '' ( fixed point ) value of the gauge parameter , @xmath261 , and the technically more convenient @xmath314 are at the level of about 10 to 20 per - cent .+ the above results suggest that the uv attractive non - gaussian fixed point occuring in the einstein - hilbert truncation is very unlikely to be an artifact of this truncation but should rather be the projection of a fixed point in the exact theory .we interpreted them as nontrivial indications supporting the conjecture that 4-dimensional qeg is `` asymptotically safe '' in weinberg s sense .the actual justification of a truncation is that when one adds further terms to it its physical predictions do not change significantly any more . in order to test the stability of the einstein - hilbert truncation against the inclusion of other invariants we shall now reanalyze the non - gaussian fixed point in the generalized truncation ( [ trunc ] ) including the @xmath1-term .starting from the @xmath360-functions of the @xmath1-truncation , eqs .( [ del ] ) , ( [ deg ] ) and ( [ deb ] ) , we determine the location of the fixed point in @xmath43-@xmath70-@xmath104space and the linearized flow in its vicinity .then we investigate the residual cutoff scheme dependence of the associated universal quantities , and we compare our results to those obtained from the pure einstein - hilbert truncation . note that , contrary to the pure einstein - hilbert truncation , only a cutoff of type b is used in the context of the generalized truncation .therefore we omit the specification of the cutoff type when we refer to results obtained from the @xmath1-truncation . + * location of the fixed point @xmath567 * + in a first attempt at finding the non - gaussian fixed point in the @xmath1-truncation we neglect the cosmological constant and the coupling of the @xmath1-invariant .we approximate @xmath568 , @xmath569 , thereby projecting the renormalization group flow onto the one - dimensional space parametrized by @xmath70 . in this casethe non - gaussian fixed point is obtained as the nontrivial solution of @xmath570 .it is determined in appendix [ approxngfp ] with the result given by eq .( [ gstar1 ] ) . for any @xmath30 , this solution coincides precisely with the analogous approximate solution ( h2 ) of ref .@xcite with @xmath314 , obtained in the pure einstein - hilbert truncation . in order to get a numerical value for the fixed point we have to specify @xmath245 . inserting the exponential shape function with @xmath459 into eq .( [ gstar1 ] ) and setting @xmath73 leads to @xmath571 . assuming that for the combined @xmath43-@xmath70-@xmath492 system the numbers @xmath545 , @xmath396 and @xmath501 are of the same order of magnitude as @xmath396 above we expand the @xmath360-functions about @xmath572 and neglect terms of higher orders in the couplings . again in appendix [ approxngfp ]we determine the non - gaussian fixed point from the corresponding system of differential equations . inserting the shape function ( [ expcut ] ) and setting @xmath73 , we find @xmath573 .quite remarkably , for any cutoff @xmath545 and @xmath396 agree perfectly with the corresponding values obtained in @xcite by the same approximation applied to the pure einstein - hilbert truncation . in order to determine the _ exact _ position of the non - gaussian fixed point @xmath574 we have to resort to numerical methods .given a starting value for the fixed point , for instance one of the approximate solutions above , the program we use determines a numerical solution which is exact up to an arbitrary degree of accuracy . under the same conditions as above , i.e. for @xmath459 and @xmath73, we obtain @xmath575 in the pure einstein - hilbert truncation the corresponding coordinates of the fixed point are @xmath576 @xcite .obviously the values of @xmath545 and @xmath396 are almost the same in both cases . while @xmath545 and @xmath396 are of the same order of magnitude , we find that @xmath501 is significantly smaller than @xmath545 and @xmath396 . in order to test whether these properties of the fixed point coordinates are universal we study their scheme dependence by looking at the @xmath577- or @xmath578-dependence introduced via the one - parameter families of shape functions ( [ expshape ] ) or ( [ supp ] ) , respectively .here @xmath577 parametrizes the family of exponential shape functions ( [ expshape ] ) , while the shape parameter @xmath578 allows us to change the profile of the shape functions with compact support ( [ supp ] ) . as for the family of exponential shape functions ,we are forced to restrict our considerations to shape parameters @xmath579 .this is because for @xmath580 the numerical integrations are plagued by convergence problems .they are due to the fact that in @xmath73 some of the threshold functions appearing in @xmath361 , @xmath362 and @xmath363 diverge in the limit @xmath581 , see also @xcite . as for the family of shape functions with compact support , we have to restrict ourselves to @xmath582 for similar reasons . here@xmath583 approaches a sharp cutoff as @xmath584 , which introduces discontinuities into the integrands of the threshold functions @xmath585 and @xmath586 .already for @xmath587 the @xmath360-functions start to `` feel '' the sharp cutoff limit , which leads to convergence problems .= 7.9 cm = 5.2 cm ( a ) = 7.9 cm = 5.2 cm ( b ) = 7.9 cm = 5.2 cm ( a ) = 7.9 cm = 5.2 cm ( b ) as in the case of the pure einstein - hilbert truncation @xcite our results establish the existence of the non - gaussian fixed point in a wide range of @xmath577- and @xmath578-values . as expected , the position of the fixed point turns out to depend on @xmath577 or @xmath578 , i.e. on the cutoff scheme , but the crucial point is that it exists for any of the cutoffs employed . figs .[ plot1 ] and [ plot4 ] show its coordinates @xmath574 as well as the product @xmath88 for the shape functions ( [ expshape ] ) and ( [ supp ] ) , respectively .in fig [ plot1](a ) we plotted the various quantities in the range @xmath588 where the largest changes in @xmath545 and @xmath396 occur , but we calculated them for @xmath589 . for every shape parameter @xmath577 or @xmath578 , the values of @xmath545 and @xmath396 are almost the same as those obtained with the einstein - hilbert truncation @xcite .as a consequence , the product @xmath88 is again almost constant and its value differs only slighlty from the one in @xcite for the same gauge @xmath314 . both figs .[ plot1](a ) and [ plot4](a ) suggest the universal value @xmath566 while we obtained @xmath565 from the pure einstein - hilbert truncation .thus we may expect that our @xmath88-value is precise at the 10 to 20 per - cent level .presumably this degree of precision is the best we can achieve in the present calculation because we saw already that the error due to using @xmath314 instead of the `` correct '' @xmath261 leads to an uncertainty of the same size .furthermore , our results show that @xmath501 is always significantly smaller than @xmath396 and @xmath545 for both families of shape functions , which is quite remarkable . within the limited precision of our calculationthis means that in the three - dimensional @xmath43-@xmath70-@xmath104space the fixed point practically lies in the @xmath43-@xmath44plane with @xmath491 , i.e. on the parameter space of the pure einstein - hilbert truncation .it is also interesting to note that the scheme dependence of @xmath501 is unexpectedly small . as for the family of exponential shape functions( [ expshape ] ) , the function @xmath590 depicted in fig . [ plot1](b ) develops a plateau - like shape for not too small values of @xmath577 . employing the family of shape functions with compact support ,the scheme dependence of @xmath501 is even weaker. the function @xmath591 plotted in fig .[ plot4](b ) is almost constant in the range @xmath592 .moreover , the positions of the two plateaus are nearly identical . while fig .[ plot1](b ) suggests the value @xmath593 , we obtain @xmath594 from fig .[ plot4](b ) .this indicates that in @xmath73 dimensions also @xmath501 might be a universal quantity .+ * the linearized flow @xmath567 * + let us now analyze the critical behavior near the non - gaussian fixed point .quite remarkably , the non - gaussian fixed point of the @xmath1-truncation proves to be uv attractive in any of the three directions of @xmath43-@xmath70-@xmath104space , for all cutoffs used . the linearized flow in its vicinity is always governed by a pair of complex conjugate critical exponents @xmath595 with @xmath596 and a single real , positive critical exponent @xmath597 .( we define @xmath533 as the critical exponent with the positive imaginary part so that @xmath598 . ) the general solution to the linearized flow equations is obtained by taking the real part of eq .( [ gfp5 ] ) . introducing the rg time@xmath599 it may be written as @xmath600 { \rm re}\,v\nonumber\\ & & + \left[{\rm re}\,c\,\sin\left(\theta''\,t\right)-{\rm im}\,c \,\cos\left(\theta''\,t\right)\right]{\rm im}\,v\bigg\}\,e^{-\theta ' t } + c_3 v^3\,e^{-\theta_3 t}\end{aligned}\ ] ] with arbitrary complex @xmath601 and arbitrary real @xmath602 .furthermore , @xmath603 and @xmath604 are the right - eigenvectors of the stability matrix @xmath605 with eigenvalues @xmath606 and @xmath607 , respectively .obviously the conditions for uv stability are @xmath596 and @xmath597 .they are indeed satisfied for all cutoffs . as a consequence , all rg trajectories which reach its basin of attraction hit the fixed point as @xmath173is sent to infinity .the trajectories ( [ linflow2 ] ) comprise three independent normal modes with amplitudes proportional to @xmath608 , @xmath609 and @xmath602 , respectively .the first two are of the spiral type , the third one is a straight line .let us illustrate these features by means of an example .for the exponential shape function ( [ expshape ] ) with @xmath459 , for instance , we have @xmath610 .the corresponding stability matrix @xmath402 takes the form @xmath611 it leads to the pair of complex critical exponents @xmath553 with @xmath612 , @xmath613 , and to the real critical exponent @xmath614 . for the associated right - eigenvectors we find @xmath615 ( the vectors are normalized such that @xmath616 . ) in fig .[ plot2 ] we show a typical trajectory which has all three normal modes excited with equal strength @xmath617 , @xmath618 . all its way down from @xmath557 to about @xmath71 it is confined to a very thin box surrounding the @xmath477plane , i.e the parameter space of the einstein - hilbert truncation .in fact , the linearized flow is characterized by the following quite remarkable properties , independently of the cutoff .they all indicate that , close to the non - gaussian fixed point , the rg flow is rather well approximated by the pure einstein - hilbert truncation .= 7.9 cm = 5.2 cm ( a ) = 7.9 cm = 5.2 cm ( b ) * ( a ) * the @xmath492-components of @xmath619 and @xmath620 are very tiny . hence these two vectors span a plane which virtually coincides with the @xmath43-@xmath44subspace at @xmath491 , i.e. with the parameter space of the einstein - hilbert truncation . as a consequence , the @xmath608- and @xmath609- normal modes are essentially the same trajectories as the `` old '' normal modes already found without the @xmath1-term . also the corresponding @xmath554- and @xmath556-valuescoincide within the scheme dependence , see below . *( b ) * for all cutoffs employed , the new eigenvalue @xmath534 introduced by the @xmath1-term is significantly larger than @xmath554 , see below . when a trajectory approaches the fixed point from below @xmath621 , the `` old '' normal modes @xmath622 are proportional to @xmath623 , but the new one is proportional to @xmath624 , so that it decays much more quickly . for every trajectory running into the fixed point ,i.e. for every set of constants @xmath625 , we find therefore that , once @xmath173 is sufficiently large , the trajectory lies entirely in the @xmath619-@xmath626subspace , i.e. the @xmath477plane practically . due to the large value of @xmath534 ,the new scaling field is very `` relevant '' .however , when we start at the fixed point @xmath627 and lower @xmath173 it is only at the low energy scale @xmath558 @xmath628 that @xmath624 reaches unity , and only then , i.e. far away from the fixed point , the new scaling field starts growing rapidly . *( c ) * since the matrix @xmath387 is not symmetric its eigenvectors have no reason to be orthogonal .in fact , we find that @xmath604 lies almost in the @xmath619-@xmath626plane . for the angles between the eigenvectors given above we obtain @xmath629 , @xmath630 , @xmath631 .their sum is @xmath632 which confirms that @xmath619 , @xmath620 and @xmath604 are almost coplanar .this implies that when we lower @xmath173 and move away from the fixed point so that the @xmath604- scaling field starts growing , it is again predominantly the @xmath37- and @xmath38-invariants which get excited , but not @xmath89 in the first place . summarizing the three points abovewe can say that very close to the fixed point the rg flow seems to be essentially two - dimensional , and that this two - dimensional flow is well approximated by the rg equations of the einstein - hilbert truncation .+ * scheme dependence of the critical exponents ( @xmath73 ) * + as we pointed out already the critical exponents are universal in an exact treatment , but in a truncated parameter space a scheme dependence is expected to occur as an artifact of the truncation .we may use it to judge the quality of our truncation .also in this respect the @xmath1-truncation yields satisfactory results , which we display in figs .[ plot3 ] and [ plot5 ] . figs .[ plot3](a ) and [ plot5](a ) show the real and the imaginary part @xmath554 and @xmath556 of the complex conjugate pair @xmath553 while @xmath534 is depicted in figs.[plot3](b ) and [ plot5](b ) .the plots in fig .[ plot3 ] are based on the family of exponential shape functions ( [ expshape ] ) and those in fig .[ plot5 ] are obtained by employing the family of shape functions with compact support ( [ supp ] ) .they display the @xmath577- and the @xmath578-dependence of the critical exponents , respectively .= 7.9 cm = 5.2 cm ( a ) = 7.9 cm = 5.2 cm ( b ) = 7.9 cm = 5.2 cm ( a ) = 7.9 cm = 5.2 cm ( b ) as for the complex conjugate pair of critical exponents , the scheme dependence is of the same order of magnitude as in the case of the einstein - hilbert truncation @xcite .while the scheme dependence of @xmath556 is weaker than that found in @xcite we see that it is slightly larger for @xmath554 . forthe exponential shape functions with @xmath589 , @xmath554 and @xmath556 assume values in the ranges @xmath633 and @xmath634 , respectively .employing the shape functions with compact support leads to a weaker dependence on the shape parameter @xmath578 .however , the corresponding values @xmath635 and @xmath636 are in good agreement with those obtained with the exponential cutoffs .in fact , they all lie in the @xmath637- and @xmath638intervals given above .the average values of @xmath554 and @xmath556 are slightly larger than those obtained from the pure einstein - hilbert truncation .the difference between the corresponding average values is approximately 1 for both @xmath554 and @xmath556 .let us now come to the new critical exponent @xmath534 which was not present in the einstein - hilbert truncation .using the exponential shape functions ( [ expshape ] ) it suffers from relatively strong variations as the shape parameter @xmath577 is changed .it assumes values in the range @xmath639 . as compared to the exponential cutoffs , the cutoffs with compact support lead to a much weaker scheme dependence . for @xmath640$ ] we have @xmath641 . however , the results obtained with the two families of shape functions agree within the scheme dependence .moreover , @xmath534 is always systematically larger than @xmath554 ( and @xmath556 ) with both families of cutoffs . asa consequence , the hierarchy of critical exponents which was mentioned in ( b ) above and which squeezes the trajectories into a thin box is a universal feature .obviously the critical exponents , in particular @xmath534 , exhibit a much stronger scheme dependence than @xmath88 .this is most probably due to neglecting further relevant operators in the truncation so that the @xmath387-matrix we are diagonalizing is still too small .+ * in @xmath642 dimensions * + the above results and their mutual consistency strongly suggest that 4-dimensional quantum einstein gravity indeed possesses a rg fixed point with precisely the properties needed for its nonperturbative renormalizability or `` asymptotic safety '' .however , with the present approach it is clearly not possible to determine the dimensionality @xmath643 of the uv critical hypersurface , which coincides with the number of invariants relevant at the non - gaussian fixed point . according to the canonical dimensional analysis ,the @xmath644va@xmath645-invariants in 4 dimensions are _ classically _ marginal for @xmath646 and irrelevant for @xmath647 .the results for @xmath534 indicate that there are large _ nonclassical _ contributions so that there might be relevant operators perhaps even beyond @xmath646 .however , as it is hardly conceivable that the quantum effects change the signs of arbitrarily large ( negative ) classical scaling dimensions , @xmath643 should be finite @xcite .a first confirmation of this picture comes from our @xmath1-calculation in @xmath58 where the dimensional count is shifted by two units . in this casewe find indeed that the third scaling field is _ irrelevant _ for any cutoff employed , @xmath648 . for our analysis of the @xmath1-truncation in @xmath58 dimensions with @xmath649 we had to resort to numerical methods . using the @xmath650-expansion we calculated the fixed point coordinates and the critical exponents for selected values of the shape parameter @xmath577 . for all quantities only the leading nontrivial order of the @xmath650-expansion was retained . in table 1we present the corresponding numerical results .@xmath577 & @xmath651 ) & @xmath652 & @xmath653 & @xmath654 & @xmath655 & @xmath656 + 1 & @xmath657 & @xmath658 & @xmath659 & @xmath660 & @xmath661 & @xmath662 + 5 & @xmath663 & @xmath664 & @xmath665 & @xmath660 & @xmath666 & @xmath667 + 10 & @xmath668 & @xmath669 & @xmath670 & @xmath660 & @xmath666 & @xmath671 + for all cutoffs used we obtain three _ real _ critical exponents , the first two are positive and the third is negative .thus , the corresponding @xmath604-direction is uv repulsive .this suggests that the dimensionality of @xmath57 could be as small as @xmath672 , but this is not a proof , of course .if so , the quantum theory would be characterized by only two free parameters , the renormalized newton constant @xmath673 and the renormalized cosmological constant @xmath674 , for instance .let us now compare the results to those from the einstein - hilbert truncation @xcite .the @xmath43- and @xmath70-coordinates of the fixed point and the critical exponents @xmath533 and @xmath535 are found to be similar to those in @xcite. however , in the einstein - hilbert truncation the leading - order results @xmath675 and @xmath676 are scheme independent , which is not quite true for the results above .both truncations agree on @xmath677 . + * summary : * + our main results concerning the non - gaussian fixed point in the @xmath1-truncation are : @xmath678 : the fixed point is found to exists for all cutoffs used . this result is highly nontrivial since the example of the gaussian fixed point clearly shows that a fixed point of the einstein - hilbert truncation does not necessarily generalize to a fixed point of the @xmath1-truncation . for every shape parameterthe fixed point practically lies on the @xmath43-@xmath44plane , and its position almost exactly coincides with that from the einstein - hilbert truncation .@xmath679 : the fixed point is uv attractive in any of the three directions of the @xmath43-@xmath70-@xmath104space for all cutoffs employed . the linearized flow in its vicinity is always governed by a pair of complex conjugate critical exponents @xmath595 with @xmath596 and a single real , positive critical exponent @xmath597 .it is essentially two - dimensional , and this two - dimensional flow is well descibed by the rg equations of the einstein - hilbert truncation .@xmath680 : the scheme dependence of the critical exponents and of the product @xmath88 is of the same order of magnitude as in the case of the einstein - hilbert truncation . while the scheme dependence of @xmath556 is weaker than in the case of the einstein - hilbert truncation we find that it is slightly larger for @xmath554 .the exponent @xmath534 shows a relatively strong dependence on the cutoff .the product @xmath88 again exhibits an impressively weak scheme dependence .@xmath681 : the dimensionality @xmath643 of the uv critical hypersurface can not be determined within the present approach .however , the results from our @xmath1-calculation in @xmath642 dimensions suggest that @xmath643 should be finite also in 4 dimensions . on the basis of the above results we believe that the non - gaussian fixed point occuring in the einstein - hilbert truncation is very unlikely to be an artifact of this truncation butrather should be the projection of a fixed point in the exact theory .we demonstrated explicitly that the fixed point and all its qualitative properties are stable against the inclusion of a further invariant in the truncation .these results strongly support the hypothesis that 4-dimensional qeg is indeed nonperturbatively renormalizable .[ 3s6 ] it is a well known problem that in @xmath417 dimensions the euclidean einstein - hilbert action @xmath682=\frac{1}{16\pi\bar{g}}\int d^dx\,\sqrt{g}\left\{-r(g ) + 2\bar{\lambda}\right\}\end{aligned}\ ] ] is not bounded below . in fact , decomposing the metric as @xmath683 where @xmath33 is a fixed reference metric we obtain @xmath684=\frac{1}{16\pi \bar{g}}\int d^dx\ , \sqrt{\bar{g}}\,e^{(d-2)\chi}\,\left[-\bar{r}+2\bar{\lambda}\,e^{2\chi } -(d-1)(d-2)\,\bar{g}^{\mu\nu}\ , \left(\bar{d}_\mu\chi\right)\left(\bar{d}_\nu\chi\right)\right]\,.\end{aligned}\ ] ] this shows that @xmath685 can become arbitrarily negative if the conformal factor @xmath686 varies rapidly enough so that @xmath687 is large. therefore it seems difficult to define a path integral @xmath688 for euclidean quantum gravity .the situation improves by including the term @xmath689 with a positive coefficient since the resulting action is _ bounded below _ @xcite .while the einstein - hilbert term @xmath690 leads to a negative contribution to the kinetic term of the conformal factor , which dominates at small momenta , the @xmath1-term gives rise to a positive contribution dominating at large momenta . as a consequence , both the truncated action functional @xmath253 $ ] of eq .( [ trunc ] ) and the bare action @xmath691= \gamma_{\widehat{k}\rightarrow\infty}[g , g]$ ] possess an absolute minimum .moreover , rewriting the truncation ansatz ( [ trunc ] ) with @xmath42 as @xmath692=\int d^dx\,\sqrt{g}\left\{\bar{\beta}_k\left(r-\frac{z_{nk } \kappa^2}{\bar{\beta}_k}\right)^2+z_{nk}\kappa^2\left(4\bar{\lambda}_k -\frac{z_{nk}\kappa^2 } { \bar{\beta}_k}\right)\right\}\end{aligned}\ ] ] one can easily determine a sufficient condition for a manifestly _ positive _ action @xmath693>0 $ ] . in terms of the dimensionless couplings it reads : @xmath694 , @xmath695 , and @xmath696 at the level of the flow equation , @xmath3 appears on the rhs in terms of its hessian @xmath20 to which the cutoff operator @xmath21 is adapted by the rule ( [ rule ] ) .thus , only if @xmath20 is a positive definite operator we can obtain a cutoff which leads to a `` correct '' mode suppression . since we expect the @xmath1-truncation anyhow to be reliable only for large @xmath2 , it is actually sufficient if @xmath20 and @xmath21 are positive definite for sufficiently large momenta @xmath697 .the reason is that , due to the factor @xmath698 which emphasizes the region @xmath699 , the traces on the rhs of the rg equation ( [ truncflow ] ) receive the dominant contributions from modes whose @xmath23 is close to @xmath72 .in general @xmath312 $ ] depends on both @xmath32 and the background metric @xmath33 .here we concentrate on @xmath700\equiv\gamma_k^{(2)}$ ] with the two metrics identified .furthermore , we assume that @xmath267 is the metric of a @xmath30-sphere with radius @xmath274 since our projection technique requires these backgrounds only . in this casethe eigenvalues @xmath701 depend on the discrete quantum number @xmath702 .the explicit expressions for @xmath703 are tabulated in appendix [ harm ] .they are strictly monotonically increasing functions of @xmath702 with @xmath704 . in the followingwe show that the operator @xmath20 with @xmath2 very large indeed becomes positive definite if it is restricted to the subspace spanned by the @xmath705-eigenfunctions with sufficiently large eigenvalues , certain assumptions on the couplings being made .the spherical harmonics @xmath282 , @xmath283 , and @xmath284 with @xmath702 larger than a certain minimum value @xmath706 provide a basis of this subspace .we shall concentrate on the conditions implied by the leading large-@xmath702 behavior .the hessian @xmath700 $ ] as given by the quadratic form ( [ gravquad ] ) is a symmetric blockdiagonal matrix .therefore , according to the jacobi criterion , the condition for positivity takes the simple form @xmath707 , @xmath708 , @xmath709 , and @xmath710 . for sufficiently large values of @xmath702the leading @xmath702-powers of @xmath703 are the dominating contributions to the entries of @xmath20 in eq .( [ gravquad ] ) so that , in this limit , the above condition boils down to @xmath711 for nonnegative values of the gauge parameter , @xmath712 , this leads to the following restrictions on the dimensionless couplings : @xmath713 in the uv fixed point regime of the @xmath73-dimensional case we have @xmath714 and @xmath715 with @xmath716 .hence , close to the non - gaussian fixed point , the first two conditions of eq .( [ rs ] ) are obviously satisfied .furthermore , the third condition then takes the form @xmath717 . for @xmath718fixed this condition is satisfied as well provided @xmath2 is sufficiently large .thus , for @xmath2 large and on modes with large eigenvalues of @xmath705 , the restricted operator @xmath20 is positive .the cutoff should have the desired suppression properties therefore .the above argument treats @xmath718 as a constant parameter .recalling the derivation of the projected flow equation where we compared powers of the radius @xmath719 it is indeed clear that in this context @xmath274 and @xmath718 should be regarded as fixed , @xmath2-independent quantities .it is instructive to look also at the operator @xmath720 $ ] where @xmath721 is the @xmath2-dependent `` on - shell '' @xmath273-metric which solves the equation of motion @xmath722 for @xmath267 .the difference to the situation discussed before is that @xmath718 is a function of @xmath2 now , to be computed from @xmath721 .( the operator @xmath723 $ ] would appear in a standard one - loop ( saddle point ) calculation based upon the `` classical '' action @xmath3 . ) for the truncated action functional @xmath3 of eq .( [ trunc ] ) with @xmath42 the field equation takes the form @xmath724 + \bar{\beta}_k\left[-\left(g_{\mu\nu}+r_{\mu\nu}\right)r+2d_\mu d_\nu r -2g_{\mu\nu}d^2r\right]=0\end{aligned}\ ] ] with @xmath725 the einstein tensor . precisely for @xmath73 , the maximally symmetric solutions to eq .( [ fieldeqn ] ) satisfy einstein s equation @xmath726 , they are not affected by the @xmath1-term . inserting the contracted equation @xmath727 intothe third condition of ( [ rs ] ) leads to @xmath728 remarkably , this condition is satisfied precisely if @xmath693 $ ] is _ not _ a manifestly positive functional of @xmath32 , as follows from ( [ posit ] ) .this implies that for @xmath73 the @xmath76 solution of eq .( [ fieldeqn ] ) can not correspond to the absolute minimum of @xmath693 $ ] if this functional is manifestly positive . in the uv fixed point regime, the condition ( [ oscond ] ) becomes @xmath729 .for all cutoffs employed we found that @xmath730 so that this condition is indeed satisfied .it is reassuring that also , upon inserting the @xmath76 solution of eq .( [ fieldeqn ] ) , the hessian @xmath20 becomes a positive operator for sufficiently large values of @xmath702 and @xmath2 , independently of the cutoff .furthermore , the concomitant violation of ( [ posit ] ) implies that in the vicinity of the fixed point the functional @xmath693 $ ] is bounded below but not positive . by adding an appropriate constant it is trivial though to turn it into a manifestly positive functional .the cutoff @xmath154 is expected to be positive definite under the same conditions as found for the hessian @xmath20 in the previous subsection . in order to obtain more quantitative information about @xmath706 and the momentum regime where @xmath8 is positive we continue our analysis with an explicit investigation of the cutoff operator @xmath21 . for simplicitywe again restrict our considerations to the most interesting case of @xmath73 and to spherical backgrounds . after setting @xmath731 and @xmath732 , and inserting the eigenvalues of the covariant laplacians , the entries of the cutoff matrix ( [ cutoff ] )assume the form @xmath733^{\frac{3}{2}}\nonumber\\ & & \times\sqrt{\lambda_l(4,0)/k^2+r^{(0 ) } ( \lambda_l(4,0)/k^2)-\frac{r}{3k^2}}\nonumber\\ & & -\left[\lambda_l(4,0)/k^2\right]^{\frac{3}{2}}\sqrt{\lambda_l(4,0)/k^2 -\frac{r}{3k^2}}\bigg]\bigg\}\;,\\ \left({\cal r}_k\right)_{\bar{\phi}\bar{\phi}}&= & \frac{k^4r^{(0)}(\lambda_l(4,0)/k^2 ) } { 32\pi\,g_k}\left\ { 36\pi\,g_k\beta_k \left(2\lambda_l(4,0)/k^2 + r^{(0)}(\lambda_l(4,0)/k^2)\right)-\frac{1}{4}\right\}\;.\nonumber\end{aligned}\ ] ] as compared to eq .( [ cutoff ] ) which was written in terms of the dimensionful quantities @xmath734 , @xmath41 , and @xmath251 , we switched here to a description in terms of the dimensionless couplings . in analogy with the hessian @xmath186 the condition for the cutoff matrix @xmath21 to be positive definite reads : @xmath735 , @xmath736 , @xmath737 , and @xmath738 .these conditions indeed reproduce the restrictions on the couplings obtained from @xmath20 for sufficiently large momenta @xmath701 .provided that @xmath712 , they take the form @xmath694 , @xmath695 and @xmath739 , which coincides with eq .( [ rs ] ) . given an arbitrary set of parameters @xmath740 satisfying these three inequalities, we have @xmath735 and @xmath736 ( and also @xmath741 ) for _ any _ allowed value of @xmath702 .this is not the case for the other two conditions which stem from the scalar sector of the cutoff .clearly @xmath742 of eq .( [ posdef ] ) can assume negative values for sufficiently small values of @xmath702 , provided @xmath47 , @xmath346 and @xmath743 are small enough .since @xmath741 , a negative @xmath742 implies that also @xmath744 .the @xmath702-values for which @xmath737 satisfy @xmath745 in appendix [ estimation ] we derive a similar inequality involving the @xmath309-@xmath179 cross term .there we find that @xmath746 at least for all values of @xmath702 satisfying @xmath747 both inequalities , ( [ phiphicond ] ) and ( [ phisigma ] ) , depend on @xmath748 , on @xmath749 with @xmath750 , and on the shape function @xmath245 .given a specific set @xmath751 with a certain scalar eigenmode @xmath284 , they tell us whether the contribution from this mode is suppressed correctly or not . the more restrictive inequality ( [ phisigma ] ) applies to the scalar eigenmodes @xmath752 with @xmath753 , while ( [ phiphicond ] ) can be used for the constant mode @xmath315 and the pckv s @xmath754 only .let us now focus on rg trajectories which run into the non - gaussian fixed point as @xmath7 .furthermore , we assume that @xmath718 is either kept fixed or that @xmath755 with a constant @xmath756 .then , for large enough values of @xmath2 , we have @xmath757 , @xmath758 and @xmath759 so that the conditions ( [ rs ] ) for the positivity of @xmath20 and @xmath21 are satisfied .moreover , the rhs of ( [ phiphicond ] ) and ( [ phisigma ] ) may be expressed as @xmath760 and @xmath761 , respectively .now we are in a position to determine the @xmath702-regime for which the cutoff is manifestly positive definite . using the family of exponential shape functions ( [ expshape ] ) with @xmath589, a numerical analysis reveals that _ any _ value of the ratio @xmath762 satisfies ( [ phiphicond ] ) or ( [ phisigma ] ) provided @xmath763 or @xmath764 , respectively .hence , under the above conditions , all cutoffs employing an exponential shape function with @xmath764 are manifestly positive definite for _ all _ momenta , i.e. for all quantum numbers @xmath702 .this is a rather intriguing result .it might indicate that cutoffs with @xmath765 are particularly reliable .conversely , for any @xmath766 there exists a specific value @xmath767 such that all @xmath768 with @xmath769 violate ( [ phisigma ] ) .furthermore , there exists a specific @xmath770 for any @xmath771 such that any @xmath772 leads to a violation of ( [ phiphicond ] ) . in fig .[ plota ] we show @xmath767 and @xmath770 in the ranges @xmath773 and @xmath774 , respectively .it is important to note that @xmath775 and @xmath776 for any value of @xmath577 considered .this implies that in the uv fixed point regime the cutoff has the desired suppression properties for all modes with momenta ranging from infinity down to values well below @xmath2 .= 7.9 cm = 5.2 cm ( a ) = 7.9 cm = 5.2 cm ( b ) to complete the analysis let us study the inequalities ( [ phiphicond ] ) and ( [ phisigma ] ) also at the spherically symmetric stationary point @xmath721 of @xmath693 $ ] which we discussed in the previous subsection . in the vicinity of the fixed pointthe on - shell value of the curvature is @xmath777 .hence , we obtain from ( [ phiphicond ] ) and ( [ phisigma ] ) , respectively , @xmath778\equiv\frac{2l(l+3)\lambda_*}{3}+r^{(0)}(l(l+3)\lambda_*/3)- \frac{1}{144\pi\,g_*\beta_*}>0\;,\;\;l=0,1\end{aligned}\ ] ] and @xmath779\equiv\frac{2l(l+3)\lambda_*}{3}+r^{(0)}(l(l+3)\lambda_*/3)- \frac{1}{96\pi\,g_*\beta_*}>0\;,\;\;l\ge 2\;.\end{aligned}\ ] ] the first inequality stems from the scalar eigenmodes @xmath284 with @xmath780 , and the second from those with @xmath753 . both ( [ phiphi2 ] ) and ( [ phisigma2 ] ) depend on @xmath702 and @xmath245 .again we restrict our investigation to the family of exponential shape functions with @xmath589 .then the lhs of ( [ phiphi2 ] ) and ( [ phisigma2 ] ) are functions of @xmath577 and @xmath702 : @xmath781\equiv f_1(s , l)$ ] , @xmath782 \equiv f_0(s , l)$ ] .for @xmath753 we have @xmath783 independently of the shape parameter .numerically we find that @xmath784 is always positive .hence , _ any _ momentum with @xmath753 satisfies the condition ( [ phisigma2 ] ) for all values of @xmath577 considered .furthermore , our numerical analysis shows that also @xmath785 for all cutoffs employed .however , @xmath786 is not always positive .we obtain @xmath787 for @xmath763 and @xmath788 for @xmath771 .our results are illustrated in fig .[ plotb ] .= 7.9 cm = 5.2 cm ( a ) = 7.9 cm = 5.2 cm ( b ) the above results have to be interpreted as follows .assume that @xmath2 lies in the uv scaling regime , and consider the cutoff operator at the spherically symmetric stationary point , @xmath789 $ ] .then this operator is strictly positive on the space spanned by _ all _ spherical harmonics and would correctly suppress all modes in a path integral containing this `` on - shell '' cutoff , provided we choose an exponential shape function with @xmath763 . for @xmath771 , only the contributions from the constant mode with @xmath790 are not suppressed correctly . for any other mode the cutoff term is positive even in this case . * to summarize : * in this section we found that at least in the asymptotic domain relevant in our investigation of the uv fixed point the cutoff which is adapted to the @xmath1-truncation is positive definite and therefore has all the required mode suppression properties .no conformal factor problem and no growing exponentials produced by @xmath791 are encountered .[ conclusio ] in this paper we evaluated the exact rg equation of quantum einstein gravity in a truncation which generalizes the einstein - hilbert approximation used so far by the inclusion of a higher - derivative term .we derived the beta - functions of the resulting @xmath43-@xmath70-@xmath104system which turned out to be by far more complicated than the old @xmath43-@xmath44system .we used these beta - functions in order to investigate how the two fixed points known to exist in the einstein - hilbert truncation manifest themselves in the enlarged theory space .we found that the gaussian fixed point of the einstein - hilbert truncation does not generalize to a corresponding fixed point of the @xmath1-truncation .nevertheless , the 2-dimensional projection of the @xmath43-@xmath70-@xmath104flow onto the @xmath43-@xmath44plane at @xmath491 , near the origin @xmath792 , is well approximated by the flow resulting from the einstein - hilbert truncation .the projected flow does indeed have a fixed point at @xmath793 . in the einstein - hilbert truncationthere exists a distinguished rg trajectory , the `` separatrix '' @xcite , which gives rise to a vanishing renormalized cosmological constant , @xmath794 . in @xmath73 , it turned out that this trajectory possesses a 3-dimensional `` lift '' which is characterized by a logarithmic running of the @xmath1-coupling @xmath346 . for @xmath506its running is power - like , and there exists a `` quasi - gaussian '' fixed point at @xmath408 , @xmath795 .this picture puts the older perturbative calculations in @xmath1-gravity into a broader context . quite differently , the non - gaussian fixed point @xmath796 implied by the einstein - hilbert truncation does `` lift '' to a corresponding fixed point of the 3-dimensional flow , with a tiny but nonzero third component @xmath797 .it is uv attractive in all 3 directions of @xmath43-@xmath70-@xmath104space .we demonstrated in detail that close to the fixed point the flow on the extended theory space is essentially 2-dimensional , and that the 2-dimensional projected flow is very well approximated by the einstein - hilbert flow .for the @xmath545- and @xmath396-coordinates both truncations yield virtually identical values , and the same is true for the critical exponents pertaining to the 2-dimensional subspace . for universal quantitiesthe differences between the two truncations are typically smaller than their weak residual scheme dependence .this stability of the einstein - hilbert truncation against the inclusion of a further invariant , together with the other pieces of evidence which we summarized in sections [ 3s5.3.1 ] and [ 3s5.3.2 ] strongly support the conjecture that this approximation is at least qualitatively reliable in the uv .hence it appears increasingly unlikely that the very existence of the non - gaussian fixed point is an artifact of the truncation .we believe that qeg has indeed very good chances of being nonperturbatively renormalizable .a notorious difficulty of euclidean quantum gravity is the conformal factor problem . in the exact rg approach, it appears in the einstein - hilbert truncation , but not in the @xmath1-truncation provided @xmath2 is large enough .when this complication occurs the construction of an appropriate cutoff operator is rather subtle .however , it was possible to show that our investigation of the non - gaussian fixed point in the @xmath1-truncation is not affected by this problem , and that a straightforward positive definite cutoff can be employed .the numerical agreement of the results with those from the einstein - hilbert truncation indicates that the rule for constructing an adapted cutoff in presence of the conformal factor problem which was proposed in @xcite ( `` @xmath207-rule '' ) should indeed be correct .+ acknowledgments : we would like to thank c. wetterich for many helpful discussions .we are also grateful to a. ashtekar , a. bonanno , w. dittrich , p. forgcs , h. gies , e. gozzi , d. litim , m. niedermaier , h. b. nielsen , r. percacci , j. b. pitts , m. salmhofer , f. saueressig , l. smolin , t. thiemann , and g. veneziano for interesting conversations and communications .[ eval ] in this section we present several rather lengthy calculations needed for the discussion of the @xmath1-truncation in section [ @xmath1-trunc ] . in the following , all calculations are performed with @xmath267 where @xmath33 is assumed to correspond to a @xmath273 background . for simplicitythe bars are omitted from the metric , the curvature and the operators .in section [ @xmath1-trunc ] we derived explicit expressions for the kinetic operators @xmath800 and @xmath801 .they may be represented as matrix differential operators acting on the column vectors @xmath802 and @xmath803 , respectively . in this representationthey take the form @xmath804= \left(\begin{array}{cccc } \left(\widetilde{\gamma}_k^{(2)}[g , g]\right)_{\bar{h}^t\bar{h}^t } & 0 & 0 & 0_{1\times 2}\\0 & \left(\widetilde{\gamma}_k^{(2)}[g , g]\right)_{\bar{\xi } \bar{\xi } } & 0 & 0_{1\times 2}\\ 0 & 0 & \left(\widetilde{\gamma}_k^{(2)}[g , g]\right)_{\bar{\phi}_0 \bar{\phi}_0 } & 0_{1\times 2}\\ 0_{2\times 1 } & 0_{2\times 1 } & 0_{2\times 1 } & { \cal q}_k \end{array}\right)\end{aligned}\ ] ] and @xmath805= \left(\begin{array}{cccc } 0 & \left(\widetilde{s}_{\rm gh}^{(2)}[g , g]\right)_{\bar{v}^tv^t } & 0 & 0\\ \left(\widetilde{s}_{\rm gh}^{(2)}[g , g]\right)_{v^t\bar{v}^t } & 0 & 0 & 0\\ 0 & 0 & 0 & \left(\widetilde{s}_{\rm gh}^{(2)}[g , g]\right)_{\bar{\varrho } \varrho}\\ 0 & 0 & \left(\widetilde{s}_{\rm gh}^{(2)}[g , g]\right)_{\varrho\bar{\varrho } } & 0 \end{array}\right ) \end{aligned}\ ] ] where @xmath806\right)_{\bar{\sigma } \bar{\sigma } } & \left(\widetilde{\gamma}_k^{(2)}[g , g ] \right)_{\bar{\phi}_1\bar{\sigma}}\\ \left(\widetilde{\gamma}_k^{(2)}[g , g]\right)_{\bar{\phi}_1\bar{\sigma } } & \left(\widetilde{\gamma}_k^{(2)}[g , g]\right)_{\bar{\phi}\bar{\phi } } \end{array}\right)\;.\end{aligned}\ ] ] the entries of these matrices are given in eq .( [ entries ] ) . on the rhs of the flow equation ( [ truncflow ] ) the inverse operators @xmath807^{-1}$ ] and @xmath808^{-1}$ ]appear which we determine in the following .at this point it is important to note that , because of the maximally symmetric background , all covariant derivatives contained in the operators ( [ 83 ] ) and ( [ 84 ] ) appear as covariant laplacians , and that the various entries are @xmath768-independent otherwise .this implies that these entries are _ commuting _ differential operators which allows for particularly simple manipulations .therefore it is not difficult to verify that the inverse operators assume the form @xmath809\right)^{-1}=}\nonumber\\ & & \left(\begin{array}{cccc } \left[\left(\widetilde{\gamma}_k^{(2)}[g , g]\right)_{\bar{h}^t\bar{h}^t } \right]^{-1 } & 0 & 0 & 0_{1\times 2}\\ 0 & \left[\left(\widetilde{\gamma}_k^{(2)}[g , g]\right)_{\bar{\xi}\bar{\xi } } \right]^{-1 } & 0 & 0_{1\times 2}\\ 0 & 0 & \left[\left(\widetilde{\gamma}_k^{(2)}[g , g]\right)_{\bar{\phi}_0 \bar{\phi}_0}\right]^{-1 } & 0_{1\times 2}\\ 0_{2\times 1 } & 0_{2\times 1 } & 0_{2\times 1 } & { \cal q}_k^{-1 } \end{array}\right)\end{aligned}\ ] ] and @xmath810\right)^{-1}=\\ & & \!\!\!\!\!\left(\begin{array}{cccc } 0 & \left[\left(\widetilde{s}_{\rm gh}^{(2)}[g , g]\right)_{\bar{v}^tv^t } \right]^{-1 } & 0 & 0\\ \left[\left(\widetilde{s}_{\rm gh}^{(2)}[g , g]\right)_{v^t\bar{v}^t } \right]^{-1 } & 0 & 0 & 0\\ 0 & 0 & 0 & \left[\left(\widetilde{s}_{\rm gh}^{(2)}[g , g]\right ) _ { \bar{\varrho}\varrho}\right]^{-1}\\ 0 & 0 & \left[\left(\widetilde{s}_{\rm gh}^{(2)}[g , g]\right)_{\varrho \bar{\varrho}}\right]^{-1 } & 0 \end{array}\right)\nonumber\end{aligned}\ ] ] with @xmath811\right)_{\bar{\sigma } \bar{\sigma } } \left(\widetilde{\gamma}_k^{(2)}[g , g]\right)_{\bar{\phi}_1\bar{\phi}_1 } -\left(\widetilde{\gamma}_k^{(2)}[g , g]\right)^2_{\bar{\phi}_1\bar{\sigma } } \right]^{-1}\nonumber\\ & & \times\left(\begin{array}{cc } \left(\widetilde{\gamma}_k^{(2)}[g , g]\right)_{\bar{\phi}_1\bar{\phi}_1 } & -\left(\widetilde{\gamma}_k^{(2)}[g , g ] \right)_{\bar{\phi}_1\bar{\sigma}}\\ -\left(\widetilde{\gamma}_k^{(2)}[g , g]\right)_{\bar{\phi}_1\bar{\sigma } } & \left(\widetilde{\gamma}_k^{(2)}[g , g]\right)_{\bar{\sigma}\bar{\sigma } } \end{array}\right)\;.\end{aligned}\ ] ] inserting these expressions into the rhs of the flow equation ( [ truncflow ] ) leads to @xmath812 + { \cal r}_k[g]\right)_{\zeta\zeta}\right]^{-1 } \partial_t\left({\cal r}_k[g]\right)_{\zeta\zeta}\right ] } \nonumber\\ & & + \frac{1}{2}{\rm tr}'\left [ \left\{\left(\widetilde{\gamma}_k^{(2)}[g , g]\right)_{\bar{\sigma}\bar{\sigma } } \left(\widetilde{\gamma}_k^{(2)}[g , g]\right)_{\bar{\phi}_1\bar{\phi}_1 } -\left(\widetilde{\gamma}_k^{(2)}[g , g]\right)^2_{\bar{\phi}_1\bar{\sigma } } \right\}^{-1}\right.\nonumber\\ & & \times\left\{\left(\gamma_k^{(2)}[g , g ] + { \cal r}_k[g]\right)_{\bar{\sigma}\bar{\sigma } } \partial_t\left({\cal r}_k[g]\right)_{\bar{\phi}_1\bar{\phi}_1 } + \left(\gamma_k^{(2)}[g , g]+{\cal r}_k[g]\right)_{\bar{\phi}_1\bar{\phi}_1 } \partial_t\left({\cal r}_k[g]\right)_{\bar{\sigma } \bar{\sigma}}\right.\nonumber\\ & & \left.\left.-2\left(\gamma_k^{(2)}[g , g]+{\cal r}_k[g ] \right)_{\bar{\phi}_1\bar{\sigma } } \partial_t\left({\cal r}_k[g]\right)_{\bar{\phi}_1\bar{\sigma } } \right\}\right]\nonumber\\ & & -{\rm tr}'\left[\sum\limits_{\psi\in\{v^t,\varrho\ } } \left[\left(s_{\rm gh}^{(2)}[g , g ] + { \cal r}_k[g]\right)_{\bar{\psi}\psi}\right]^{-1 } \partial_t\left({\cal r}_k[g]\right)_{\bar{\psi}\psi}\right]\end{aligned}\ ] ] where we used the relations @xmath813^{\mu x}}_{\nu y } & = & -{\left[\left(s_{\rm gh}^{(2)}\right)_{v^t\bar{v}^t } \right]_{\nu y}}^{\mu x } = \frac{1}{\sqrt{g(y)}}\frac{\delta}{\delta v^{t\nu}(y ) } \frac{1}{\sqrt{g(x)}}\frac{\delta s_{\rm gh}}{\delta\bar{v}^t_\mu(x ) } \nonumber\\ { \left[\left(s_{\rm gh}^{(2)}\right)_{\bar{\varrho}\varrho } \right]^x}_y & = & -{\left[\left(s_{\rm gh}^{(2)}\right)_{\varrho\bar{\varrho } } \right]_y}^x = \frac{1}{\sqrt{g(y)}}\frac{\delta}{\delta\varrho(y ) } \frac{1}{\sqrt{g(x)}}\frac{\delta s_{\rm gh}}{\delta\bar{\varrho}(x)}\;.\end{aligned}\ ] ] the trace of the @xmath296-term appearing in the first line of eq .( [ 061 ] ) may be evaluated easily since only the scalar eigenmodes @xmath814 and @xmath815 contribute .we obtain @xmath816 + { \cal r}_k[g]\right)_{\phi_0\phi_0}\right]^{-1 } \partial_t\left({\cal r}_k[g]\right)_{\phi_0\phi_0}\right]}\nonumber\\ & = & \frac{1}{2z_{nk}\kappa^2}\sum\limits_{l=0}^{1}\sum\limits_{m=1}^{d_l(d,0 ) } \int d^dx\,\sqrt{g(x)}\,t^{lm}(x ) \bigg[c_{s2}(d,\alpha)\,c_{s1}(d,\alpha)\,\big(p_k+a_{s1}(d,\alpha)\,r \nonumber\\ & & + b_{s1}(d,\alpha)\,\bar{\lambda}_k\big ) + ( z_{nk}\kappa^2)^{-1}\bar{\beta}_k\big(h_s(d)\,p_k^2 + g_{s2}(d)\,r\,p_k+g_{s3}(d)\,r^2\big)\bigg]^{-1 } \nonumber\\ & & \times\partial_t\bigg [ c_{s2}(d,\alpha)\,c_{s1}(d,\alpha)\,z_{nk}\kappa^2 \,k^2r^{(0)}(-d^2/k^2 ) + \bar{\beta}_k\big(h_s(d)\,\left(-2d^2\,r^{(0)}(-d^2/k^2)\right.\nonumber\\ & & + \left.k^4r^{(0)}(-d^2/k^2)^2\right ) + g_{s2}(d)\,r\,k^2r^{(0)}(-d^2/k^2)\big)\bigg ] t^{lm}(x)\nonumber\\ & = & \frac{1}{2z_{nk}\kappa^2}\sum\limits_{l=0}^{1}\bigg [ d_l(d,0)\bigg\{c_{s2}(d,\alpha)\,c_{s1}(d,\alpha)\,\big(\lambda_l(d,0)+ k^2r^{(0)}(\lambda_l(d,0)/k^2)\nonumber\\ & & + a_{s1}(d,\alpha)\,r + b_{s1}(d,\alpha)\,\bar{\lambda}_k\big)\nonumber\\ & & + ( z_{nk}\kappa^2)^{-1}\bar{\beta}_k\big(h_s(d)\,\left(\lambda_l(d,0)+ k^2r^{(0)}(\lambda_l(d,0)/k^2)\right)^2\nonumber\\ & & + g_{s2}(d)\,r\left(\lambda_l(d,0)+ k^2r^{(0)}(\lambda_l(d,0)/k^2)\right)+g_{s3}(d)\,r^2\big)\bigg\}^{-1 } \nonumber\\ & & \times\partial_t\bigg\ { c_{s2}(d,\alpha)\,c_{s1}(d,\alpha)\,z_{nk}\kappa^2 \,k^2r^{(0)}(\lambda_l(d,0)/k^2)\nonumber\\ & & + \bar{\beta}_k\big(h_s(d)\ , \left(2\lambda_l(d,0)\,k^2r^{(0)}(\lambda_l(d,0)/k^2 ) + k^4r^{(0)}(\lambda_l(d,0)/k^2)^2\right)\nonumber\\ & & + g_{s2}(d)\,r\,k^2r^{(0)}(\lambda_l(d,0)/k^2)\big)\bigg\ } \bigg ] \;.\end{aligned}\ ] ] here @xmath817 is the eigenvalue with respect to @xmath705 corresponding to @xmath284 . inserting also the remaining entries given in eq .( [ entries ] ) into eq .( [ 061 ] ) finally leads to eq .( [ squareflow ] ) . in this part of the appendix we expand @xmath332 of eq .( [ squareflow ] ) with respect to @xmath274 and evaluate the traces appearing in the resulting equation ( [ derivexp ] ) below by applying the heat kernel expansion .thereby we extract the contributions proportional to @xmath37 , @xmath38 and @xmath89 .this puts us in a position to read off the rg equations for the three couplings . for technical conveniencewe restrict our considerations to the gauge @xmath314 .we start our evaluation of @xmath332 , eq . ( [ squareflow ] ) , by expanding it with respect to @xmath347 .since we are only interested in the contributions proportional to @xmath818 , @xmath277 and @xmath278 , only terms of order @xmath819 , @xmath820 and @xmath281 are needed .this leads to @xmath821 + { \rm tr}_{(1t)}'\left[{\cal a}_1^{-1}{\cal n}\right ] -h_1(d)\,{\rm tr}_{(0)}''\left[{\cal a}_2^{-1}{\cal n}\right ] + a_k\,{\rm tr}_{(0)}''\left[{\cal a}_2^{-1}{\cal t}_2\right]}\nonumber\\ & & + a_k\,{\rm tr}_{(0)}''\left[({\cal a}_1{\cal a}_2)^{-1}p_k^2{\cal n } \right ] -2{\rm tr}_{(1t)}\left[p_k^{-1}{\cal n}_0\right ] -2{\rm tr'}_{(0)}\left[p_k^{-1}{\cal n}_0\right]\nonumber\\ & & + \bigg\{-a_k\,{\rm tr}_{(2st^2 ) } \left[{\cal a}_1^{-1}{\cal t}_1\right ] + a_k\,{\rm tr}_{(2st^2)}\left[{\cal a}_1^{-2}p_k{\cal n}\right ] -a_t(d)\,{\rm tr}_{(2st^2)}\left[{\cal a}_1^{-2}{\cal n}\right]\nonumber\\ & & -a_v(d,1)\,{\rm tr}_{(1t)}'\left[{\cal a}_1^{-2}{\cal n}\right ] + h_2(d)\,a_k\,{\rm tr}_{(0)}''\left[({\cal a}_1{\cal a}_2)^{-1}{\cal t}_2 \right]\nonumber\\ & & + h_3(d)\,a_k\,{\rm tr}_{(0)}''\left[({\cala}_1{\cal a}_2)^{-1}p_k{\cal n } \right ] + h_4(d)\,{\rm tr}_{(0)}''\left[({\cal a}_1{\cal a}_2)^{-1}{\cal n}\right ] + h_3(d)\,a_k\,{\rm tr}_{(0)}''\left[{\cal a}_2^{-1}{\cal t}_1\right]\nonumber\\ & & -h_2(d)\,a_k^2\,{\rm tr}_{(0)}''\left[{\cal a}_1^{-1}{\cal a}_2^{-2}p_k^2 { \cal t}_2\right ] -h_3(d)\,a_k^2\,{\rm tr}_{(0)}''\left[{\cal a}_2^{-2}p_k{\cal t}_2\right ] -h_4(d)\,a_k\,{\rm tr}_{(0)}''\left[{\cal a}_2^{-2}{\cal t}_2\right ] \nonumber\\ & & -h_2(d)\,a_k^2\,{\rm tr}_{(0)}''\left[({\cal a}_1{\cal a}_2)^{-2 } p_k^4{\cal n}\right ] -h_3(d)\,a_k^2\,{\rm tr}_{(0)}''\left[{\cal a}_1^{-1}{\cal a}_2^{-2 } p_k^3{\cal n}\right]\nonumber\\ & & -2h_4(d)\,a_k\ , { \rm tr}_{(0)}''\left[{\cal a}_1^{-1}{\cal a}_2^{-2}p_k^2{\cal n}\right ] + h_1(d)\,h_3(d)\,a_k\,{\rm tr}_{(0)}''\left[{\cal a}_2^{-2}p_k{\caln } \right]\nonumber\\ & & + h_1(d)\,h_4(d)\,{\rm tr}_{(0)}''\left[{\cal a}_2^{-2}{\cal n}\right ] -\frac{2}{d}\,{\rm tr}_{(1t)}\left[p_k^{-2}{\cal n}_0\right ] -\frac{4}{d}\,{\rm tr}_{(0)}'\left[p_k^{-2}{\cal n}_0\right]\nonumber\\ & & + \frac{\delta_{d,2}}{4\pi}\left[(\bar{\beta}_k\,k^4)^{-1 } \,\partial_t(\bar{\beta}_k\,k^4)\right]\int d^dx\,\sqrt{g}\bigg\}\,r\nonumber\\ & & + \bigg\ { -a_k^2\,{\rm tr}_{(2st^2)}\left[{\cal a}_1^{-2}p_k{\cal t}_1\right ] + a_k\,a_t(d)\,{\rm tr}_{(2st^2)}\left[{\cal a}_1^{-2}{\cal t}_1\right ] -g_t(d)\,a_k\ , { \rm tr}_{(2st^2)}\left[{\cal a}_1^{-2}{\cal n}\right]\nonumber\\ & & + a_k^2\,{\rm tr}_{(2st^2)}\left[{\cal a}_1^{-3}p_k^2{\cal n}\right ] -2 a_t(d)\,a_k\,{\rm tr}_{(2st^2)}\left[{\cal a}_1^{-3}p_k{\cal n}\right ] + a_t(d)^2\,{\rm tr}_{(2st^2)}\left[{\cal a}_1^{-3}{\cal n}\right]\nonumber\\ & & -g_v(d)\,a_k\,{\rm tr}_{(1t)}'\left[{\cal a}_1^{-2}{\cal n}\right ] + a_v(d,1)^2\,{\rm tr}_{(1t)}'\left[{\cal a}_1^{-3}{\cal n}\right ] -h_5(d)\,a_k^2\,{\rm tr}_{(0)}''\left[({\cal a}_1{\cal a}_2)^{-1}{\cal t}_2 \right]\nonumber\\ & & + h_6(d)\,a_k\,{\rm tr}_{(0)}''\left[({\cal a}_1{\cal a}_2)^{-1}{\cal n } \right ] -\frac{2}{d^2}\,a_k^2\,{\rm tr}_{(0)}''\left[({\cal a}_1{\cal a}_2)^{-1}p_k { \cal t}_1\right]\nonumber\\ & & + h_2(d)\,h_3(d)\,a_k\ , { \rm tr}_{(0)}''\left[({\cal a}_1{\cal a}_2)^{-1}{\cal t}_1\right ] -h_2(d)^2\,a_k^2\,{\rm tr}_{(0)}''\left[({\cal a}_1{\cal a}_2)^{-2}p_k^2 { \cal t}_2\right]\nonumber\\ & & + \frac{1}{2}h_2(d)\,a_k^3\ , { \rm tr}_{(0)}''\left[{\cal a}_1^{-1}{\cal a}_2^{-2}p_k^2{\cal t}_2\right ] -2h_2(d)\,h_3(d)\,a_k^2\ , { \rm tr}_{(0)}''\left[{\cal a}_1^{-1}{\cal a}_2^{-2}p_k{\cal t}_2\right ] \nonumber\\ & & -\frac{3}{2}h_2(d)\,h_4(d)\,a_k\ , { \rm tr}_{(0)}''\left[{\cal a}_1^{-1}{\cal a}_2^{-2}{\cal t}_2\right ] -h_6(d)\,a_k^2\,{\rm tr}_{(0)}''\left[{\cal a}_2^{-2}{\cal t}_2\right ] \nonumber\\ & & + \frac{1}{2}h_2(d)\,a_k^3\ , { \rm tr}_{(0)}''\left[({\cal a}_1{\cal a}_2)^{-2}p_k^4{\cal n } \right ] -2h_2(d)\,h_3(d)\,a_k^2\ , { \rm tr}_{(0)}''\left[({\cal a}_1{\cal a}_2)^{-2}p_k^3{\caln } \right]\nonumber\\ & & -\frac{3}{2}h_2(d)\,h_4(d)\,a_k\ , { \rm tr}_{(0)}''\left[({\cal a}_1{\cal a}_2)^{-2}p_k^2{\cal n}\right ] -h_7(d)\,a_k^2\ , { \rm tr}_{(0)}''\left[{\cal a}_1^{-1}{\cal a}_2^{-2}p_k^2{\caln } \right]\nonumber\\ & & -3h_3(d)\,h_4(d)\,a_k\,{\rm tr}_{(0)}''\left[{\cal a}_1^{-1}{\cal a}_2^{-2 } p_k{\cal n}\right ] -\frac{3}{2}h_4(d)^2\ , { \rm tr}_{(0)}''\left[{\cal a}_1^{-1}{\cal a}_2^{-2}{\cal n}\right]\nonumber\\ & & + h_1(d)\,h_6(d)\,a_k\,{\rm tr}_{(0)}''\left[{\cal a}_2^{-2}{\cal n}\right ] -h_2(d)\,h_3(d)\,a_k^2\,{\rm tr}_{(0)}''\left[{\cal a}_1^{-1 } { \cal a}_2^{-2}p_k^2{\cal t}_1\right]\nonumber\\ & & -h_3(d)^2\,a_k^2\ , { \rm tr}_{(0)}''\left[{\cal a}_2^{-2}p_k{\cal t}_1\right ] -h_3(d)\,h_4(d)\,a_k\,{\rm tr}_{(0)}''\left[{\cal a}_2^{-2}{\cal t}_1\right ] \nonumber\\ & & + h_2(d)^2\,a_k^3\,{\rm tr}_{(0)}''\left[{\cal a}_1^{-2 } { \cal a}_2^{-3}p_k^4{\cal t}_2\right ] + 2h_2(d)\,h_3(d)\,a_k^3\ , { \rm tr}_{(0)}''\left[{\cal a}_1^{-1}{\cal a}_2^{-3}p_k^3{\cal t}_2\right ] \nonumber\\ & & + 2h_2(d)\,h_4(d)\,a_k^2\ , { \rm tr}_{(0)}''\left[{\cal a}_1^{-1}{\cal a}_2^{-3}p_k^2{\cal t}_2\right ] + h_3(d)^2\,a_k^3\,{\rm tr}_{(0 ) } '' \left[{\cala}_2^{-3}p_k^2{\cal t}_2\right]\nonumber\\ & & + 2h_3(d)\,h_4(d)\,a_k^2\,{\rm tr}_{(0 ) } '' \left[{\cal a}_2^{-3}p_k{\cal t}_2\right ] + h_4(d)^2\,a_k\,{\rm tr}_{(0 ) } '' \left[{\cal a}_2^{-3}{\cal t}_2\right]\nonumber\\ & & + h_2(d)^2\,a_k^3\,{\rm tr}_{(0)}''\left[({\cal a}_1 { \cal a}_2)^{-3}p_k^6{\cal n}\right ] + 2h_2(d)\,h_3(d)\,a_k^3\,{\rm tr}_{(0)}''\left[{\cal a}_1^{-2}{\cal a}_2^{-3 } p_k^5{\caln}\right]\nonumber\\ & & + 3h_2(d)\,h_4(d)\,a_k^2\ , { \rm tr}_{(0)}''\left[{\cal a}_1^{-2}{\cal a}_2^{-3}p_k^4{\cal n}\right ] + h_3(d)^2\,a_k^3\,{\rm tr}_{(0)}''\left[{\cal a}_1^{-1}{\cal a}_2^{-3}p_k^4 { \cal n}\right]\nonumber\\ & & + 4h_3(d)\,h_4(d)\,a_k^2\,{\rm tr}_{(0)}''\left[{\cal a}_1^{-1 } { \cal a}_2^{-3}p_k^3{\cal n}\right ] + 3h_4(d)^2\,a_k\,{\rm tr}_{(0)}''\left[{\cal a}_1^{-1}{\cal a}_2^{-3 } p_k^2{\cal n}\right]\nonumber\\ & & -h_1(d)\,h_3(d)^2\,a_k^2\,{\rm tr}_{(0 ) } '' \left[{\cal a}_2^{-3}p_k^2{\cal n}\right ] -2h_1(d)\,h_3(d)\,h_4(d)\,a_k\,{\rm tr}_{(0 ) } '' \left[{\cal a}_2^{-3}p_k{\cal n}\right]\nonumber\\ & & -h_1(d)\,h_4(d)^2\,{\rm tr}_{(0)}''\left[{\cal a}_2^{-3}{\cal n}\right ] -\frac{2}{d^2}\,{\rm tr}_{(1t)}\left[p_k^{-3}{\cal n}_0\right ] -\frac{8}{d^2}\,{\rm tr}_{(0)}'\left[p_k^{-3}{\cal n}_0\right]\nonumber\\ & & + \int d^dx\,\sqrt{g}\bigg [ \frac{\delta_{d,2}}{8\pi\,k^4}\left(1 - 3r^{(0)'}(0)\right ) \left((\bar{\beta}_k\,k^2)^{-1}\partial_t(\bar{\beta}_k\,k^4 ) -\bar{\beta}_k^{-1}\partial_t(\bar{\beta}_k\,k^2)\right ) \nonumber\\ & & + \frac{\delta_{d,4}}{8(4\pi)^2}\big(9a_k\,k^4 - 2(k^2 - 2\bar{\lambda}_k ) \big)^{-1}\big(9a_k\,\bar{\beta}_k^{-1}\partial_t(\bar{\beta}_k\,k^4 ) -2z_{nk}^{-1}\partial_t(z_{nk}\,k^2)\big ) \bigg]\bigg\}\,r^2\nonumber\\ & & + { \cal o}(r^{<d-4})\;.\end{aligned}\ ] ] here we set @xmath822 and @xmath823 the quantities @xmath336 , @xmath337 , @xmath338 and @xmath339 are defined as in eq .( [ ntop ] ) .furthermore , @xmath824 means that terms @xmath825 with powers @xmath826 are neglected .the terms in eq .( [ derivexp ] ) proportional to @xmath827 and @xmath828 arise from the last term in eq .( [ squareflow ] ) .contrary to the other terms of eq .( [ squareflow ] ) , its expansion does not contain @xmath30-dependent powers of @xmath274 , but is of the form @xmath829 with @xmath830 a set of @xmath274-independent coefficients .as for comparing powers of @xmath274 , this has the following consequence .since , for all @xmath831 and @xmath832 , @xmath833 or @xmath834 are satisfied only if @xmath835 , and since @xmath836 can not be satisfied at all , this term contributes to the evolution equation only in the two- and the four - dimensional case . using eq .( [ spheres ] ) the pieces contributing , i.e. @xmath837 in @xmath838 and @xmath73 , and @xmath839 in @xmath838 , may be expressed in terms of the operators @xmath840 , @xmath841 or @xmath842 .this yields the terms in eq .( [ derivexp ] ) which are proportional to the @xmath843 s .as the next step we evaluate the @xmath274-expansion of the traces appearing in eq .( [ derivexp ] ) by applying the heat kernel expansion . in its original formit has often been used to compute traces of operators acting on unconstrained fields .for our purposes we need the * heat kernel expansions for operators acting on constrained fields * , i.e. fields satisfying appropriate transversality conditions . in ref .@xcite these expansions are derived in detail for laplacians @xmath324 on @xmath273 backgrounds acting on symmetric transverse traceless tensors , on transverse vectors and on scalars , with the following results : @xmath844 = \left(\frac{{\rm i}}{4\pi ( s+{\rm i}\varepsilon)}\right)^{d/2}\int d^dx\ , \sqrt{g}\bigg\ { \frac{1}{2}(d-2)(d+1)}\nonumber\\ & & -\frac{(d+1)(d+2)(d-5 + 3\delta_{d,2})}{12(d-1 ) } ( { \rm i } s-\varepsilon ) r\\\ & & -\frac{(d+1)(5d^4 - 22d^3 - 83d^2 - 392d-228 + 1440\,\delta_{d,2 } + 3240\,\delta_{d,4})}{720d(d-1)^2}(s+{\rm i}\varepsilon)^2 r^2\nonumber\\ & & + { \cal o}(r^3)\bigg\}\;,\nonumber\end{aligned}\ ] ] @xmath845 = \left(\frac{{\rm i}}{4\pi ( s+{\rm i}\varepsilon)}\right)^{d/2}\int d^dx\ , \sqrt{g}\bigg\{d-1}\nonumber\\ & & -\frac{(d+2)(d-3)+6\delta_{d,2}}{6d}({\rm i}s-\varepsilon ) r\\ & & -\frac{5d^4 - 12d^3 - 47d^2 - 186d+180 + 360\,\delta_{d,2 } + 720\,\delta_{d,4}}{360d^2(d-1 ) } ( s+{\rm i}\varepsilon)^2 r^2 + { \cal o}(r^3)\bigg\}\;,\nonumber\end{aligned}\ ] ] @xmath846 = \left(\frac{{\rm i}}{4\pi ( s+{\rm i}\varepsilon)}\right)^{d/2}\int d^dx\ , \sqrt{g}}\nonumber\\ & & \times\bigg\{1-\frac{1}{6}({\rm i } s-\varepsilon ) r -\frac{5d^2 - 7d+6}{360d(d-1 ) } ( s+{\rm i}\varepsilon)^2 r^2 + { \cal o}(r^3)\bigg\}\;.\end{aligned}\ ] ] here the terms proportional to the @xmath843 s arise from the exclusion of unphysical modes of the type discussed in section [ 3s2 ] .let us now consider an arbitrary function @xmath847 with a fourier transform @xmath848 .for such functions @xmath9 , we may express the trace of the operator @xmath849 that results from replacing the argument of @xmath9 with @xmath705 in terms of @xmath848 : @xmath850=\lim\limits_{\varepsilon\searrow 0 } \int\limits^{\infty}_{-\infty}ds\,\widetilde{w}(s)\,{\rm tr } \left[e^{-({\rm i } s-\varepsilon)d^2}\right]\end{aligned}\ ] ] we obtain the asymptotic expansion of @xmath851 $ ] by inserting the heat kernel expansion for @xmath852 $ ] into eq .( [ fourier ] ) . for laplacians acting on the constrained fields considered here they read as follows : @xmath853=(4\pi)^{-d/2}\bigg\ { \frac{1}{2}(d-2)(d+1)\,q_{d/2}[w]\ , \int d^dx\,\sqrt{g}}\nonumber\\ & & + \frac{(d+1)(d+2)(d-5 + 3\delta_{d,2})}{12(d-1 ) } \,q_{d/2 - 1}[w]\,\intd^dx\,\sqrt{g}r\nonumber\\ & & + \frac{(d+1)(5d^4 - 22d^3 - 83d^2 - 392d-228 + 1440\,\delta_{d,2 } + 3240\,\delta_{d,4})}{720d(d-1)^2}\,q_{d/2 - 2}[w]\nonumber\\ & & \times\int d^dx\,\sqrt{g}r^2 + { \cal o}(r^{<d-4})\bigg\}\;,\end{aligned}\ ] ] @xmath854=(4\pi)^{-d/2}\bigg\ { ( d-1)\,q_{d/2}[w]\,\int d^dx\,\sqrt{g}}\nonumber\\ & & + \frac{(d+2)(d-3)+6\delta_{d,2}}{6d } \,q_{d/2 - 1}[w]\,\int d^dx\,\sqrt{g}r\nonumber\\ & & + \frac{5d^4 - 12d^3 - 47d^2 - 186d+180 + 360\,\delta_{d,2 } + 720\,\delta_{d,4}}{360d^2(d-1)}\,q_{d/2 - 2}[w]\,\int d^dx\,\sqrt{g}r^2 \nonumber\\ & & + { \cal o}(r^{<d-4})\bigg\}\;,\end{aligned}\ ] ] @xmath855=(4\pi)^{-d/2}\bigg\{q_{d/2}[w]\ , \int d^dx\,\sqrt{g}}\\ & & + \frac{1}{6}\,q_{d/2 - 1}[w]\,\int d^dx\,\sqrt{g}r + \frac{5d^2 - 7d+6}{360d(d-1)}\,q_{d/2 - 2}[w]\,\int d^dx\,\sqrt{g}r^2 + { \cal o}(r^{<d-4})\bigg\}\;.\nonumber\end{aligned}\ ] ] here the set of functionals @xmath856 $ ] is defined as @xmath857\equiv\lim\limits_{\varepsilon\searrow 0}\int \limits^{\infty}_{-\infty } ds\,(-{\rm i } s+\varepsilon)^{-n}\ , \widetilde{w}(s)\;.\end{aligned}\ ] ] by virtue of the mellin transformation we may now reexpress @xmath858 in terms of @xmath9 so that @xmath859=\frac{(-1)^i}{\gamma(n+i)}\int\limits_{0}^{\infty}dz\,z^{n+i-1}\ , \frac{d^i w(z)}{dz^i}\;,\;\,i>-n\;,\;\,i\in \mathbb{n}\cup\{0\}\;\;{\rm arbitrary}.\end{aligned}\ ] ] in particular we obtain @xmath860=w(0)$ ] .furthermore , if @xmath861 we may choose @xmath862 for simplicity . as can be seen by an appropriate integration by parts , @xmath856 $ ]does not depend on @xmath863 . at this pointit is necessary to discuss the case where isolated eigenvalues have to be excluded from @xmath851 $ ] .as we showed in @xcite , such traces can be expressed as @xmath864={\rm tr}[w(-d^2 ) ] -\sum\limits_{l\in\{l_1,\ldots , l_n\}}d_l(d , s)w\left(\lambda_l(d , s)\right)\;.\end{aligned}\ ] ] here @xmath380 primes at @xmath865 symbolize the exclusion of all eigenmodes @xmath284 with @xmath866 , and @xmath703 and @xmath867 denote the corresponding eigenvalues of @xmath705 and their degrees of degeneracy , respectively . since @xmath868we may view @xmath869 as a function of @xmath718 .as outlined above , such a function contributes to the evolution of @xmath41 , @xmath266 and @xmath251 only for @xmath838 and for @xmath73 , with the contributions given by @xmath870 and @xmath871 , respectively . using the explicit expressions for @xmath703 and@xmath867 ( see table 2 in appendix [ harm ] ) and applying eq .( [ spheres ] ) we therefore obtain for the traces relevant to the flow equation : @xmath872={\rm tr}_{(1t)}[w(-d^2 ) ] -\frac{\delta_{d,2}}{16\pi}\left\{6w(0)\,\int d^2x\,\sqrt{g}\,r\right . }\nonumber\\ & & \left.+3w'(0)\,\int d^2x\,\sqrt{g}\,r^2\right\ } -\frac{5\delta_{d,4 } } { 12(4\pi)^2 } \,w(0)\,\int d^4x\,\sqrt{g}\,r^2+{\cal o}\left(r^{<d-4}\right)\;,\nonumber\\\end{aligned}\ ] ] @xmath873={\rm tr}_{(0)}[w(-d^2 ) ] -\frac{\delta_{d,2}}{16\pi}\left\{8w(0)\,\int d^2x\,\sqrt{g}\,r\right . }\nonumber\\ & & \left.+3w'(0)\,\int d^2x\,\sqrt{g}\,r^2\right\ } -\frac{\delta_{d,4 } } { 4(4\pi)^2 } \,w(0)\,\int d^4x\,\sqrt{g}\,r^2+{\cal o}\left(r^{<d-4}\right)\;,\end{aligned}\ ] ] @xmath874={\rm tr}_{(0)}[w(-d^2)]}\nonumber\\ & & -\frac{\delta_{d,2}}{8\pi}\,w(0)\,\int d^2x\,\sqrt{g}\,r -\frac{\delta_{d,4}}{24(4\pi)^2 } \,w(0)\,\int d^4x\,\sqrt{g}\,r^2+{\cal o}\left(r^{<d-4}\right)\;.\end{aligned}\ ] ] here @xmath875 denotes the derivative with respect to the argument : @xmath876 with @xmath877 .the next step is to insert the expansions of the traces into @xmath332 , eq .( [ derivexp ] ) , and to compare the coefficients of the operators @xmath878 , @xmath690 and @xmath689 with those on the lhs , eq .( [ lhs ] ) .this leads to the following differential equations : @xmath879 -h_1(d)\,q_{d/2}\left[{\cal a}_2^{-1}{\cal n}\right]}\nonumber\\ & & + a_k\,q_{d/2}\left[{\cal a}_2^{-1}{\cal t}_2\right ] + a_k\,q_{d/2}\left[({\cal a}_1{\cal a}_2)^{-1}p_k^2{\cal n}\right ] -2d\,q_{d/2}\left[p_k^{-1}{\cal n}_0\right]\big\}\;,\end{aligned}\ ] ] @xmath880 -\frac{1}{6}h_1(d)\,q_{d/2 - 1}\left[{\cal a}_2^{-1}{\cal n}\right]}\nonumber\\ & & + \frac{1}{6}a_k\,q_{d/2 - 1}\left[{\cal a}_2^{-1}{\cal t}_2\right ] + \frac{1}{6}a_k\,q_{d/2 - 1}\left[({\cal a}_1{\cal a}_2)^{-1}p_k^2{\cal n}\right ] + h_{10}(d)\,q_{d/2 - 1}\left[p_k^{-1}{\cal n}_0\right]\nonumber\\ & & -h_{11}(d)\,a_k\,q_{d/2}\left[{\cal a}_1^{-1}{\cal t}_1\right ] + h_{11}(d)\,a_k\,q_{d/2}\left[{\cal a}_1^{-2}p_k{\cal n}\right ] + h_{12}(d)\,q_{d/2}\left[{\cal a}_1^{-2}{\cal n}\right]\nonumber\\ & & + h_2(d)\,a_k\,q_{d/2}\left[({\cal a}_1{\cal a}_2)^{-1}{\cal t}_2 \right ] + h_3(d)\,a_k\,q_{d/2}\left[({\cal a}_1{\cal a}_2)^{-1}p_k{\cal n}\right ] \nonumber\\ & & + h_4(d)\,q_{d/2}\left[({\cal a}_1{\cal a}_2)^{-1}{\cal n}\right ] + h_3(d)\,a_k\,q_{d/2}\left[{\cal a}_2^{-1}{\cal t}_1\right ] -h_2(d)\,a_k^2\,q_{d/2}\left[{\cal a}_1^{-1}{\cal a}_2^{-2}p_k^2 { \cal t}_2\right]\nonumber\\ & & -h_3(d)\,a_k^2\,q_{d/2}\left[{\cal a}_2^{-2}p_k{\cal t}_2\right ] -h_4(d)\,a_k\,q_{d/2}\left[{\cal a}_2^{-2}{\cal t}_2\right ] -h_2(d)\,a_k^2\,q_{d/2}\left[({\cal a}_1{\cal a}_2)^{-2 } p_k^4{\cal n}\right]\nonumber\\ & & -h_3(d)\,a_k^2\,q_{d/2}\left[{\cal a}_1^{-1}{\cal a}_2^{-2 } p_k^3{\cal n}\right ] -2h_4(d)\,a_k\ , q_{d/2}\left[{\cal a}_1^{-1}{\cal a}_2^{-2}p_k^2{\cal n}\right]\\ & & + h_1(d)\,h_3(d)\,a_k\,q_{d/2}\left[{\cal a}_2^{-2}p_k{\cal n } \right ] + h_1(d)\,h_4(d)\,q_{d/2}\left[{\cal a}_2^{-2}{\cal n}\right ] + h_{13}(d)\,q_{d/2}\left[p_k^{-2}{\cal n}_0\right]\big\}\nonumber\;,\end{aligned}\ ] ] @xmath881 -h_1(d)\,h_{15}(d)\,q_{d/2 - 2}\left[{\cal a}_2^{-1}{\cal n}\right]}\nonumber\\ & & + h_{15}(d)\,a_k\,q_{d/2 - 2}\left[{\cal a}_2^{-1}{\cal t}_2\right ] + h_{15}(d)\,a_k\,q_{d/2 - 2}\left[({\cal a}_1{\cal a}_2)^{-1}p_k^2{\caln } \right]\nonumber\\ & & -h_{16}(d)\,q_{d/2 - 2}\left[p_k^{-1}{\cal n}_0\right ] -h_{17}(d)\,a_k\,q_{d/2 - 1}\left[{\cal a}_1^{-1}{\cal t}_1\right ] + h_{17}(d)\,a_k\,q_{d/2 - 1}\left[{\cal a}_1^{-2}p_k{\cal n}\right]\nonumber\\ & & -h_{18}(d)\,q_{d/2 - 1}\left[{\cal a}_1^{-2}{\cal n}\right ] + \frac{1}{6}h_2(d)\,a_k\,q_{d/2 - 1}\left[({\cal a}_1{\cal a}_2)^{-1}{\cal t}_2 \right]\nonumber\\ & & + \frac{1}{6}h_3(d)\,a_k\,q_{d/2 - 1 } \left[({\cal a}_1{\cal a}_2)^{-1}p_k{\cal n}\right ] + \frac{1}{6}h_4(d)\,q_{d/2 - 1}\left[({\cal a}_1{\cal a}_2)^{-1}{\cal n}\right ] \nonumber\\ & & + \frac{1}{6}h_3(d)\,a_k\,q_{d/2 - 1}\left[{\cal a}_2^{-1}{\cal t}_1\right ] -\frac{1}{6}h_2(d)\,a_k^2\,q_{d/2 - 1}\left[{\cal a}_1^{-1}{\cal a}_2^{-2 } p_k^2{\cal t}_2\right]\nonumber\\ & & -\frac{1}{6}h_3(d)\,a_k^2\,q_{d/2 - 1}\left[{\cal a}_2^{-2}p_k{\cal t}_2 \right ] -\frac{1}{6}h_4(d)\,a_k\,q_{d/2 - 1}\left[{\cal a}_2^{-2}{\cal t}_2\right ] \nonumber\\ & & -\frac{1}{6}h_2(d)\,a_k^2\,q_{d/2 - 1}\left[({\cal a}_1{\cal a}_2)^{-2 } p_k^4{\cal n}\right ] -\frac{1}{6}h_3(d)\,a_k^2\,q_{d/2 - 1}\left[{\cal a}_1^{-1}{\cal a}_2^{-2 } p_k^3{\cal n}\right]\nonumber\\ & & -\frac{1}{3}h_4(d)\,a_k\ , q_{d/2 - 1}\left[{\cal a}_1^{-1}{\cal a}_2^{-2}p_k^2{\cal n}\right ] + \frac{1}{6}h_1(d)\,h_{19}(d)\,a_k\,q_{d/2 - 1}\left[{\cal a}_2^{-2}p_k{\caln } \right]\nonumber\\ & & + \frac{1}{6}h_1(d)\,h_4(d)\,q_{d/2 - 1}\left[{\cal a}_2^{-2}{\cal n}\right ] -h_{20}(d)\,q_{d/2 - 1}\left[p_k^{-2}{\cal n}_0\right ] -h_{11}(d)\,a_k^2\,q_{d/2}\left[{\cal a}_1^{-2}p_k{\cal t}_1\right]\nonumber\\ & & + \frac{1}{2}h_{11}(d)\,h_{21}(d)\,a_k\,q_{d/2}\left[{\cal a}_1^{-2 } { \cal t}_1\right ] + h_{22}(d)\,a_k\,q_{d/2}\left[{\cal a}_1^{-2}{\cal n}\right ] \nonumber\\ & & + h_{11}(d)\,a_k^2\,q_{d/2}\left[{\cal a}_1^{-3}p_k^2{\cal n}\right ] -h_{11}(d)\,h_{21}(d)\,a_k\,q_{d/2}\left[{\cal a}_1^{-3}p_k{\cal n}\right ] \nonumber\\ & & + h_{23}(d)\,q_{d/2}\left[{\cal a}_1^{-3}{\cal n}\right ] -h_5(d)\,a_k^2\,q_{d/2}\left[({\cal a}_1{\cal a}_2)^{-1}{\cal t}_2 \right]\nonumber\\ & & + h_6(d)\,a_k\,q_{d/2}\left[({\cal a}_1{\cal a}_2)^{-1}{\cal n } \right ] -\frac{2}{d^2}\,a_k^2\,q_{d/2}\left[({\cal a}_1{\cal a}_2)^{-1}p_k { \cal t}_1\right]\nonumber\\ & & + h_2(d)\,h_3(d)\,a_k\ , q_{d/2}\left[({\cal a}_1{\cal a}_2)^{-1}{\cal t}_1\right ] -h_2(d)^2\,a_k^2\,q_{d/2}\left[({\cal a}_1{\cal a}_2)^{-2}p_k^2 { \cal t}_2\right]\nonumber\\ & & + \frac{1}{2}h_2(d)\,a_k^3\ , q_{d/2}\left[{\cal a}_1^{-1}{\cal a}_2^{-2}p_k^2{\cal t}_2\right ] -2h_2(d)\,h_3(d)\,a_k^2\ , q_{d/2}\left[{\cal a}_1^{-1}{\cal a}_2^{-2}p_k{\cal t}_2\right ] \nonumber\\ & & -\frac{3}{2}h_2(d)\,h_4(d)\,a_k\ , q_{d/2}\left[{\cal a}_1^{-1}{\cal a}_2^{-2}{\cal t}_2\right ] -h_6(d)\,a_k^2\,q_{d/2}\left[{\cal a}_2^{-2}{\cal t}_2\right ] \nonumber\\ & & + \frac{1}{2}h_2(d)\,a_k^3\ , q_{d/2}\left[({\cal a}_1{\cal a}_2)^{-2}p_k^4{\cal n}\right ] -2h_2(d)\,h_3(d)\,a_k^2\,q_{d/2}\left[({\cal a}_1{\cal a}_2)^{-2}p_k^3{\caln } \right]\nonumber\\ & & -\frac{3}{2}h_2(d)\,h_4(d)\,a_k\ , q_{d/2}\left[({\cal a}_1{\cal a}_2)^{-2}p_k^2{\cal n}\right ] -h_7(d)\,a_k^2\,q_{d/2}\left[{\caln } \right]\nonumber\\ & & -3h_3(d)\,h_4(d)\,a_k\,q_{d/2}\left[{\cal a}_1^{-1}{\cal a}_2^{-2 } p_k{\cal n}\right ] -\frac{3}{2}h_4(d)^2\ , q_{d/2}\left[{\cal a}_1^{-1}{\cal a}_2^{-2}{\cal n}\right]\nonumber\\ & & + h_1(d)\,h_6(d)\,a_k\,q_{d/2}\left[{\cal a}_2^{-2}{\cal n}\right ] -h_2(d)\,h_3(d)\,a_k^2\,q_{d/2}\left[{\cal a}_1^{-1 } { \cal a}_2^{-2}p_k^2{\cal t}_1\right]\nonumber\\ & & -h_3(d)^2\,a_k^2\,q_{d/2}\left[{\cal a}_2^{-2}p_k{\cal t}_1\right ] -h_3(d)\,h_4(d)\,a_k\,q_{d/2}\left[{\cal a}_2^{-2}{\cal t}_1\right ] \nonumber\\ & & + h_2(d)^2\,a_k^3\,q_{d/2}\left[{\cal a}_1^{-2 } { \cal a}_2^{-3}p_k^4{\cal t}_2\right ] + 2h_2(d)\,h_3(d)\,a_k^3\ , q_{d/2}\left[{\cal a}_1^{-1}{\cal a}_2^{-3}p_k^3{\calt}_2\right]\nonumber\\ & & + 2h_2(d)\,h_4(d)\,a_k^2\ , q_{d/2}\left[{\cal a}_1^{-1}{\cal a}_2^{-3}p_k^2{\cal t}_2\right ] + h_3(d)^2\,a_k^3\,q_{d/2}\left[{\cal a}_2^{-3}p_k^2{\cal t}_2\right ] \nonumber\\ & & + 2h_3(d)\,h_4(d)\,a_k^2\,q_{d/2}\left[{\cal a}_2^{-3}p_k{\cal t}_2\right ] + h_4(d)^2\,a_k\,q_{d/2}\left[{\cal a}_2^{-3}{\cal t}_2\right]\nonumber\\ & & + h_2(d)^2\,a_k^3\,q_{d/2}\left[({\cal a}_1 { \cal a}_2)^{-3}p_k^6{\cal n}\right ] + 2h_2(d)\,h_3(d)\,a_k^3\,q_{d/2}\left[{\cal a}_1^{-2}{\cal a}_2^{-3 } p_k^5{\cal n}\right]\nonumber\\ & & + 3h_2(d)\,h_4(d)\,a_k^2\ , q_{d/2}\left[{\cal a}_1^{-2}{\cal a}_2^{-3}p_k^4{\cal n}\right ] + h_3(d)^2\,a_k^3\,q_{d/2}\left[{\cal a}_1^{-1}{\cal a}_2^{-3}p_k^4 { \cal n}\right]\nonumber\\ & & + 4h_3(d)\,h_4(d)\,a_k^2\,q_{d/2}\left[{\cal a}_1^{-1 } { \cal a}_2^{-3}p_k^3{\cal n}\right ] + 3h_4(d)^2\,a_k\,q_{d/2}\left[{\cal a}_1^{-1}{\cal a}_2^{-3 } p_k^2{\cal n}\right]\nonumber\\ & & -h_1(d)\,h_3(d)^2\,a_k^2\,q_{d/2}\left[{\cal a}_2^{-3}p_k^2{\cal n}\right ] -2h_1(d)\,h_3(d)\,h_4(d)\,a_k\ , q_{d/2}\left[{\cal a}_2^{-3}p_k{\cal n}\right]\nonumber\\ & & -h_1(d)\,h_4(d)^2\,q_{d/2}\left[{\cal a}_2^{-3}{\cal n}\right ] + h_{24}(d)\,q_{d/2}\left[p_k^{-3}{\cal n}_0\right]\nonumber\\ & & + \delta_{d,2}\big [ -\frac{3}{2}a_k\,\frac{\partial_t\left[\bar{\beta}_k\,k^2\right ] } { \bar{\beta_k}\left(k^2 - 2\bar{\lambda}_k\right ) } + \left(\frac{11}{4}r^{(0)'}(0)+\frac{3}{2}a_k\,k^2\right)\,\frac{\partial_t \left[z_{nk}\,k^2\right]}{z_{nk}\left(k^2 - 2\bar{\lambda}_k\right)^2 } \nonumber\\ & & -\frac{11}{4}r^{(0)'}(0)\frac{\partial_t\,z_{nk } } { z_{nk}\left(k^2 - 2\bar{\lambda}_k\right)}-\frac{1}{2k^2}r^{(0)'}(0 ) \big]\nonumber\\ & & + \delta_{d,4}\big [ \frac{1}{8}\,\frac{\partial_t\left[z_{nk}\,k^2\right ] } { z_{nk}\left(k^2 - 2\bar{\lambda}_k\right ) } + \frac{1}{4}\,\frac{\partial_t\left[z_{nk}\,k^2\right ] } { z_{nk}\left(3a_k\,k^4-\left(k^2 - 2\bar{\lambda}_k\right)\right)}\nonumber\\ & & -\frac{3}{8}a_k\,k^4\,\frac{\partial_t\left[z_{nk}\,k^2\right ] } { z_{nk}\left(k^2 - 2\bar{\lambda}_k\right)\left(3a_k\,k^4-\left(k^2 -2\bar{\lambda}_k\right)\right ) } -\frac{1}{4}\,\frac{\partial_t\left[z_{nk}\,k^2\right ] } { z_{nk}\left(9a_k\,k^4 - 2\left(k^2 - 2\bar{\lambda}_k\right)\right)}\nonumber\\ & & -\frac{3}{8}a_k\,\frac{\partial_t\left[\bar{\beta}_k\,k^4\right ] } { \bar{\beta}_k\left(3a_k\,k^4-\left(k^2 - 2\bar{\lambda}_k\right)\right ) } + \frac{9}{8}a_k\,\frac{\partial_t\left[\bar{\beta}_k\,k^4\right ] } { \bar{\beta}_k\left(9a_k\,k^4 - 2\left(k^2 - 2\bar{\lambda}_k\right)\right ) } \big]\big\}\;.\end{aligned}\ ] ] here the coefficients @xmath414 are functions of the dimensionality @xmath30 .they are tabulated in ( [ coeff3 ] ) of the appendix [ betacoeff ] .now we introduce the cutoff - dependent * generalized threshold functions * @xmath882\nonumber\end{aligned}\ ] ] and @xmath883\;.\nonumber\end{aligned}\ ] ] in eqs .( [ psi ] ) and ( [ psitilde ] ) , @xmath863 is a nonnegative integer which satisfies @xmath884 , but which is arbitrary otherwise .the functions @xmath368 and @xmath369 are independent of @xmath863 which can be seen by an integration by parts .again , we may set @xmath862 if @xmath861 .furthermore , noting that @xmath885 we recover the threshold functions @xmath886 and @xmath887 originally defined in @xcite in the context of the pure einstein - hilbert truncation . using the relations @xmath888 & = & k^{2(m+n - p - q+1)}\psi^{p;q}_{n;m}(a_k\,k^2/(32\pi),\bar{\lambda}_k / k^2;d ) \nonumber\\ & & -\frac{1}{2}\eta_{\beta}(k)\,k^{2(m+n - p - q+1 ) } \widetilde{\psi}^{p;q}_{n;m;0}(a_k\,k^2/(32\pi),\bar{\lambda}_k / k^2;d ) \nonumber\\q_n\left[{\cal a}_1^{-p}{\cal a}_2^{-q}p_k^m{\cal t}_2\right ] & = & 2k^{2(m+n - p - q+2)}\psi^{p;q}_{n;m+1}(a_k\,k^2/(32\pi),\bar{\lambda}_k / k^2;d ) \nonumber\\ & & -\frac{1}{2}\eta_{\beta}(k)\,k^{2(m+n - p - q+2 ) } \widetilde{\psi}^{p;q}_{n;m;1}(a_k\,k^2/(32\pi),\bar{\lambda}_k / k^2;d ) \nonumber\\q_n\left[{\cal a}_1^{-p}{\cal a}_2^{-q}p_k^m{\cal n}\right ] & = & k^{2(m+n - p - q+1)}\psi^{p;q}_{n;m}(a_k\,k^2/(32\pi),\bar{\lambda}_k / k^2;d ) \nonumber\\ & & -\frac{1}{2}\eta_n(k)\,k^{2(m+n - p - q+1 ) } \widetilde{\psi}^{p;q}_{n;m;0}(a_k\,k^2/(32\pi),\bar{\lambda}_k / k^2;d ) \nonumber\\ q_n\left[p_k^{-p}{\cal n}_0\right ] & = & k^{2(n - p+1)}\phi^p_n(0)\end{aligned}\ ] ] the differential equations ( [ q - level1 ] ) , ( [ q - level2 ] ) and ( [ q - level3 ] ) may be rewritten in terms of the threshold functions @xmath889 and @xmath890 instead of the @xmath858 . in order to make the integrals in eq .( [ qn ] ) convergent we have to demand that @xmath195 decreases rapidly as @xmath891 .however , since from now on its form for @xmath892 does not play a role any more we identify @xmath195 with its part for nonnegative arguments and assume that @xmath195 is a smooth function defined only for @xmath893 and endowed with the properties stated in subsection [ 3s3b ] .next we introduce the dimensionless couplings @xmath48 , @xmath47 and @xmath346 of eqs . ( [ lambda])-([beta ] ) . inserting eq .( [ lambda ] ) into @xmath894 leads to the relation @xmath895 then , by using eq .( [ q - level1 ] ) , we obtain the differential equation ( [ del ] ) for the dimensionless cosmological constant .the corresponding differential equations for @xmath47 and @xmath346 may be determined as follows .taking the scale derivative of eq .( [ g ] ) and ( [ beta ] ) leads to eqs .( [ deg ] ) and ( [ deb ] ) , respectively . for the anomalous dimensions@xmath896 and @xmath897 we obtain from eqs .( [ q - level2 ] ) and ( [ q - level3 ] ) , respectively , @xmath898 and @xmath899 the set of equations ( [ eta0 ] ) and ( [ rho0 ] ) may now be solved for the anomalous dimensions @xmath896 and @xmath897 in terms of @xmath48 , @xmath47 , @xmath346 and @xmath30 which eventually leads to the expressions ( [ eta ] ) and ( [ rho ] ) .[ coefffnct ] in the following we list the coefficient functions @xmath364 , @xmath365 , @xmath366 , @xmath367 , which appear in the @xmath360-functions ( [ del ] ) , ( [ deg ] ) and ( [ deb ] ) .the coefficients @xmath900 contained in these expressions are defined in subsection [ betacoeff ] , and the generalized threshold functions @xmath368 and @xmath369 were introduced in eq .( [ psi ] ) and eq .( [ psitilde ] ) , respectively .the other threshold functions , @xmath370 and @xmath371 , are those already introduced in the context of the pure einstein - hilbert truncation @xcite .@xmath901 @xmath902 @xmath903 @xmath904 @xmath905 @xmath906 @xmath907\delta_{d,2}\nonumber\\ & & + \left[\frac{288\pi\,g_k\,\beta_k-1 } { 4(144\pi\,g_k\,\beta_k-(1 - 2\lambda_k ) ) } + \frac{1}{4(1 - 2\lambda_k)}\right.\nonumber\\ & & \left.-\frac{96\pi\,g_k\,\beta_k-1}{2(96\pi\,g_k\,\beta_k -(1 - 2\lambda_k))}-\frac{24\pi\,g_k\,\beta_k}{(1 -2\lambda_k)\left(96\pi\,g_k\,\beta_k-(1 - 2\lambda_k\right ) } \right]\delta_{d,4 } \bigg\}\end{aligned}\ ] ] @xmath908 \delta_{d,2}\nonumber\\ & & + \left[-\frac{1}{4(144\pi\,g_k\,\beta_k-(1 - 2\lambda_k ) ) } + \frac{1}{4(1 - 2\lambda_k)}\right.\nonumber\\ & & \left.+\frac{1}{2(96\pi\,g_k\,\beta_k-(1 - 2\lambda_k ) ) } -\frac{24\pi\,g_k\,\beta_k}{(1 -2\lambda_k)\left(96\pi\,g_k\,\beta_k-(1 - 2\lambda_k)\right ) } \right]\delta_{d,4 } \bigg\}\end{aligned}\ ] ] @xmath909 \delta_{d,4}\bigg\}\end{aligned}\ ] ] in eqs .( [ coeffc1 ] ) , ( [ coeffc2 ] ) and ( [ coeffc3 ] ) the terms proportional to @xmath827 or @xmath828 arise not only from the @xmath843-terms of eq .( [ derivexp ] ) , but also by evaluating the `` primed '' traces , i.e. by subtracting the contributions coming from unphysical modes , see appendix [ trace ] for details .all these contributions are obtained by expanding various functions @xmath910 with respect to @xmath718 and retaining only the terms @xmath911 in @xmath838 and @xmath912 in @xmath73 . as we explained above , these are the only pieces of @xmath913 which may contribute to the evolution in the truncated parameter space .furthermore , the heat kernel expansions of the traces corresponding to differentially constrained fields introduce additional contributions proportional to @xmath827 or @xmath828 into eqs .( [ coeffc1])-([coeffc3 ] ) .[ harm ] the spherical harmonics @xmath282 , @xmath283 and @xmath284 for symmetric transverse traceless @xmath914 tensors @xmath285 , transverse @xmath915 vectors @xmath290 , and scalars @xmath125 on @xmath273 form complete sets of orthogonal eigenfunctions with respect to the covariant laplacians .they satisfy @xmath916 and , after proper normalization , @xmath917 here @xmath918 and @xmath919 are the unit matrices in the spaces of @xmath920 tensors and transverse vectors , respectively .the @xmath703 s denote the eigenvalues of @xmath921 where @xmath577 is the spin of the field under consideration and @xmath702 takes the values @xmath922 .the index @xmath923 is a degeneracy index . in ref .@xcite explicit expressions for @xmath703 and the degeneracies @xmath867 were derived which are summarized in table 2 .the eigenvalues are expressed in terms of the curvature scalar @xmath924 of the sphere with radius @xmath274 .the spherical harmonics @xmath282 , @xmath283 and @xmath284 span the spaces of @xmath920 tensors , @xmath925 vectors , and scalars so that we may expand arbitrary functions @xmath285 , @xmath290 and @xmath125 according to @xmath926 eqs .( [ 067 ] ) may now be used to expand also any symmetric non-@xmath927 tensor and nontransverse vector in terms of spherical harmonics since they may be expressed in terms of @xmath920 tensors , @xmath925 vectors and scalars by using the decompositions ( [ tt ] ) , ( [ t ] ) , see e.g. @xcite .note that the @xmath928 modes @xmath929 and the @xmath930 modes @xmath931 satisfy the killing equation ( [ 09 ] ) and the scalar equation ( [ 010 ] ) , respectively , and that @xmath932 .arbitrary symmetric rank-2 tensors receive no contribution from these modes . in the case of arbitrary vectorsthe constant scalar mode does not contribute .such modes have no physical meaning and have to be omitted therefore .eigenfunction & spin @xmath577 & eigenvalue @xmath703 & degeneracy @xmath867 & @xmath702 + @xmath933 & 2 & @xmath934 & @xmath935 & @xmath936 + @xmath937 & 1 & @xmath938 & @xmath939 & @xmath940 + @xmath941 & 0 & @xmath942 & @xmath943 & @xmath944 +[ somecoe ] in this subsection we define the various @xmath240 s , @xmath241 s , @xmath153 s and @xmath93 s and @xmath945 which appear in eqs .( [ gravquad])-([squareflow ] ) of subsection [ 3s4c ] and in eqs .( [ phi0term ] ) and ( [ derivexp ] ) of appendix [ eval ] .@xmath946 next we define the coefficients @xmath900 contained in the @xmath360-functions ( [ del ] ) , ( [ deg ] ) and ( [ deb ] ) via the coefficient functions @xmath364 , @xmath365 , @xmath366 , @xmath367 , given in appendix [ coefffnct ] .they also appear in the approximate solutions for the non - gaussian fixed point of appendix [ approxngfp ] .[ approxngfp ] in the following we derive the approximate formula for the position of the non - gaussian fixed point discussed in subsection [ 3s5c ] . herewe restrict our considerations to the case @xmath417 . in a first approximation we set @xmath948 , @xmath949 and determine @xmath396 from the condition @xmath950 alone .since @xmath951 , we may solve this equation for @xmath396 in closed form which leads to @xmath952 as @xmath953 , it boils down to @xmath954 here the @xmath900 are again @xmath30-dependent coefficients which are defined in subsection [ betacoeff ] of appendix [ somecoe ] .it is remarkable that the solution ( [ gstar1 ] ) coincides precisely with the corresponding approximate solution ( h2 ) of ref .@xcite with @xmath314 , obtained in the framework of the einstein - hilbert truncation . employing the exponential shape function ( [ expshape ] ) with @xmath459 , and setting @xmath73 , for instance , eq .( [ gstar1 ] ) yields @xmath571 .here we used that , for this shape function , @xmath955 , @xmath485 , @xmath956 , @xmath957 , see appendix [ threshprop ] . in order to improve upon this approximation scheme, we determine @xmath574 from a set of taylor - expanded @xmath372-functions . using eqs .( [ psiexpansion1])-([phiexpansion ] ) we expand the @xmath372-functions ( [ del ] ) , ( [ deg ] ) and ( [ deb ] ) about @xmath958 and obtain @xmath959 here @xmath412 , @xmath423 , and @xmath445 are defined as in eqs .( [ gamma_d ] ) , ( [ nu_d ] ) and ( [ omega_d ] ) , respectively , and @xmath960 stands for terms of @xmath380th and higher orders in the couplings @xmath961 , @xmath962 and @xmath963 . now @xmath396 is obtained as the nontrivial solution to @xmath964 , which reads @xmath965 inserting eq .( [ gstar2 ] ) into @xmath966 leads to @xmath967 quite remarkably , also these results agree completely with those of ref .@xcite which follow from the pure einstein - hilbert truncation .( see eqs .( h6 ) and ( h7 ) of this reference . )now we use @xmath968 in order to determine @xmath501 .however , since the term linear in @xmath346 vanishes for @xmath73 , the expanded @xmath968 of eq .( [ expandbeta ] ) is not sufficient in this case .therefore we consider also those terms of second order in the couplings which are linear in @xmath346 .for these terms we find @xmath969 taking the nonvanishing term of eq .( [ additional ] ) into account , and inserting @xmath396 of eq .( [ gstar2 ] ) into @xmath970 then leads to @xmath971 employing the shape function ( [ expshape ] ) with @xmath459 we obtain from eqs .( [ gstar2 ] ) , ( [ lambdastar1 ] ) and ( [ betastar1 ] ) in @xmath73 dimensions @xmath972 here we used the expressions for the threshold functions derived in appendix [ threshprop ] . the numbers in ( [ h7 ] ) should be compared to the exact result ( [ r^2fixedpoint ] ) .[ threshprop ] in this appendix we summarize various important properties of the threshold functions @xmath973 , @xmath974 , @xmath585 and @xmath975 which are defined by eqs .( [ psi ] ) , ( [ psitilde ] ) and ( [ phi ] ) . expanding the generalized threshold functions @xmath973 , @xmath974 about vanishing couplings yields @xmath976 @xmath977 @xmath978 here @xmath979 stands for terms of third and higher orders in the couplings @xmath961 , @xmath962 and @xmath963 .quite remarkably , every fixed order of these expansions depends only on the `` conventional '' threshold functions @xmath585 and @xmath975 at vanishing arguments . by using eq .( [ phi ] ) the corresponding expansions of @xmath585 and @xmath975 about vanishing argument can be read off directly from eqs .( [ psiexpansion1 ] ) and ( [ psiexpansion2 ] ) .they are given by @xmath980 for @xmath981 the threshold functions are universal in the sense that they do not depend on @xmath195 .in fact , setting @xmath981 in @xmath973 , @xmath974 , @xmath585 and @xmath975 leads to @xmath982 and @xmath983 there exists a second class of universal values of certain threshold functions . using the boundary conditions for @xmath195one may easily verify that , for vanishing argument and for @xmath984 , @xmath585 assumes the universal value @xmath985 let us now be more specific and opt for the family of exponential cutoffs ( [ expshape ] ) . in this casethe integral that defines @xmath585 can be carried out analytically for vanishing argument . using the integral representation of the polylogarithm @xcite , @xmath986 one obtains for @xmath987 : @xmath988 @xmath989 @xmath990 @xmath991^{-1}\,s^{3-n}\,\left\ { ( 2-s)\,{\rm li}_{n-2}(1-s)-s\,{\rm li}_{n-3}(1-s)\right\ } & n\neq 1\;,\;\;s\neq 1 \\ \left[2^{n-1}(n-1)\right]^{-1 } ( 2^{n-1}-1 ) & n\neq 1\;,\;\;s=1 \\ - [ 2(1-s)^2]^{-1}\,s^{3-n}\,\left\ { ( 2-s)\,{\rm li}_{-1}^{(1,0)}(1-s ) -s\,{\rm li}_{-2}^{(1,0)}(1-s)\right\ } & n=1\;,\;\;s\neq 1\\ \ln(2 ) & n = s=1 \end{array}\right.\nonumber\end{aligned}\ ] ] @xmath992^{-1}\,s^{4-n } \,\left\{2(s^2 - 3s+3)\,{\rm li}_{n-3}(1-s)\right . & \\\hspace{0.5cm}\left.-3s(2-s)\,{\rm li}_{n-4}(1-s)+s^2\,{\rm li}_{n-5}(1-s ) \right\ } & n\neq 1,2\;,\;\,s\neq 1 \\ \left[(n-1)(n-2)\right]^{-1}(1 - 2^{3-n}+3^{2-n } ) & n\neq 1,2\;,\;\,s=1 \\\left[6(2n-3)(1-s)^3\right]^{-1}\,s^{4-n}\,\left\ { 2(s^2 - 3s+3)\,{\rm li}_{n-3}^{(1,0)}(1-s)\right . & \\ \hspace{0.5cm}\left.-3s(2-s)\,{\rmli}_{n-4}^{(1,0)}(1-s ) + s^2\,{\rm li}_{n-5}^{(1,0)}(1-s)\right\ } & n\in\{1,2\}\;,\;\,s\neq 1\\ \ln(27/16 ) & n = s=1\\\ln(4/3 ) & n=2\;,\;\,s=1 \end{array}\right.\nonumber\end{aligned}\ ] ] here we defined @xmath993 and used the relations @xmath994 with @xmath488 denoting the riemann zeta - function . for nonvanishing arguments an analytic solution to the integrals defining the threshold functionsis not known . for the exponential cutoff( [ expshape ] ) with @xmath459 there even exists a very useful relation among @xmath460 and @xmath995 .one may easily verify that @xmath996 this relation allows us to calculate the @xmath995-integrals analytically as well .[ estimation ] in this appendix we prove the inequality ( [ phisigma ] ) . as a first step we consider the function @xmath997 with @xmath998 and @xmath999 .( for @xmath1000 this function vanishes identically : @xmath1001 . ) an upper bound for this function may be obtained as follows .for the first two derivatives of @xmath913 with respect to @xmath1002 we obtain @xmath1003^{\frac{3}{2 } } } + \frac{1}{4\left(1-y\right)^{\frac{3}{2}}}\;.\end{aligned}\ ] ] solving @xmath1004 for @xmath1002 leads to the _ single _ solution @xmath1005 since @xmath1006 , we have @xmath1007 for all @xmath1008 $ ] and @xmath998 .hence , for @xmath998 fixed but arbitrary , @xmath913 monotonically decreases in the interval @xmath1009 $ ] where @xmath1010 .furthermore , @xmath1011 is a monotonically increasing function of @xmath1012 for all @xmath1013 .therefore we have that @xmath1014 . as a consequence ,@xmath913 monotonically decreases in the interval @xmath1015 for any value of @xmath998 .since we restricted our considerations to @xmath1016 $ ] we obtain @xmath1017 .@xmath1018 may be obtained from @xmath913 by replacing @xmath1019 and multiplying the result by @xmath1020 .note that for all @xmath753 we have @xmath1021 so that @xmath1022 is indeed satisfied .moreover , @xmath998 is satisfied as long as @xmath1023 .if @xmath1024 , the cutoff in the scalar sector is zero anyway : @xmath1025 .hence , for positive values of @xmath346 , @xmath1017 leads to @xmath1026 next we insert @xmath1027 and @xmath742 of eq .( [ posdef ] ) into @xmath1028 and then use ( [ glorious ] ) .this yields @xmath1029 obviously the positivity condition @xmath1030 now boils down to @xmath1031 here @xmath1032 is the nonnegative function of @xmath2 , @xmath702 and @xmath718 which represents the contributions to @xmath1033 neglected on the rhs of ( [ hauptmincrit ] ) .we see that eq .( [ phisigma ] ) is a sufficient condition for the inequality ( [ condpos ] ) to be valid , i.e. for @xmath1034 to be positive .this is what we wanted to proof .99999999 m. reuter , phys .d 57 ( 1998 ) 971 and hep - th/9605030 o. lauscher , m. reuter , phys . rev .d 65 ( 2001 ) 025013 and hep - th/0108040 o. lauscher , m. reuter , class .( 2002 ) 483 and + hep - th/0110021 o. lauscher , m. reuter , in _ proceedings of the 5th workshop on quantum field theory under the influence of external conditions , _ int .a 17 ( 2002 ) 993 and hep - th/0112089 m. reuter , f. saueressig , phys .d 65 ( 2002 ) 065016 and hep - th/0110054 m. reuter , f. saueressig , in preparation d. dou , r. percacci , class .( 1998 ) 3449 a. bonanno , m. reuter , phys .d 60 ( 1999 ) 084011 and gr - qc/9811026 a. bonanno , m. reuter , phys . rev .d 62 ( 2000 ) 043008 and hep - th/0002196 a. bonanno , m. reuter , phys .d 65 ( 2002 ) 043508 and hep - th/0106133 ; for an application of similar methods to the very late universe see also phys .b 527 ( 2002 ) 9 and astro - ph/0106468 for a review see : j. berges , n. tetradis , c. wetterich , hep - ph/0005122 ; + c. wetterich , int . j. mod . phys . a 16 ( 2001 )1951 k. g. wilson , j. kogut , phys .12 ( 1974 ) 75 ; + k. g. wilson , rev . mod .47 ( 1975 ) 773 for a review see : c. bagnuls , c. bervillier , phys .rept . 348 ( 2001 ) 91 ; + t. r. morris , prog .. suppl . 131( 1998 ) 395 m. reuter , c. wetterich , nucl .b 417 ( 1994 ) 181 and nucl . phys .b 427 ( 1994 ) 291 ; for an earlier formulation see : m. reuter , c. wetterich , nucl .b 391 ( 1993 ) 147 and nucl . phys .b 408 ( 1993 ) 91 m. reuter , phys .d 53 ( 1996 ) 4430 and hep - th/9511128 ; mod . phys .lett . a 12 ( 1997 ) 2777 and hep - th/9604124 ; + s. falkenberg , b. geyer , phys .d 58 ( 1998 ) 085004 l. f. abbott , nucl . phys .b 185 ( 1981 ) 189 s. l. adler , rev .54 ( 1982 ) 729 c. wetterich , gen .. grav . 30 ( 1998 )159 j. w. york jr ., j. math .( 1973 ) 456 w. souma , prog .phys . 102 ( 1999 ) 181 j. polchinski , nucl .b 231 ( 1984 ) 269 g. parisi , nucl .b 100 ( 1975 ) 368 , nucl . phys .b 254 ( 1985 ) 58 ; k. gawedzki , a. kupiainen , nucl . phys .b 262 ( 1985 ) 33 , phys .54 ( 1985 ) 2191 , phys .55 ( 1985 ) 363 ; b. rosenstein , b. j. warr , s. h. park , phys .( 1991 ) 59 ; c. de calan , p. a. faria da veiga , j. magnen , r. snor , phys .66 ( 1991 ) 3233 s. weinberg in _ general relativity , an einstein centenary survey , _s. w. hawking , w. israel ( eds . ) , cambridge university press , 1979 s. weinberg , hep - th/9702027 l. smolin , nucl .b 208 ( 1982 ) 439 ; nucl . phys .b 148 ( 1979 ) 333 m. reuter , c. wetterich , nucl .b 506 ( 1997 ) 483 and hep - th/9605039 ; + a. h. chamseddine , m. reuter , nucl . phys .b 317 ( 1989 ) 757 j. f. donoghue , phys .72 ( 1994 ) 2996 ; phys .d 50 ( 1994 ) 3874 a. ashtekar , phys .57 ( 1986 ) 2244 ; + a. ashtekar , _ lectures on non - perturbative canonical gravity , _ world scientific , singapore , 1991 ; + for recent reviews see : a. ashtekar , gr - qc/0112038 and math - ph/0202008 a. perez , nucl .b 599 ( 2001 ) 427 ; l. crane , a. perez , c. rovelli , phys .( 2001 ) 181301 ; preprint gr - qc/0104057 k. s. stelle , gen .rel . grav . 9( 1978 ) 353 ; phys . rev .d 16 ( 1977 ) 953 e. tomboulis , phys .70 b ( 1977 ) 361 ; j. julve , m. tonin , nuovo cimento 46 b ( 1978 ) 137 ; e. s. fradkin , a. a. tseytlin , phys .104 b ( 1981 ) 377 ; nucl .b 201 ( 1982 ) 469 ; i. g. avramidi , a. o. barvinsky , phys . lett .159 b ( 1985 ) 269 ; n. h. barth , s. m. christensen , phys .d 28 ( 1983 ) 1876 m. reuter in _ annual report 2000 of the international school in physics and mathematics , tbilisi , georgia _ and hep - th/0012069 l. n. granda , europhys .42 ( 1998 ) 487 ; l. n. granda , s. d. odintsov , phys .b 409 ( 1997 ) 206 s. falkenberg , s. d. odintsov , int . j. mod. phys . a 13 ( 1998 )607 w. souma , gr - qc/0006008 a. a. bytsenko , l. n. granda , s. d. odintsov , jetp lett .65 ( 1997 ) 600 ; + l. n. granda , s. d. odintsov , grav .cosmol . 4 ( 1998 )85 c. wetterich , phys .b 301 ( 1993 ) 90 a. dasgupta , r. loll , nucl .b 606 ( 2001 ) 357 ; + j. ambjrn , j. jurkiewicz , r. loll , nucl .b 610 ( 2001 ) 347 ; + r. loll , nucl .proc . suppl .94 ( 2001 ) 96 ; + j. ambjrn , gr - qc/0201028 j. ambjrn , b. durhuus , t. jonsson , _ quantum geometry - a statistical field theory approach _ , cambridge university press , 1997 o. lauscher , m. reuter , c. wetterich , phys .d 62 ( 2000 ) 125021 and hep - th/0006099 for a comparison and optimization of shape functions see : d. f. litim , phys .b 486 ( 2000 ) 92 ; phys . rev .d 64 ( 2001 ) 105007 ; hep - th/0203006 u. ellwanger , m. hirsch , a. weber , z. phys .c 69 ( 1996 ) 687 ; d. f. litim , j. m. pawlowski , phys .b 435 ( 1998 ) 181 ; hep - th/9901063 ; f. freire , d. f. litim , j. m. pawlowski , phys . lett .b 495 ( 2000 ) 256 ; inta 16 ( 2001 ) 2035 ; d. f. litim , j. m. pawlowski , hep - th/0203005 j .-sumi , w. souma , k .-aoki , h. terao , k. morikawa , hep - th/0002231 h. kawai , m. ninomiya , nucl .b 336 ( 1990 ) 115 m. a. rubin , c. r. ord@xmath1035ez , j. math .phys . 25 ( 1984 ) 2888 ; j. math .( 1985 ) 65 b. allen , phys . rev .d 34 ( 1986 ) 3670 e. s. fradkin , a. a. tseytlin , nucl . phys .b 234 ( 1984 ) 472 t. r. taylor , g. veneziano , nucl . phys .b 345 ( 1990 ) 210 o. lewin , _ polylogarithms and associated functions , _ elsevier , 1981

<S> motivated by recent evidence indicating that quantum einstein gravity ( qeg ) might be nonperturbatively renormalizable , the exact renormalization group equation of qeg is evaluated in a truncation of theory space which generalizes the einstein - hilbert truncation by the inclusion of a higher - derivative term @xmath0 . </S><S> the beta - functions describing the renormalization group flow of the cosmological constant , newton s constant , and the @xmath1-coupling are computed explicitly . </S><S> the fixed point properties of the 3-dimensional flow are investigated , and they are confronted with those of the 2-dimensional einstein - hilbert flow . the non - gaussian fixed point predicted by the latter is found to generalize to a fixed point on the enlarged theory space . in order to test the reliability of the @xmath1-truncation near this fixed point we analyze the residual scheme dependence of various universal quantities ; it turns out to be very weak . </S><S> the two truncations are compared in detail , and their numerical predictions are found to agree with a suprisingly high precision . due to the consistency of the results it appears increasingly unlikely that the non - gaussian fixed point is an artifact of the truncation . </S><S> if it is present in the exact theory qeg is probably nonperturbatively renormalizable and `` asymptotically safe '' . </S><S> we discuss how the conformal factor problem of euclidean gravity manifests itself in the exact renormalization group approach and show that , in the @xmath1-truncation , the investigation of the fixed point is not afflicted with this problem . </S><S> also the gaussian fixed point of the einstein - hilbert truncation is analyzed ; it turns out that it does not generalize to a corresponding fixed point on the enlarged theory space . </S><S> hep - th/0205062 + mz - th/02 - 07 + flow equation of + quantum einstein gravity in a + higher - derivative truncation + _ institute of physics , university of mainz + staudingerweg 7 , d-55099 mainz , germany _ + </S>

1304.5423

the investigations in this paper are focused on the analysis of the low - speed , or low - froude , wave models proposed by @xcite , in which blunt - bodied ships are studied in the context of potential flow and asymptotic expansions in powers of the froude number . as a particularly interesting case that draws our attention, we recall the work of @xcite , who showed that by attaching a bulbous - like obstruction to an otherwise rectangular ship s stern , one could produce a dramatic effect on the production of transverse waves . as they reported in their paper : _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ at this [ froude number ] , a rectangular stern generates waves with steepness @xmath0 , whereas the stern with the downward - pointing bulb [ ] yields waves with steepness @xmath1 .it is clear that the addition of the downward - pointing bulb has had a dramatic effect on the downstream wave steepness , reducing it by a factor of 7.2 , although it has still not eliminated the waves entirely . __ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ our goal is to give an analytical criterion that explains why this phenomenon occurs ; that is to say , what distinguishes the two ships , one with a bulb and one without a bulb , in the context of the ` slow - ship ' approximation ?the advantage of the slow - ship potential - flow approximation , is that allows us to directly relate the generation of waves to the shape of the ship s hull without the need for numerical simulations .in addition to addressing the @xcite issue , we are also interested in a more general question : in the low - speed limit , when a blunt ship is modeled as a two - dimensional semi - infinite body , can it ever be made waveless ?these waveless or wave - minimisation questions in the context of the @xcite approximation were studied by @xcite , @xcite , @xcite , and @xcite for ship hulls of varying geometries , and we are interested in continuing their line of inquiry . in the previous part of our work ( @xcite ) , we demonstrated that piecewise - linear hulls with a single , submerged corner can never be made waveless .for the case of piecewise - linear hulls with multiple corners , the answer to this question is not as clear .for example , are any of the eight hulls presented in figure [ fig : nhull_ships ] waveless ? if not , then which ones produce the smallest waves ? for the case of potential flow over a submerged obstruction ,waveless configurations are certainly possible ,as was demonstrated by @xcite and @xcite , but the same question for surface - piercing ships of general form remains open .certainly , there are notable difficulties in studying this problem .for example , waveless ships were proposed by @xcite and @xcite , but these were later refuted in the more comprehensive numerical study by @xcite , in which they showed that _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ the free surface would at first sight appear to be waveless , but on closer examination of the numerical data , there are very small waves present and they have a steepness of @xmath2 _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ in reference to the bulbous hull in @xcite .our desire , then , is to study these issues in terms of the low - froude asymptotic expansions . ]having progressed through the theory of @xcite , we know that at low froude numbers , the waves generated by a ship become exponentially small and are thus invisible to a regular asymptotic expansion .the ineffectiveness of traditional asymptotics in capturing the low - speed limit was first remarked by @xcite and later termed the _ low - speed paradox_. techniques in exponential asymptotics @xcite allow us to demonstrate the fact that these hidden waves are switched - on when the regular expansion is continued across critical curves ( _ stokes lines _ ) in the complex plane ; this process is known as the _ stokes phenomenon_. most important , however , is the valuable insight that these approximations give : an explicit formula that relates the shape of an arbitrary hull with its resultant waves .the question of wavelessness in the low - froude limit is then simplified to examining whether the sum of the stokes line contributions can ever be zero in regions far from the ship .the requisite background in exponential asymptotics can be found in @xcite .the techniques we apply are based on the use of a factorial over power ansatz to capture the divergence of the asymptotic expansions , then optimal truncation and stokes line smoothing to relate the late - order terms to the exponentially small waves [ see for example , papers by @xcite , @xcite , and @xcite ] .our paper also parallels the works of @xcite and @xcite on the application of exponential asymptotics to the study of gravity or capillary waves produced by flow over a submerged object .it is important for us to mention that the low - froude model of @xcite is indeed a very _ idealised _approximation for understanding the production of ship waves . real stern andbow flows are very complex , and viscosity , turbulence , and necklace vorticies can all play an important role in the production of waves .we refer the reader to , for example , some of the numerical simulations of @xcite and @xcite that demonstrate some of the complex dynamics that arise in ship flows once , for example , vorticity and viscosity are included . in [ sec : discuss ] of this paper , we shall return to discuss the caveats of the low - speed approximation . ultimately , we are interested in obtaining analytical intuition about the connection between the ship s hull and the waves produced .the more usual routes towards analytical solutions assumes an asymptotically small geometry , which leads to the ` thin - ship ' , ` flat - ship ' or ` streamline - ship ' approximations ; in such regimes , a waveless ship is impossible ( see for example , @xcite and krein in @xcite ) , but these theories say very little about the case of non - thin ships .other examples of ship wave models can be studied , including the kelvin - neumann formulation in which the free surface condition is linearised about a steady uniform stream annd the boundary condition on the ship s hull is satisfied exactly , but generally these problems do require a degree of numerical computation .a discussion of such problems can be found in the book by @xcite and the review and discussion by @xcite ( see _ e.g. _@xcite for more recent work on rigorous results applicable to non - slender geometries ) .we finally refer readers to the reviews by @xcite and @xcite for a summary of the role played by low - froude approximations , particularly in connection with problems in which we require asymptotic solutions that preserve the nonlinearity of the geometry .the paper will proceed as follows .first , the mathematical formulation of the ship - wave problem is briefly recapitulated in [ sec : nhull_form ] .this is followed by the asymptotic analysis of the low - froude problem in [sec : nhull_asymptotic ] , which culminates with the derivation of explicit expressions for the wake of an arbitrary multi - cornered ship . from these analytical results , we explain why certain classes of multi - cornered ships can never be made waveless in [ sec : nhull_nonexistence ] , then in [ sec : nhull_numerical ] , these theoretical results are vindicated by comparisons with numerical computations .consider steady , incompressible , irrotational , inviscid flow in the presence of gravity , past the semi - infinite body shown in figure [ fig : nhull_nhull ] , which consists of a flat bottom and a piecewise linear front face .there is a uniform stream of speed @xmath3 as @xmath4 , and we assume that the flow attaches to stern at a stagnation point , @xmath5 and @xmath6 .the dimensional problem can be reposed in terms of a non - dimensional boundary - integral formulation in the potential @xmath7-plane .the unknowns are the fluid speed @xmath8 , and streamline angles , @xmath9 , measured from the positive @xmath10-axis .the body and free - surface are given by the streamline @xmath11 , and we assume the free - surface ( @xmath12 ) attaches to the hull ( @xmath13 ) at @xmath14 . the free - surface , with @xmath11 ,is then obtained by solving a boundary - integral equation , coupled with bernoulli s condition : [ eq : nhulleqtogether ] @xmath15 where @xmath16 is related to the square of the froude draft number with upstream flow @xmath3 , and @xmath17 is defined by @xmath18 , where @xmath19 is the number of corners and @xmath20 is the dimensional value of the potential at each of the corners . in this way , if @xmath21 for @xmath22 denotes the value of the potential at the corners in the non - dimensionalised problem , we have the property that @xmath23 .the derivation of ( [ eq : nhulleqtogether ] ) can be found in @xcite , and the only difference is the choice of scaling for the froude number .we shall refer to the @xmath19-cornered piecewise - linear hull as an @xmath19-hull .for @xmath24 the geometry of the hull can be described by @xmath25 for @xmath26 where @xmath27 , @xmath28 is the stagnation point , and @xmath29 is the exterior angle at the @xmath30th corner ( see figure [ fig : nhull_nhull ] ) .when @xmath31 , we will sometimes also refer to the ship as a @xmath32$]-hull .the @xmath19 corners of exterior angles @xmath33 in the @xmath34 plane ( left ) are mapped to @xmath35 in the complex potential plane ( right ) .the stagnation point is @xmath36 .[ fig : nhull_nhull ] ] in ( [ eq : nhull_bdint ] ) , we write the portion of the boundary integral over the negative real axis as @xmath37 \equiv \log q_s(\phi),\ ] ] where @xmath38 is defined according to the requirement that the free - surface approaches the uniform stream , with @xmath39 and @xmath40 as @xmath41 ; this gives @xmath42 the function @xmath43 in ( [ eq : nhull_rigid ] ) serves to distinguish the different sorts of piecewise - linear ships .note also that the product representation of the complex quantity @xmath43 can be alternately derived by using a schwartz - christoffel mapping applied to the polygonal hull shape and a rigid , flat free surface . as explained in @xcite , in order to study the stokes phenomenon , we must ` complexify ' the free - surface , and thus allow @xmath44 and @xmath45 to be complex functions of the complex potential , @xmath46 . analytically continuing( [ eq : nhull_bdint ] ) and ( [ eq : nhull_bern ] ) gives @xmath47 where the @xmath48 signs correspond to analytic continuation in the upper and lower - half @xmath49 planes , respectively , and @xmath50 denotes the hilbert transform operator on the free - surface , @xmath51a single - cornered ship will always produce exponentially small waves in the low froude limit , @xmath52 ; these waves are explained by the presence of a stokes line which emerges from the singularity at the corner . for a multi - cornered ship , the analysis proceeds almost identically to @xcite , except now , each corner of the hull has the potential to produce stokes lines and its own separate wave contribution . in this section, we shall briefly re - apply the methodology of the previous work , and provide the corresponding formulae for the case of an @xmath19-hull . here , we perform the asymptotic analysis which corresponds to analytically continuing the free - surface into the upper - half @xmath49-plane ; continuation into the lower - half plane produces a complex conjugate contribution , which we add to our results , later .we begin by substituting the usual asymptotic expansions @xmath53 into ( [ eq : nhull_bdint2 ] ) and ( [ eq : nhull_bern2 ] ) ( with the @xmath54 sign ) . in the limit @xmath52, the leading - order solution is the rigid - wall flow of ( [ eq : nhull_rigid ] ) , @xmath55 notice that the leading - order solution , @xmath56 in ( [ eq : nhull_rigid2 ] ) , possesses a singularity at each of the corners , @xmath57 .however , the solution at each subsequent order involves a derivative of the previous order , so we would thus expect that as @xmath58 , the power of the singularity grows , and the asymptotic expansions ( [ eq : nhull_asympt ] ) exhibit factorial over power divergence : @xmath59 where @xmath60 is complex constant , @xmath61 and @xmath62 are functions of the complex potential @xmath63 , and @xmath64 . for most of the analysis , however , we can simply choose one of the corners of interest and add the individual contributions at the end .the singularities , @xmath57 , are located off the free surface , where the hilbert transform in ( [ eq : nhull_bdint2 ] ) is evaluated and so , as justified in @xcite , @xmath65 is exponentially subdominant to the terms on the left - hand side for large @xmath66 . at @xmath67 , ( [ eq : nhull_bdint2 ] ) gives @xmath68 as @xmath69 , and substitution into ( [ eq : nhull_bern2 ] ) gives the relevant terms at @xmath67 : @xmath70 + \biggl [ 2q_0 ^ 2q_0'q_{n-1 } + 2q_0 ^ 2q_1q_{n-2 } ' - i\frac{q_{n-1 } q_1}{q_0}\biggr ] + \ldots = 0 .\label{eq : nhull_oen}\ ] ] we substitute the ansatzes of ( [ eq : nhull_ansatz ] ) into ( [ eq : nhull_oen ] ) , and this yields , at leading order as @xmath71 , @xmath72 using @xmath64 , we integrate this result , to give @xmath73 at the next order as @xmath58 , we find that @xmath74,\ ] ] where @xmath75 is constant , and @xmath76 is any point for which the integral is defined .finally , ( [ eq : nhull_oen_bdint2 ] ) allows us to relate @xmath61 with @xmath77 , using @xmath78 , so that @xmath79.\ ] ] with @xmath62 , @xmath61 , and @xmath77 now determined , we have thus derived the late - orders behaviour in ( [ eq : nhull_ansatz ] ) , subject to the values of @xmath60 and @xmath75 ; these must be determined by applying the method of matched asymptotics near the singularity , @xmath57 . from @xcite , we know that the late - order terms ( [ eq : nhull_ansatz ] ) play a crucial role in determining the free - surface waves . using the expression of @xmath62 in ( [ eq : nhull_chi ] ) ,stokes lines can be traced from each of the ship s corners , across which the stokes phenomenon necessitates the switching - on of waves . from @xcite , these special lines are given by the points @xmath80 where @xmath81 = 0 \text{\quad and \quad } \re[\chi_k(w ) ] \geq 0.\ ] ] if we write @xmath82 near @xmath57 , then from ( [ eq : nhull_rigid2 ] ) , we have @xmath83 and thus from ( [ eq : nhull_chi ] ) , in the limit that @xmath84 , we have @xmath85 ( w+a_k)^{1 + 3\sigma_k}.\ ] ] the condition that@xmath64 thus requires that @xmath86 . in other words , for there to be a singularity , the local deviation of the corner must be greater than @xmath87 .this is a _ necessary _ ( but not _ sufficient _ ) condition for there to be a free - surface wave produced by the corner .in fact , a stronger condition for the existence of a stokes line emerging on the relevant riemann sheet can be derived . since @xmath88 , we can write @xmath89 for analytic continuation into the upper - half plane , where @xmath90 is the angle of the hull as @xmath91 , measured from the positive @xmath10-axis ( shown in figure [ fig : nhull_nhull ] ). if we write @xmath92 , then stokes lines must leave at angles of @xmath93 for @xmath94 and we thus need @xmath95 in order for the line to emerge in the upper - half plane . the general requirements for a stokes line to intersect the free - surface is a global function of the leading - order flow , but for most hulls , the requirement that @xmath95 with ( [ eq : nhull_stokescond ] ) is adequate . in any case , we will let @xmath96 denote those corners which have stokes lines crossing the free - surface . as an example , consider figure [ fig : nhull_stokes ] , which illustrates the stokes lines for various @xmath19-hulls , including a simple @xmath97-hull , the @xmath98-hull of @xcite , a bulbous @xmath99-hull , and a step - like @xmath100-hull . with the exception of a single configuration, the condition that a stokes line emerges into the upper - half plane is enough to guarantee that it intersects the free - surface .the exception is with the @xmath98-hull , for which the second singularity has a stokes line emerging at an angle of @xmath101 , but which does not later encounter the free - surface .-hull , farrow and tuck s @xcite @xmath98-hull , the @xmath100-hull , and the @xmath99-hull shown before in figure [ fig : nhull_nhull ] . for the @xmath97-hull and @xmath99-hull ,the corner angles diverge at @xmath102 ; for the remaining hulls , the corner angles are all rectangular . [fig : nhull_stokes],title="fig : " ] -hull , farrow and tuck s @xcite @xmath98-hull , the @xmath100-hull , and the @xmath99-hull shown before in figure [ fig : nhull_nhull ] . for the @xmath97-hull and @xmath99-hull ,the corner angles diverge at @xmath102 ; for the remaining hulls , the corner angles are all rectangular .[ fig : nhull_stokes],title="fig : " ] -hull , farrow and tuck s @xcite @xmath98-hull , the @xmath100-hull , and the @xmath99-hull shown before in figure [ fig : nhull_nhull ] . for the @xmath97-hull and @xmath99-hull ,the corner angles diverge at @xmath102 ; for the remaining hulls , the corner angles are all rectangular .[ fig : nhull_stokes],title="fig : " ] -hull , farrow and tuck s @xcite @xmath98-hull , the @xmath100-hull , and the @xmath99-hull shown before in figure [ fig : nhull_nhull ] .for the @xmath97-hull and @xmath99-hull , the corner angles diverge at @xmath102 ; for the remaining hulls , the corner angles are all rectangular . [fig : nhull_stokes],title="fig : " ] to derive the form of the exponentials that appear whenever a stokes line intersects the free - surface , we optimally truncate the asymptotic expansions ( [ eq : nhull_asympt ] ) , and examine the remainder as the stokes line is crossed .we let @xmath103 with a similar expression for the series for @xmath104 .when @xmath105 is chosen to be the optimal truncation point , the remainder @xmath106 is found to be exponentially small , and by re - scaling near the stokes line , it can be shown that a wave of the following form switches on : @xmath107.\ ] ] to ( [ eq : nhull_finalpre ] ) , we must also include the complex conjugate due to the contributions from analytic continuation of the free - surface into the lower - half @xmath49-plane [ see ( [ eq : nhull_bdint2 ] ) ] .the sum of the two contributions is then @xmath108 \biggr\},\ ] ] with , of course , one such expression for every @xmath109 .thus , for any given arbitrary @xmath19-hull with a geometry such that @xmath110 is nonempty , the appearance of exponentially small waves is a _ necessary consequence _ of the divergent low - froude problem ; in order to check that such a ship can never be waveless , we need only verify that the sum of all the contributions incurred can never be zero , so that there is a non - zero wave amplitude far downstream .the constants @xmath60 and @xmath75 , which appear in the final form of the waves ( [ eq : nhull_final ] ) [ the latter as a prefactor in @xmath61 in ( [ eq : nhull_q ] ) ] , can be determined by re - scaling @xmath63 and @xmath111 near each of the singularities , and then matching the leading - order nonlinear ( inner ) solutions with the late - order ( outer ) terms of ( [ eq : nhull_ansatz ] ) .it can be shown ( see ( 6.8 ) in @xcite ) that @xmath112 and @xmath113,\ ] ] where @xmath114 is given by @xmath115 the terms @xmath116 are given by the recurrence relation , @xmath117 we will often make reference to the limiting ratio in ( [ eq : nhull_lambdak ] ) , so we define the function : @xmath118 the value of @xmath119 only depends on the local divergence of the @xmath120 corner , and its values are given in @xcite .since @xmath121 for all choices of the local angle @xmath122 , @xmath75 is also non - zero and this verifies that each of the @xmath123 corners of an @xmath19-hull must necessarily generate a non - zero wave on the free surface .with @xmath61 given by ( [ eq : nhull_q ] ) , @xmath75 given by ( [ eq : nhull_lambdak ] ) and ( [ eq : nhull_omega_k ] ) , and @xmath124 from ( [ eq : argck ] ) , we have from ( [ eq : nhull_final ] ) the result that @xmath125 \exp\left[-\frac{\re(\chi_k)}{\epsilon}\right ] \times \\ \cos\left[-\frac{\im(\chi_k)}{\epsilon } + \frac{\pi}{2}+ \frac{\pi \gamma_k}{2 } + ( 6 - 3 \gamma_k)\theta_k + \re\left ( 3 \int_{-a_k}^w \frac{\mathscr{h}\theta_1}{q_0 ^ 3}\ , { \rm d}{\phi}\right)\right].\label{eq : nhull_qexpk } \end{split}\ ] ] then , for each @xmath126 , we add the waves together , so that the total wave contribution after all the stokes lines have been crossed is @xmath127let us see what would be needed to produce a waveless ship .the wave contributions ( [ eq : nhull_qexpk ] ) are written in terms of different denominators @xmath62 ( also referred to as the ` _ _ singulants _ _ ' , _ c.f ._ @xcite ) . to make it easier to sum them we rewrite them in terms of the single singulant , @xmath128 .note that @xmath129 where in order for the integral to exist , we may have to avoid the intermediate corners by deforming the contour into the upper half plane .consider now @xmath130 in ( [ eq : nhull_qexpk ] ) when @xmath63 is evaluated on the free - surface , that is , for @xmath131 . from the appendix of @xcite, it was shown that @xmath132 = q_0 ^ 3(w ) \ , { \rm e}.\ ] ] moreover , the real part of @xmath133 comes from the pole at infinity , giving @xmath134 putting these together in ( [ eq : nhull_qexpk ] ) we find , for @xmath131 , @xmath135 \cos\left[-\frac{\im(\chi_1(w))}{\epsilon}+ \re\left ( 3 \int_{-a_1}^w \frac{\mathscr{h}\theta_1}{q_0 ^ 3}\ , { \rm d}{\phi}\right ) + \psi_k \right],\end{aligned}\ ] ] where the dependence on @xmath30 arises only through the constants @xmath136 \exp\left[-\frac{1}{\epsilon}\,\im\left(\int_{-a_1}^{-a_k } \frac{{\rm d}{\phi}}{q_0 ^ 3}\right ) \right],\\ \psi_k & = & \frac{1}{\epsilon}\,\re\left ( \int_{-a_1}^{-a_k}\frac{{\rm d}{\phi}}{q_0 ^ 3 } \right)- \re\left ( 3 \int_{-a_1}^{-a_k } \frac{\mathscr{h}\theta_1}{q_0 ^ 3}\ , { \rm d}{\phi}\right ) + \frac{\pi}{2}+ \frac{\pi \gamma_k}{2 } + ( 6 - 3 \gamma_k)\theta_k.\end{aligned}\ ] ] we have already shown in @xcite that a single - cornered hull must produce waves .therefore let us consider first the next simplest case of a 2-hull .for such a ship to be waveless , both corners must generate stokes lines which intersect the free surface , and the waves generated by each must exactly cancel ; then , there will be a finite section of free surface containing waves , but no wavetrain at infinity ( as happens in the case of capillary waves in @xcite ) . in order for the waves from the two corners to cancel ,we need @xmath137 .now for a fixed value of @xmath138 this may indeed be possible ( and we give such an example in [ sec : nhull_numerical ] ) , but what if we want the waves to vanish for _ all _ ( small ) values of @xmath138 ? then , since each @xmath75 is of the form @xmath139 as @xmath140 , we need the exponentials to be equal , the powers of @xmath138 to be equal , and the prefactors to be equal . in order for the exponentials to be equal, we require @xmath141 and since @xmath142 the only way ( [ 1 ] ) can hold is if @xmath143 , so that @xmath144 is real for @xmath145 . now , for the algebraic factors of @xmath138 to be equal , we require @xmath146 , which implies @xmath147 . thus the only possibility for a waveless 2-hull is for a ship with @xmath148 . to eliminate this final possibility we need to consider the prefactors .since @xmath149 and @xmath144 is real for @xmath145 , then @xmath150 the only remaining difference between the two prefactors is in @xmath151 .however , since @xmath152 the only way that we can have @xmath153 is if @xmath154 .thus the two prefactors must be different , one corner always dominates the other one , and the waves can never cancel .waveless ships with two corners are not possible .what can we say about more general ships ?let us take a general @xmath19-hull with the following assumptions : suppose that all the stokes lines intersect the free surface , that @xmath155 for each @xmath30 , and that @xmath156 . in figure 1 , hulls ( a ) to ( e )satisfy this requirement , whereas hulls ( f ) to ( h ) do not . under these assumptions, @xmath157 is monotonically increasing with @xmath30 , so that @xmath158 is monotonically decreasing .thus the argument of the exponential @xmath159 is convex in @xmath30 , increasing while @xmath160 and then decreasing while @xmath161 .thus if @xmath162 for all @xmath163 then we can see immediately that the waves generated at the corner @xmath30 such that @xmath164 exponentially dominate all the others . on the other hand , if @xmath165 then the waves from corners @xmath30 and @xmath166 have the same exponential factor ( as in the 2-hull case ) .if we further impose @xmath167 then they have the same algebraic factor , and there is possibility of wave cancellation if the prefactors are equal . of course , even if we could get the prefactors to be equal , we still have to worry about the waves generated by all the other corners .in fact , even from these two corners there would be higher - order correction terms ( both in the form @xmath168 , @xmath169 , _ etc . _ and also in the form of a trans - series @xmath170 , @xmath171 _ etc .. thus it does not seem to be worth pursuing the analysis further .however , even if we can not get the waves to cancel exactly , we might expect a significant reduction in the amplitude of the waves in the case when leading - order cancellation occurs .this brings us to the natural question of whether the analysis we have presented may be used to design a hull to minimise the wave drag .before we address this question , let us first demonstrate that the hulls shown in figure 1 ( f ) through ( h ) must generate waves on the free surface .we consider them in reverse order .the 9-hull shown in figure 1(h ) has @xmath172 ( as shown in figure 3 ) , with @xmath158 alternating between zero and @xmath173 ; thus the argument above can be used to show that the contribution from @xmath174 exponentially dominates all the others . the hull shown in figure 1(g )has @xmath175 , with three positive angles @xmath176 and one negative angle .thus the algebraic factor in the contribution from @xmath177 is different to the others , and those waves must always exist on the free surface . finally let us consider the 3-hull shown in figure 1(f ) , which is found in the work of @xcite , and for which the addition of a downward pointing bulb was shown to dramatically reduce the wave resistance compared to a rectangular ship . in this case ,@xmath178 , and @xmath179 is the only relevant corner , so there are always waves on the free surface . the principal effect of the bulb is to lower the usual singularity farther away from the free surface , thereby decreasing the amplitude of the waves .this last example not only highlights the difficulty in trying to minimise the wave drag , but also the advantage of our semi - analytic approach : we have gained considerable insight into the mechanism of wave production ; from this , we can immediately see why farrow & tuck obtained the results that they did . in trying to design reduced - wave hulls , it is crucial to specify what exactly is the optimisation process .for example , if we simplify the hull of farrow & tuck to a 2-hull by reducing the width of the downward pointing bulb to zero , then we have one parameter , @xmath174 , over which we can optimise ( since @xmath180 ) .we find the smallest waves occur for @xmath181 , _i.e. _ when there is no bulb .however , in our current nondimensionalisation , as we vary @xmath174 , both the depth of the hull and that of the bulb vary .if instead we fix the depth of the hull , and allow the depth of the bulb to vary , we find that the waves get smaller as the bulb gets deeper .on the other hand , if we fix the depth of the bulb , and allow the depth of the hull to vary , we again find that the smallest waves correspond to the hull depth being equal to the bulb depth , _i.e. _ to there being no bulb .the numerical algorithms developed in @xcite can be used to verify the asymptotic predictions . here, we focus on the particular case of a @xmath97-hull , which we refer to as a @xmath182$]-hull ; this is a ship with divergent corner - angles @xmath183 and @xmath184 , and with leading - order flow given by ( [ eq : nhull_rigid ] ) , or @xmath185 with @xmath186 .as we discussed in @xcite , the numerical computation of the stern problem at small values of @xmath138 can be particularly difficult , and the culprit is the presence of the attachment singularity at @xmath187 , associated with a small boundary layer ; this singularity is responsible for most of the numerical error . for hulls where the in - fluid attachment angle between the free - surface and body is less than @xmath188 , a simple finite difference scheme based on the methods outlined in @xcite can be used , provided that we limit our search to waves larger than @xmath189 .figure [ fig : nhull_specsol ] provides an example of solutions found using this method .$ ] -hull ( dashed line ) and @xmath190$]-hull ( solid line ) at @xmath191 and @xmath192 , respectively .both ships have corners set at @xmath193 and @xmath194 .the solutions were computed using algorithm a of @xcite with @xmath195 and @xmath196 for the former ship and @xmath197 and @xmath196 for the latter . [fig : nhull_specsol ] ] the theory of [ sec : nhull_asymptotic ] and [ sec : nhull_nonexistence ] can be verified by comparing the analytical predictions with numerically computed wave amplitudes far from the ship .first , consider the effect of varying the froude number on ships of a fixed geometry .this is shown in figure [ fig : nhull_num_final ] for three @xmath97-hulls with their corners fixed with @xmath193 and @xmath194 .the individual cosine waves are calculated from ( [ eq : nhull_qexpk ] ) with @xmath198 downstream , and then the final amplitude computed using the sum ( [ eq : nhull_qexpsum ] ) ( see 4.5 in @xcite for additional details ) . the match between numerical and asymptotic solutions is quite good , and like the previous work , we remark that the results are applicable over a wide range of froude numbers . and @xmath194 .the solutions were computed using algorithm a of @xcite .the parameters used were the following : @xmath195 , @xmath199 ; @xmath200 , @xmath201 ; and @xmath197 , @xmath202 . [fig : nhull_num_final ] ] .the ship is a @xmath203$]-hull with @xmath186 and @xmath204 .the dotted line indicates the one - cornered approximation for a rectangular hull .[ fig : nhull_cc2525.1 ] ] next , we would like to consider the effect of fixing the froude number , but varying the positions of the corners .the problem , however , is that the interesting effects of this procedure are only seen at values of @xmath138 much smaller than what we can achieve using the above numerical methods . in appendix[ sec : cc_simpnon ] , we present a slightly simplified version of the ship - wave problem ( [ eq : nhull_bdint2])([eq : nhull_bern2 ] ) that preserves the asymptotic structure of the waves , but also enables us to compute numerical solutions to much higher accuracy . this simplified problem was used to create figure [ fig : nhull_cc2525.1 ] , which shows the effect of varying the positions of the corners on the wave amplitude for a @xmath203$]-hull with @xmath204 , @xmath186 , and values of @xmath205 .the figure contains a number of interesting effects : the first is that the numerically computed wave amplitude shows a significant dip ( an order of magnitude ) near @xmath206 , and that this effect is also captured fairly accurately in the asymptotic solution .the reason for the dip is that at this value of @xmath174 , the waves from the two corners exhibit partial destructive interference .however , from the of the previous section , we know that the waves generated by the corner at @xmath207 should exponentially dominate those from @xmath208 .how then are they cancelling each other ?the answer is that the prefactor @xmath209 is over ten times larger than @xmath210 ; at this particular value of @xmath138 , the difference in prefactors is enough to compensate for the difference in exponentials , since @xmath211 is not very large .thus , the value of @xmath174 at which cancellation occurs depends on @xmath138 ; for somewhat smaller of @xmath138 the stokes line from @xmath212 does indeed dominate and no cancellation is possible .this indicates that it should be possible to design hulls with reduced wave drag at a particular speed ( froude number ) .it is also reassuring to note that including the leading - order term from each of our exponentially - small waves captures the behaviour of the solution very well , even though formally , one of the terms is exponentially subdominant . the second effect illustrated by figure [ fig : nhull_cc2525.1 ] is the divergence between our asymptotic expansion and the numerical solution for values of @xmath174 close to @xmath213 .the reason for the divergence is that the prefactors @xmath214 and @xmath215 are singular as the corners approach each other ; our analysis in [ sec : nhull_asymptotic ] , in fact , implicitly assumes that the corners of the ship are spaced sufficiently far from one another . to be more specific , in order to determine the constants @xmath60 and @xmath75 in ( [ eq : nhull_gammak ] ) and ( [ eq : nhull_lambdak ] ) in the previous analysis , the outer solution was required to match a nonlinear inner solution .the size of this inner region can be derived by observing where the breakdown in the outer expansion ( [ eq : nhull_asympt ] ) first occurs , _i.e. _ where @xmath216 . from ( [ eq : nhull_rigid2 ] ) and ( [ eq : nhull_q1 ] ) , the required re - scaling of @xmath63 near a singularity at @xmath217 can be seen to be @xmath218 thus , if two ( or more ) singularities are spaced within a distance of ( [ eq : nhull_innerscaling ] ) apart , then the previous asymptotic methodology breaks down as @xmath52 .if we combine the two corners into a single corner of angle @xmath219 , we find the wave amplitude is given by the dotted line in figure [ fig : nhull_cc2525.1 ] .this approximation clearly works well for @xmath174 close to 0.5 .a uniform approximation would need to smoothly match with the one - cornered approximation at one end , and the ( separated ) two - cornered analysis at the other , and thus bridges the dashed and dotted lines in figure [ fig : nhull_cc2525.1 ] .such an approximation requires us to consider the distinguished limit in which the corners of the ship are allowed to approach one another as @xmath52 .this is similar to the situation in ( * ? ? ?* appendix b ) where the asymptotic solutions for flow over a rectangular step in a channel was analysed in the case where both corners of the step lie in the same ` inner region ' . for the case of the multi - cornered ship ,the details of this analysis are very technical , and will be published in a future paper ._ in the end , what is the definitive answer to the conundrum of existence and non - existence of waveless ships ?_ unlike our results for the single - cornered ships of @xcite , there does not seem to be a simple answer to this question , applicable to the most general piecewise - linear hulls . despite this, however , we have offered several new insights into the study of ship - wave resistance : we have offered explicit formulae for the computation of waves given the shape of the ship s hull ; we have offered simple interpretations of the production of such waves in terms of stokes line crossings and the stokes phenomenon ; and perhaps most importantly , we offered a methodology which , given specific ships , provides an immediate and intuitive understanding of the effect of the body on the free - surface . in the previous work, we highlighted the importance of distinguishing between local and global properties of the analysis .consider the factorial over power divergence of the asymptotic series in ( [ eq : nhull_ansatz ] ) , or the emergence - conditions of stokes lines in ( [ eq : nhull_stokescond ] ) , or the numerically - determined pre - factor , @xmath220 , in ( [ eq : nhull_omega_k])these are all _ local _ properties of the problem ; indeed their derivation only depends on the behaviour of the asymptotic solution near the relevant singularities .these local properties , we understand well . in contrast ,global _ questions remain unanswered . for example , given a ship , represented by @xmath56 , what are the necessary and sufficient conditions for stokes lines to intersect the free surface ? or perhaps more difficult : what are the necessary restrictions on @xmath56 so that total phase cancellation occurs ?we have provided a few preliminary results on this global problem , but a more exhaustive analysis of these issues remains an open problem .naturally , our study of ship waves would be incomplete without a theory applicable for smooth hulls , with the eventual goal of addressing the well - known technique of using a bulb to reduce the wave resistance of a ship @xcite .however , the difficulty here is that analytic continuation is an ill - posed process and small perturbations in the shape of a hull can have large effects on the associated singularities a unified theory for arbitrary ship geometries will likely prove difficult , if not downright impossible .perhaps , then , we should only consider specific classes of smooth ships .ship waves associated with continuous geometries have been considered in the numerical work of @xcite , @xcite , and @xcite , where there , the hulls are specified by piecewise - entire functions .for example , @xcite consider the family of hulls given by @xmath221 which , given parameters @xmath222 and @xmath223 , provides a ship consisting of a horizontal bottom and a vertical line , joined by a rounded section ; @xmath224 yields rounded corners and @xmath225 yields bulbous sterns .the key is that if we restrict ourselves to classes of ships given by piecewise entire functions , then the complex singularities must be located at the points joining each piece . as a simpler example, we may consider the ship with @xmath226 & \text{for $ w\in(-a_1,-a_2)$ } \\\frac{\pi}{2 } & \text{for $ w\in(-a_2,0)$ } , \end{cases}\ ] ] which is similar to the vertically - faced one - cornered ships studied previously , but with a rounded edge .analysis of the stokes lines shows that the relevant line emerges from @xmath227 ; this is shown in figure [ fig : remarks_smhull ] .the study of these piecewise - entire ships is the subject of ongoing investigation . ) with @xmath193 and @xmath228 .the stokes line leaves tangentially along the boundary , but later intersects the free - surface .[ fig : remarks_smhull ] ] a similar direction for research is towards the development of a low - froude asymptotic theory for flows past three - dimensional , full - bodied ships .this builds upon the works of , for example , @xcite and @xcite , who apply geometrical ray theory to the case of streamline ( thin ) ships . in theory , the interpretation we have presented in this paper of free - surface waves arising due to stokes line crossings is still valid in three dimensions , except now , singularities are associated with stokes _ surfaces _ rather than _lines_. in practice , however , the analysis is complicated due to the loss of complex variable techniques .we refer the reader to the work of @xcite , which provides a first step towards extensions of exponential asymptotics to partial differential equations .in addition to our study , which solely focuses on the low - froude model of @xcite , it is important for us to question the place of these simplified mathematical models in terms of the bigger picture : that which includes the effects of vorticity , viscosity , and time - dependence in ship - wave interactions .as we elucidated in the introduction , numerical work ( as particular examples , see @xcite and @xcite ) show that in practice , these neglected effects can have significant roles in the production of waves .extended discussions of the role of low - froude theories appear in @xcite ( @xcite ) , @xcite , and @xcite .thus , while it is certainly true that in order to obtain _ analytical _ approximations directly relating ship geometries to free - surface waves , the low - froude approximation provides enormous simplification , we hope that similar analytical theories can be developed which include a more complete host of effects .37 natexlab#1#1 1976 wave breaking resistance of ships . in _ proc .seminar on wave resistance , tokyo _ , pp .1998 _ weakly nonlocal solitary waves and beyond - all - orders asymptotics_. kluwer academic publishers . 1985 a quasi - linear free surface condition in slow ship theory ._ schiffstechnik bd . _ * 32 * , 2541 .1998 exponential asymptotics and stokes lines in nonlinear ordinary differential equations .lond . a _ * 454 * , 27332755 .2005 exponential asymptotics and stokes lines in a partial differential equation .r. soc . a _ * 461 * ( 2060 ) , 23852421 .2002 exponential asymptotics and capillary waves ._ siam j. appl ._ * 62 ( 6 ) * , 18721898 .2006 exponential asymptotics and gravity waves ._ j. fluid mech . _ * 567 * , 299326 .1969 bow waves before blunt ships .rep._. office of naval research , department of the navy .1972 two - dimensional free - surface gravity flow past blunt bodies ._ j. fluid mech . _ * 51 ( 3 ) * , 529543 .1973 _ asymptotic expansions : their derivation and interpretation_. academic press , london .1995 further studies of stern wavemaking ._ j. austral .b _ * 36 * , 424437. 1989 nonlinear free - surface flow at a two - dimensional bow ._ j. fluid mech . _ * 209 * ( 1 ) , 5775 . 2012a note on waveless subcritical flow past a submerged semi - ellipse ._ journal of engineering mathematics _1979 the ray theory of ship waves and the class of streamlined ships ._ j. fluid mech . _ * 91 * , 465487 .1968 _ theory of ship waves and wave resistance_. iowa city : effective communications inc .1964 a sequence of submerged dipole distributions whose wave resistance tends to zero . _ j. math .* 13 * , 693700 .2002 _ linear water waves : a mathematical approach_. cambridge university press .2012 free surface flow past topography : a beyond - all - orders approach .j. appl . math ._ * 1 * ( 1 ) , 127 . 1988 splashless ship bows with stagnant attachment . _ j. ship .res . _ * 32 * ( 3 ) , 194202 .1986 ship bows with continuous and splashless flow attachment ._ j. austral .b _ * 27 * ( 442 - 452 ) . 1991the quest for a three - dimensional theory of ship - wave interactions [ and discussion ] .lond . a _ * 334 * ( 1634 ) , 213227 .1968 wave resistance : the low speed limit . _ tech .rep._. michigan university , ann arbor .1970 singular perturbation problems in ship hydrodynamics . _ tech .rep._. michigan university , ann arbor .1995 stokes phenomenon and matched asymptotic expansions ._ siam j. appl .* 55(6 ) * , 14691483 .2004 the subcritical motion of a semisubmerged body : solvability of the free boundary problem ._ siam j. math .* 36 * ( 1 ) , 6993 .2010 _ asymptotic methods in fluid mechanics : survey and recent advances _ , chap .exponential asymptotics and stokes line smoothing for generalized solitary waves , pp. 121126 .springerwiennewyork .2010 exponential asymptotics and free - surface flows .phd thesis , university of oxford .2013 new gravity - capillary waves at low speeds .part 1 : linear theory ._ j. fluid mech . _ * ( in production)*. 2013 new gravity - capillary waves at low speeds .part 2 : nonlinear theory ._ j. fluid mech . _ * ( in production)*. 2011 do waveless ships exist ? results for single - cornered hulls . _ j. fluid mech ._ * 685 * , 413439 . 1991 ship - hydrodynamic free - surface problems without waves ._ j. ship res ._ * 35 * ( 4 ) , 277287 . 1991 waveless solutions of wave equations . in _proceedings 6th international workshop on water waves and floating bodies_. wood s hole , mass .: m.i.t .1984 splashless bow flows in two - dimensions . in _ proc .15th symp .naval hydrodynamics_. hamburg , germany : national academy press .2005 reminiscences and reflections : ship waves , 1950 - 2000 ._ j. ship res . _ * 49 * ( 4 ) , 238246 .1978 divergent low - froude - number series expansion of nonlinear free - surface flow problems .a _ * 361 * , 207224. 1977 computation of near - bow or stern flows using series expansion in the froude number . in _2nd internatinal conference on numerical ship hydrodynamics_. berkeley , california : university of california , berkeley . 1997 viscosity and surface - tension effects on wave generation by a translating body ._ journal of engineering mathematics _ * 32 * ( 2 ) , 257280 .the full problem in ( [ eq : nhull_bdint2])([eq : nhull_bern2 ] ) can be studied using the methods we develop here , but can also work with a simpler problem that nevertheless contains all of the the key components .the reason for this simplification is that , in order to verify the asymptotic analysis in the regime where the ship s corners are closely spaced , wave amplitudes must be computed to five or six digits of precision otherwise , the fine effects of adjusting the ship s geometry are easily missed ; this precision can only be easily achieved for the simpler problem , which we now derive . as we know, when exponentially small terms are sought from ( [ eq : nhull_bdint2 ] ) , the integral term , @xmath229 , only serves a minor role throughout the analysis .if we return to the derivation of the late - orders ansatz ( [ eq : nhull_ansatz ] ) , we recall that the subdominance of @xmath229 as @xmath58 ensures that it plays no part in the derivation of @xmath62 .in fact , the only role of the hilbert transform is to change the expression for @xmath230 in ( [ eq : nhull_q1 ] ) .consequently , in the final form of the waves @xmath130 in ( [ eq : nhull_qexpk ] ) , the presence of @xmath231 only serves to change the amplitude coefficient and the phase shift by an @xmath232 amount .therefore , the salient features of the problem can still be retained if we use @xmath233 instead of ( [ eq : nhull_bdint2 ] ) ; this way , we simplify the full problem in ( [ eq : nhull_bdint2 ] ) to ( [ eq : nhull_bern2 ] ) to a single nonlinear differential equation in @xmath111 . analytic continuation into the upper - half plane , and substituting @xmath234 into ( [ eq : nhull_bern2 ] ) gives @xmath235 = 0,\ ] ] as a simplified nonlinear model of ship waves .simplifications of the boundary integral problem ( [ eq : nhull_bdint2 ] ) and ( [ eq : nhull_bern2 ] ) were also proposed in @xcite , but there , the simplifications were argued based on behavioural requirements . here , ( [ eq : nhull_simpbern ] ) is a justified reduction based on the @xmath52 limit .notice that in this new problem , we chose to analytically continue into the upper half-@xmath63-plane , and thus the exponentially small waves of ( [ eq : nhull_simpbern ] ) will possess both a real and imaginary part . if we wish , we can mirror the analysis for the lower half-@xmath63-plane and add the complex conjugate as we did before for ( [ eq : nhull_final ] ) .however , it is somewhat simpler to examine ( [ eq : nhull_simpbern ] ) as a problem on its own ; thus we shall only concern ourselves with studying the real component of the solution to ( [ eq : nhull_simpbern ] ) , which we write @xmath239 .now instead of ( [ eq : nhull_qexpk ] ) , the form of the waves for the simplified problem ( with well - separated corners ) is given by @xmath240 \times \\ \cos\left[-\frac{\im(\chi_k)}{\epsilon } + \frac{\pi \gamma_k}{2 } + ( 6 - 3 \gamma_k)\theta_k \right],\end{gathered}\ ] ] which is effectively ( [ eq : nhull_qexpk ] ) with @xmath241 and without a phase shift of @xmath219 .the reduction by a factor of @xmath97 in ( [ eq : nhull_cc_qexpk ] ) compared to ( [ eq : nhull_qexpk ] ) occurs because there is no need to add the complex conjugate wave contribution .analytical and numerical results for the simplified nonlinear problem of ( [ eq : nhull_simpbern ] ) in the context of a one - cornered ship can be found in @xcite , whereas we have already discussed the numerical solution of the @xmath203$]-hull for the simplified problem in [ sec : nhull_numerical ] and figure [ fig : nhull_cc2525.1 ] .

<S> in the low - speed limit , a blunt ship modeled as two - dimensional semi - infinite body with a single corner _ can never be made waveless_. this was the conclusion of the previous part of our work in @xcite , which focused on the @xcite model of ship waves in the low speed limit . in this accompanying paper </S><S> , we continue our investigations with the study of more general , piecewise - linear , or multi - cornered ships . </S><S> the low - speed or low - froude limit , coupled with techniques in exponential asymptotics allows us to derive explicit formulae relating the geometry of the hull to the form of the waves . </S><S> configurations with closely spaced corners present a non - trivial extension of the theory , and we present the general methodology for their study . </S><S> lastly , numerical computations of the nonlinear ship - wave problem are presented in order to confirm the analytical predictions . </S><S> surface gravity waves , wave - structure interactions </S>

astro-ph0206130

subdwarf b ( sdb ) stars were originally defined by sargent and searle as stars with colours corresponding to those of b stars in which the balmer lines are abnormally broad compared to those seen in population i main - sequence stars .subdwarf o ( sdo ) stars and subdwarf ob ( sdob ) stars are related stars of correspondingly earlier spectral type ( see , e.g. , vauclair & liebert 1987 ) . based on an interpretation of their evolutionary state , sdb stars are also sometimes referred to as extreme horizontal branch stars .they are generally considered to be core helium burning stars with extremely thin hydrogen envelopes ( @xmath4 ) , and most of them are believed to have masses around @xmath5 @xcite .indeed , a recent asteroseismological analysis by brassard et al .( 2001 ) has confirmed a mass of 0.49@xmath6 for the sdb star pg 0014 + 067 . in this paper, we collectively refer to core helium burning stars with thin hydrogen envelopes as sdb stars if they are located in the corresponding region in a ( @xmath7 , @xmath3 ) diagram , even if some of them may in reality be sdo or sdob stars .subdwarf b stars form an important class of objects in several respects . at the galactic level , they are the dominant population in surveys of blue objects @xcite and constitute a population of stars that are important for our understanding of the structure and evolution of the galaxy .pulsating sdb stars @xcite provide a standard candle for distance determinations . on a larger cosmological scalethey have been used to constrain the ages of the oldest galaxies and hence cosmological models .the latter is based on measuring the age of giant elliptical galaxies from the ultraviolet ( uv ) excess , or `` upturn '' , with the help of evolutionary population synthesis models where low mass core - helium burning stars provide the dominant source of uv radiation @xcite . more importantly , sdb stars are exotic objects because of their thin hydrogen - rich envelopes . understanding the process of their formation helps to improve our understanding of the theory of stellar and binary evolution. there have been extensive surveys of sdb stars in the past .magnitude - limited and colour - selected samples have been obtained from the palomar green ( pg ) survey @xcite ( magnitude limit @xmath8 ) and the kitt peak downes ( kpd ) survey @xcite ( magnitude limit @xmath9 ) .saffer et al . ( 1994 ) measured atmospheric parameters , such as effective temperature , surface gravity and photospheric helium abundance , for 68 sdb stars .ferguson , green & liebert found 19 sdb stars with main sequence ( ms ) companions from the pg survey and derived a binary frequency of about 50 percent .allard et al . found 31 sdb binaries from 100 candidates chosen from the pg and the kpd surveys and estimated that 54 to 66 percent of sdb stars are in binaries with ms companions after taking selection effects into account .thejll , ulla & macdonald and ulla & thejll also found that more than half of their sdb star candidates showed infrared flux excesses , indicating the presence of binary companions .aznar cuadrado & jeffery obtained atmospheric parameters for 34 sdb stars from spectral energy distributions and found that 19 were binaries with ms companions , while 15 appeared to be single .these observations showed that at least half of the sdb stars were in binaries .a major recent development has been the identification of many sdb stars as short - period binaries @xcite . in particular , maxted et al . concluded that more than two thirds of their candidates were binaries with short orbital periods from hours to days and that most of the known companions were white dwarfs ( wds ) .a variety of formation channels for sdb stars have been proposed in the past but mainly for single sdb stars because of the absence of identified sdb star binaries at the time . in the merger channel , two helium white dwarfs in a close binaryare driven together by the orbital angular momentum loss due to gravitational wave radiation . when the white dwarfs merge and the merged object ignites helium , this produces a single sdb star @xcite . alternatively ,stellar wind mass loss near the tip of the first giant branch ( fgb ) may strip off a giant s envelope and leave an almost bare helium core .if helium is ignited in the core , the star will appear as a single sdb star @xcite .sweigart has studied the evolution of globular - cluster stars and suggested that helium mixing driven by internal rotation substantially increases the helium abundance in the envelope ; this may lead to enhanced mass loss along the fgb and the formation of a sdb star . on the other hand , mengel , norris & gross carried out conservative binary evolution calculations for a binary system with initial masses of @xmath10 and @xmath11 and a composition @xmath12 , @xmath13 , and showed that there exists a range of initial separations for which stable mass transfer can produce an sdb star of @xmath14 in a wide binary . from a binary evolution point of view , these formation channels are not complete .when a star fills its roche lobe near the tip of the fgb , mass transfer begins and may be dynamically unstable .this leads to the formation of a common envelope ( ce ) @xcite , where the ce engulfs the helium core and the secondary .due to friction between the envelope and the immersed binary , the orbit shrinks , depositing a large amount of orbital energy in the envelope .if this energy is enough to eject the envelope and if helium is subsequently ignited in the core , a sdb star in a short - period binary is formed with a mass near @xmath5 .these are exactly the types of objects identified in large numbers by maxted et al . .if mass transfer near the tip of the fgb is dynamically stable , the envelope of the primary is lost as a result of stable rlof , and the remnant core will be in a binary system with a long orbital period .it becomes a sdb star when helium in the primary s remnant is ignited .an additional channel for the formation of sdb stars in wide binaries , which has not received much attention in the past , involves binaries that experience stable rlof when passing through the hertzsprung gap ( so - called early case b mass transfer ) ( han , tout & eggleton 2000 ; han et al .2002 ; in preparation [ henceforth , paper ii ] ) .all of the sdb binaries produced through stable rlof channels are consistent with the observations by green , liebert & saffer who showed that some sdb stars appear to be members of long - period binaries .the main purpose of this study is to re - examine the various scenarios for the formation of sdb stars in some detail . in this first paper, we concentrate on the individual evolutionary channels . using detailed stellar and binary calculations ,we model the physics and appearance of sdb stars and then test individual evolutionary channels using binary population synthesis ( bps ) .we demonstrate that all of the main evolutionary channels proposed previously can lead to the formation of sdb stars . as a by - product we constrain the ce ejection efficiency from the observed period distribution of compact sdb binaries to arrive at a physically motivated and experimentally calibrated prescription for the ce phase .the outline of this paper is as follows . in section 2 ,we describe the stellar evolution code and the binary population synthesis code adopted in this study . in section 3 , we present the conditions for the formation of sdb stars from the ce ejection channel , their evolutionary tracks and simplified bps models to constrain the ce ejection efficiency . in section 4 , we derive the conditions for helium ignition in objects that result from the merger of two he white dwarfs and use monte carlo simulations to determine their mass distribution . in section 5, we investigate the criterion for stable rlof and the formation of sdb stars in wide binaries . in the follow - up paper ( paper ii )we will apply these results to a comprehensive binary population synthesis study and will estimate the relative importance of these individual channels .in this study , we employ two numerical computer codes : a stellar evolution code to determine the structure and follow the evolution of sdb stars , and a binary population synthesis code to examine different evolutionary channels .the stellar evolution code used is the one that has been originally developed by p. p. eggleton , which has been updated with the latest input physics over the last 3 decades as described by han , podsiadlowski & eggleton and pols et al . .the code distinguishes itself by the use of a self - adaptive non - lagrangian mesh , the treatment of both convective and semiconvective mixing as diffusion processes and the simultaneous and implicit solution of both the stellar structure equations and the chemical composition equations which includes convective mixing .these characteristics make the code very stable and easy to use .the current code uses an equation of state that includes pressure ionization and coulomb interaction , recent opacity tables derived from rogers & iglesias and alexander & ferguson , nuclear reaction rates from caughlan & fowler and caughlan et al . , and neutrino loss rates from itoh et al .. we set @xmath15 , the ratio of the mixing length to the local pressure scale height , to 2 .for population i ( pop i ) stars ( with a typical composition of hydrogen abundance @xmath16 , helium abundance @xmath17 and metallicity @xmath18 ) , such a value for @xmath19 gives a roughly correct lower main sequence , as determined observationally by andersen .it also well reproduces the location of the red giant branch in the hr diagram for stars in the hyades supercluster @xcite , as determined by bessell et al . .a fit to the sun also leads to @xmath20 as the most appropriate choice @xcite . for convective overshooting, the code does not employ a prescription in terms of the pressure scaleheight @xmath21 where the overshooting length is a fixed fraction of @xmath21 .instead the code uses an approach based on the stability criterion itself , the ` @xmath22 prescription ' , by incorporating a condition that mixing occurs in a region with @xmath23 , where @xmath24 is the ratio of radiation pressure to gas pressure and @xmath22 is a specified constant , the overshooting model parameter .critical tests of stellar evolution by means of double - lined eclipsing binaries @xcite show that @xmath25 gives the best fit to the observed systems , where @xmath25 corresponds to an overshooting length of @xmath26 pressure scale heights .roche lobe overflow is treated directly within the code .it has been tested thoroughly and works very reliably . because the mesh - spacing is computed along with the structure , the inclusion of rlof is almost trivial: it just requires a modification of one surface boundary condition .the boundary condition is written as @xmath27 \label{boundary}\ ] ] where @xmath28 gives the rate at which the mass of the star changes , @xmath29 is the radius of the star , and @xmath30 the radius of its roche lobe .@xmath31 is a constant . with @xmath32yr, rlof proceeds steadily , and the lobe - filling star overfills its roche lobe as necessary but never overfills its lobe by a substantial amount ( typically @xmath33 ) . the stellar evolution code described above evolves only a single star or both components of a binary at a time .however , stellar evolution theory should give and predict the _ statistical _ properties of a whole stellar population as well as the properties of individual stars or binaries . in order to investigate statistical properties of stars and check evolutionary mechanisms for different types of stars , han , podsiadlowski & eggleton developed a monte carlo simulation code , or ` binary population synthesis ' ( bps ) code , which is able to evolve a sample of 1 million or more stars ( including binaries ) simultaneously by interpolating the properties of individual stars as a function of evolutionary age in a specially prepared grid of stellar models .this code has been steadily updated ever since @xcite .the bps code needs a grid of stellar evolution models .taking pop i as the standard population model , we carried out a large number of stellar evolution calculation for a wide range of masses ( @xmath34 to @xmath35 ) , including the evolution of helium stars from @xmath36 to @xmath37 , at an interval of @xmath38 in @xmath39 . in the calculation of the stellar models ,we did not include stellar wind mass loss for most of the grids calculated .however , for massive stars the evolution was terminated at the humphreys - davidson ( hd ) limit @xcite ; we assumed that at the hd limit the envelope of the massive star was lost completely , and then treated the remnant core as a helium star .the evolution of low- and intermediate - mass stars was terminated at the point when the total energy of the envelope became positive , assuming that the envelope was ejected at this point , leaving a white dwarf remnant @xcite . for a given starwe obtain the required stellar parameters at a particular evolutionary age ( e.g. the luminosity , effective temperature , radius , core mass , core radius , envelope binding energy ) by interpolation from our set of stellar evolutionary tracks . in bps , one has to evolve binaries as well as single stars . at present stellar wind mass loss and binary interactionsare included in the form of simple prescriptions .we plan to replace these by ever more realistic modelling as part of our ongoing work .one important improvement we have implemented in the present work is that we adopted full binary evolution calculations for systems where rlof occurs in the hertzsprung gap @xcite .another uncertainty which we will address in more detail in the future is the criterion for dynamically unstable rlof . if the primary fills its roche lobe as a red giant , rlof may be dynamically unstable if the mass ratio at the onset is larger than some critical value @xmath40 , given e.g. by hjellming & webbink and webbink .however , this critical mass ratio only applies to conservative rlof and does not fully take into account the detailed dynamics at the onset of mass transfer . from observationsit is clear that rlof on the fgb / agb is non - conservative @xcite .part of the transferred mass is lost from the system . in our modelwe assume that the lost matter carries away the same specific angular momentum as pertains to the system . defining a mass transfer efficiency , @xmath41 , as the ratio of the mass accreted by the secondary to the mass transferred from the primary ( @xmath42 for conservative rlof ) , we find that the critical ratio @xmath40 depends strongly on the mass transfer efficiency @xcite .dynamically unstable mass transfer may result in the formation of a common envelope ( ce ) @xcite , leading either to the formation of a close binary or the complete merger of the two components ( see section 3 ) .given a binary sample , the bps code performs all the necessary interpolations in the model grid , integrates the mass loss along evolutionary tracks for an assumed stellar - wind law and deals with all the binary interactions .the output of the code are all the parameters for different types of binaries or single objects formed as a consequence of the evolution and various interactions . in the original version ,the bps code used the following stellar model grids as input : 1 .z=0.02 , no stellar wind , no overshooting , 0.08 - 126.0@xmath43 for normal stars( @xmath16 , @xmath17 ) , 0.32 - 8.0@xmath43 for helium stars .2 . z=0.004 ,no stellar wind , no overshooting , 0.1 - 126.0@xmath43 for normal stars( @xmath44 , @xmath45 ) , 0.32 - 8.0@xmath43 for helium stars .z=0.001 , no stellar wind , no overshooting , 0.1 - 126.0@xmath43 for normal stars( @xmath46 , @xmath47 ) , 0.32 - 8.0@xmath43 for helium stars . for the purpose of this paper, we have also calculated several additional grids that include convective overshooting and stellar wind mass loss : 1 .z=0.02 , no stellar wind but with overshooting , 0.63 - 3.2@xmath43 for normal stars .z=0.02 , 1/4 of reimers wind @xcite and with overshooting , 0.63 - 3.2@xmath43 for normal stars .3 . z=0.02 , 1/4 of reimers wind but without overshooting , 0.63 - 3.2@xmath43 for normal stars .4 . z=0.02 , 1/2 of reimers wind and with overshooting , 0.63 - 3.2@xmath43 for normal stars . 5 . z=0.02 , 1/2 of reimers wind but without overshooting , 0.63 - 3.2@xmath43 for normal stars .6 . z=0.004, 1/4 of reimers wind and with overshooting , 0.63 - 3.2@xmath43 for normal stars .in the common - envelope ( ce ) channel , the sdb star forms in a close binary as a consequence of dynamical mass transfer and a ce phase where the progenitor , a giant star , starts to fill its roche lobe when it is relatively close to the tip of the first red - giant branch ( fgb ) .this situation generally occurs when the radius of the mass - losing star increases faster than its roche - lobe radius .this leads to mass transfer on a dynamical timescale and the formation of a common envelope where the envelope of the giant engulfs both its degenerate core and the companion star .friction between these orbiting components and the envelope causes the orbit of the immersed binary to shrink .if the orbital energy released in the process is able to eject the envelope , this process leaves a very tight binary consisting of the degenerate core of the giant and the companion star .this is believed to be the main mechanism , originally proposed by paczyski ( 1976 ) , by which an initially wide binary is transformed into a very close system .if this happens when the giant was sufficiently close to the tip of the fgb at the beginning of mass transfer , i.e. the core was close to experiencing the helium flash , the remnant core of the giant may still ignite helium ( as first demonstrated by castellani & castellani 1993 ) and hence become a helium core - burning sdb star , where the companion can be either a white dwarf ( wd ) or in some cases a low - mass star .these are exactly the objects observed in large numbers by maxted et al .. note that the companion star can in principle also be a normal dwarf star with a mass as high as @xmath48 ( which depends on the condition for dynamical mass transfer ; see 5.1 ) .however , such a system would have a composite spectrum .since such systems were excluded from the pg catalog , they would not appear in radial - velocity studies based on this catalog .the details of ce evolution , in particular the conditions for which the envelope can be ejected are far from well understood at the present time ( see e.g. iben & livio 1993 ) .on the other hand , the identification of a large number of sdb stars in very close binaries , which must all have passed through such a well - defined evolutionary channel , provides a unique opportunity to test particular models of ce evolution and may even help to calibrate the criterion for ce ejection .indeed , it is one of the surprises of the observations by maxted et al . that the period distribution is extremely wide ranging from 2hr to more than @xmath49d .lllll @xmath50 & @xmath51 & @xmath52 & @xmath53 & @xmath54 + ( @xmath43 ) & ( @xmath43 ) & ( @xmath43 ) + & & & & + + 0.794 & 0.4529 & 0.4742 & 2.2245 & 2.2923 + 1.000 & 0.4502 & 0.4725 & 2.1838 & 2.2539 + 1.259 & 0.4492 & 0.4720 & 2.1492 & 2.2161 + 1.585 & 0.4461 & 0.4699 & 2.1069 & 2.1772 + 1.995 & 0.4174 & 0.4422 & 1.9537 & 2.0453 + 2.114 & 0.4010 & 0.4217 & 1.8628 & 1.9532 + & & & & + + 0.800 & 0.4538 & 0.4746 & 2.2482 & 2.3076 + 1.000 & 0.4509 & 0.4727 & 2.1985 & 2.2707 + 1.260 & 0.4496 & 0.4723 & 2.1578 & 2.2264 + 1.600 & 0.4351 & 0.4601 & 2.0656 & 2.1465 + 1.900 & 0.3889 & 0.4087 & 1.8113 & 1.9050 + + 2.050 & 0.3190 & 0.3224 & 1.4290 & 1.4355 + 2.265 & 0.3181 & 0.3460 & 1.2881 & 1.5317 + 2.512 & 0.3223 & 0.3715 & 0.6871 & 1.5612 + 3.162 & 0.4385 & 0.4669 & 0.7489 & 1.6891 + 3.981 & 0.5947 & 0.6146 & 0.8068 & 1.8685 + 5.012 & 0.8049 & 0.8286 & 0.8598 & 2.0618 + 6.310 & 1.1108 & 1.1284 & 0.9181 & 2.2587 + & & & & + + 0.800 & 0.4642 & 0.4855 & 2.0886 & 2.1475 + 1.000 & 0.4604 & 0.4826 & 2.0530 & 2.1184 + 1.260 & 0.4578 & 0.4810 & 2.0186 & 2.0856 + 1.600 & 0.4314 & 0.4564 & 1.8829 & 1.9710 + 1.750 & 0.3933 & 0.4149 & 1.6800 & 1.7825 + + 1.850 & 0.3238 & 0.3367 & 1.4003 & 1.4439 + 2.000 & 0.3175 & 0.3403 & 1.2903 & 1.4594 + 2.512 & 0.3381 & 0.3935 & 0.5740 & 1.5122 + 3.162 & 0.4492 & 0.4942 & 0.6343 & 1.6480 + 3.981 & 0.6006 & 0.6398 & 0.6910 & 1.8080 + 5.012 & 0.8143 & 0.8579 & 0.7515 &1.9889 + 6.310 & 1.1187 & 1.1539 & 0.8126 & 2.1707 + note - @xmath50 : initial zero - age main - sequence mass ; @xmath51 : minimum mass of the he core for helium ignition ; @xmath52 : mass of the he core at the tip of the first red - giant branch ; @xmath55 : stellar radius corresponding to @xmath51 ; @xmath56 : the radius at the tip of the first red - giant branch .the models were calculated by taking off mass from the envelope at a rate of @xmath57 the mass of the star in solar units until the envelope collapsed .note , however , that @xmath51 and @xmath55 for @xmath58 correspond to models at the end of the main sequence , while those with @xmath59 are models on the fgb .[ min - core - tab ] in order for the degenerate core to ignite helium in a helium flash after the envelope has been ejected , its giant progenitor had to be relatively close to the tip of the fgb at the beginning of mass transfer .dcruz et al .have shown by varying the wind mass - loss rate on the fgb that the core mass for helium ignition can be as low as @xmath60 .the situation considered here is slightly different since mass ejection in a ce phase is a sudden event , occurring on a timescale short compared to the evolutionary timescale of a giant .we therefore performed a comprehensive series of stellar calculations where we assumed that a giant loses its hydrogen - rich envelope near the tip of the fgb to determine the minimum core mass above which the core will still ignite helium after the ejection of the envelope .this critical mass generally depends on the zams mass of the giant , but also on the assumptions about mass loss , metallicity and the degree of convective overshooting from the core . to examine these various situations, we therefore considered three sets of calculations .set ( 1 ) assumes no stellar wind , no convective overshooting and a solar metallicity of @xmath18 . in set ( 2 )we again use a solar metallicity , but also include a stellar wind parametrized by a reimers wind mass - loss law @xmath61 where for the standard model we use an efficiency @xmath62 @xcite . in this set , we also take into account convective overshooting based on the calibration of this parameter by schrder et al .and pols et al . , which corresponds to overshooting of @xmath26 pressure scale heights from the core .set ( 3 ) is similar to set ( 2 ) except that we use a metallicity of @xmath63 , characteristic of a thick disc population @xcite .we find that the maximum initial zams mass below which stars experience a helium flash decreases from @xmath64 for set ( 1 ) to @xmath65 for set ( 2 ) and @xmath66 in set ( 3 ) .stars with zams masses larger than these values ignite helium non - degenerately . some of these can also become sdb stars . to determine the minimum core mass for the helium flash for each zams mass, we considered a series of models near the tip of the fgb and took the h - rich envelopes off at a high rate ( chosen to be @xmath67 the mass of the star in solar units ) until the envelopes collapsed ( note that we switched hydrogen burning off when the envelope mass became less than @xmath68 to prevent the occurrence of hydrogen shell flashes ) .we then followed the subsequent evolution of the core either until it had cooled to a surface temperature of less than @xmath69k or until it ignited helium in the core . in table[ min - core - tab ] and figure [ min - core ] we present the minimum core mass as a function of zams mass above which helium is ignited and also the core mass at the tip of the fgb .these results show that the minimum core mass has to be typically within 5 per cent of the core mass at the tip of the fgb where the minimum decreases from 0.45/0.46@xmath43 for the less massive progenitors to 0.39/0.40@xmath43 for the most massive stars that still experience a helium flash .while the range in mass is relatively small , it corresponds to a fairly large range ( @xmath70 per cent ) in radius ( also shown in table [ min - core - tab ] ) , since giants expand quite significantly close to the tip of the fgb .note that the more massive stars do not experience a helium flash .if the envelopes are stripped off near the tip of the fgb for stars with @xmath71 ( set 2 , pop i ) or with @xmath72 ( set 3 , @xmath73 ) , the cores ignite helium under non - degenerate conditions ( as shown in table 1 and figure 1 ) . for even more massive stars, the cores will burn helium even when the envelopes are lost in the hertzsprung gap .the envelopes of these massive stars are generally much more tightly bound than those near the tip of the fgb for stars with degenerate cores and hence are much harder to eject .generally , the more massive the star , the more tightly bound is its envelope .therefore we find in the bps calculations that ce ejection is only possible for stars that are not too massive .these , however , leave fairly low - mass sdb stars ( @xmath74 ) with very short orbital periods .since helium ignition in a degenerate core causes a mild thermonuclear runaway which leads to the expansion of the core on a dynamical timescale , we are unable to follow it with our hydrostatic stellar - evolution code . however , once the core has expanded sufficiently and the core has become non - degenerate , helium burning quickly stabilizes and the core regains hydrostatic equilibrium . in figures [ sdb - evol ] and [ sdb - hrd ] we show the subsequent evolution of various sdb stars in a @xmath7 @xmath3 diagram and a standard hertzsprung - russell ( h - r ) diagram . the @xmath7@xmath3 is particularly useful , since sdb stars can be placed on it based on their spectra alone with the help of model stellar atmospheres independent of their distance and their luminosity .the initial models ( i.e. on the zero - age horizontal branch ) in panels ( a ) , ( b ) and ( c ) of figure [ sdb - evol ] were constructed in such a way that they had the same core mass and chemical composition as the appropriate models from set ( 2 ) in 3.1 at the point of helium ignition ( i.e. at the helium flash ) .panel ( a ) shows the evolution of various selected sdb models with different total masses , originating from a range of progenitor masses and with a variety of different envelope masses .the symbols give the position of observed sdb stars ( from maxted et al.2001 ; note , however , the uncertainties in @xmath3 are about 0.15 ; see saffer et al .1994 ) , where the different symbols correspond either to different ranges of orbital periods ( if known ; filled symbols ) or indicate the magnitude of radial velocity variations for systems without known orbital periods ( open symbols ) .as the figure shows , there is generally excellent overlap between the evolutionary tracks and the observed systems . to show this more systematically , panel ( b ) of figure [ sdb - evol ] presents the tracks for sdb stars of a typical mass of @xmath75 , but with different masses of the hydrogen - rich envelope , ranging from 0 to @xmath76 ( here the envelope is defined as the outer part of the sdb star with a hydrogen mass fraction larger than 0.01 ) .the two dashed curves give the location of the zahb ( right curve ) and the location where the sdb stars exhaust helium in their centres ( left curve ) .tick marks are also shown along the tracks to indicate the speed of evolution across the diagram .as one would expect , sdb stars without hydrogen envelopes are the hottest and the most compact , and the evolutionary tracks are shifted towards lower temperatures and lower gravity as the mass in the hydrogen - rich envelope is increased .the majority of observed systems lie between the dashed curves , marking the range of helium core - burning objects .there appears to be , however , a fairly large range of envelope masses from 0 to @xmath77 .there are also a few objects that lie significantly outside the helium core - burning band .while evolutionary tracks will pass through most of these observational points after the end of helium core burning , this corresponds to a fast evolutionary phase , as can be seen from the wide separation of tick marks .a more likely explanation is that some of these are more massive objects . in panel ( c ) of figure [ sdb - evol ]we show the dependence of evolutionary tracks of sdb stars on convective overshooting .the sdb stars without convective overshooting display ` breathing pulses ' ( see e.g. castellani et al .1985 ) . in panel ( d ) of figure [ sdb - evol ]we show the dependence of the location of the helium core - burning sdb stars on the mass of the sdb star for 3 envelope masses ( 0.001 , 0.002 , @xmath78 ) .the masses of the sdb stars range from 0.35 to @xmath79 ( note that sdb stars more massive than @xmath80 can form in some of the other evolutionary channels discussed later in this paper ) helium core - burning star that was artificially transformed into a sdb star of the chosen total mass and envelope mass . ] .four of the six objects that were outside the helium core - burning band in panel ( b ) fall onto a helium core burning track with a more massive sdb star .this may provide an indication that some of the sdb stars have masses as high as @xmath81 .the two remaining objects ( pg 1553 + 273 and pg 1051 + 501 ) , both of which show low radial velocity variations , lie well above all of these tracks .this suggests that they have either substantially more massive envelopes ( see section 5 ) or have already exhausted helium in their cores and are now evolving quickly away from the horizontal branch .ce evolution is one of the most important but also one of the least understood phases of binary evolution ( see iben & livio 1993 ; taam & sandquist 2000 ; podsiadlowski 2001 for reviews with different emphasis ) .one of the uncertainties is related to the conditions under which a binary experiences dynamical mass transfer and a ce phase ( this will be further discussed in 5.1 ) . a second area of uncertaintyis related to the criterion for the ejection of the ce , which crucially determines the orbital period distribution of post - ce binaries .the latter depends on what fraction of the orbital energy that is released in the spiral - in process can be used to drive the ejection , which depends on the efficiency with which energy can be transported to the stellar surface where it can be radiated away .it also depends on the efficiency of the dynamics of the ejection process ; e.g. if the envelope is ejected with a velocity much larger than the surface escape velocity , the efficiency per unit mass will be reduced ( see e.g. the discussion in taam & sandquist 2000 ) .a further factor is related to the question what fraction of the thermal energy , in particular the ionization energy , can be converted into kinetic energy and can help to drive the expansion of the envelope .this is particularly important for giants that fill their roche lobes near the tip of the fgb or the agb since in such extended stars the total binding energy of the envelope , including the ionization energy , is greatly reduced and ultimately becomes 0 ( see han , podsiadlowski & eggleton 1994 ; hpe ) .this has the consequence that very little spiral - in is required to release , in principle , enough energy to eject these loosely bound envelopes ; this leads to post - ce binaries with relatively long orbital periods . in binary population synthesis ( bps )studies it is commonly assumed that the common envelope is ejected when the change in orbital energy times some efficiency factor , @xmath82 , exceeds the binding energy of the envelope , where the latter is often approximated by a simple analytical expression . our approach is rather different from this and it is designed to provide a more physical parametrization of the ce - ejection process .our common - envelope ejection criterion can be written as @xmath83 the left - hand side represents the fraction of the change in the orbital energy that can be used for the ejection , as in most other commonly used prescriptions . however , on the right - hand side , we include both the gravitational energy of the envelope ( @xmath84 ) and a fraction @xmath85 of its thermal energy ( @xmath86 ) , which in particular includes the ionization energy .moreover , instead of using analytical approximations for these energies , we use the values obtained from full stellar structure calculations ( see hpe for details and also dewi & tauris 2000 ) . the fact that we have 2 parameters , @xmath82 and @xmath85 , instead of one allows us to specifically assess the importance of the thermal energy contribution .indeed , the observations of post - ce sdb binaries from this well - defined evolutionary channel may allow the calibration of this parameter at least in principle . in order to test the ce ejection criterion, we need to perform a binary population study where we simulate the period distribution of sdb binaries after the ejection of the common envelope . to avoid unnecessary complications, we use a simplified bps model in this section , simplified in the sense that we do not model the evolution of the system before the ce phase that leads to the formation of the close sdb binary .we also restrict ourselves to systems where the companion star is likely to be a white dwarf ( wd ) .the model is completely specified by three distributions : the distribution of the white - dwarf mass , the mass of the giant and the orbital separation before the ce phase .we use a uniform distribution in @xmath87 from 1 to 4 where @xmath88 is the orbital separation , a simple wd mass distribution ( @xmath89 for @xmath90 or @xmath91 ) and a miller & scalo mass distribution for the giant between 0.8 and @xmath92 . the wd mass distribution may seem a little bit odd at first sight , but it actually mimics the bimodal mass distribution of wds after the first rlof phase as found in previous bps studies @xcite .we emphasize that because of the wide distribution in @xmath93 it is almost certain that the resulting orbital period distribution of post - ce sdb binaries will be wider than what would be obtained in a more realistic simulation , since in a full bps simulation only a subset of this parameter space would be realized by actual systems ( see paper ii ) .nevertheless , as we shall show below , this method still provides a good diagnostic for the ce efficiency . in our first series of bps simulations ,we assume a population i metallicity of @xmath18 and include convective overshooting ( of 0.25 pressure scale heights ) .we consider 3 sets of stellar models : ( 1 ) models with no stellar wind ; ( 2 ) models with a reimers wind mass loss with @xmath94 ; and ( 3 ) models with a reimers wind with @xmath95 .for each of these 3 sets we take six combinations of ce parameters @xmath82 and @xmath85 , including both efficient and inefficient models .1 . @xmath96 , @xmath97 2 .@xmath98 , @xmath97 3 .@xmath98 , @xmath99 4 .@xmath100 , @xmath101 5 .@xmath102 , @xmath97 6 .@xmath102 , @xmath103 the resulting post - ce orbital period distributions for these 3 sets are shown in panels ( a ) , ( b ) and ( c ) of figure [ sbps - p ] .first , note by comparing the three panels that the different assumptions about the stellar wind make relatively little difference . on the other hand, the variation of the ce parameters affects the orbital period distributions quite dramatically .the overall behaviour of the various curves is relatively easy to understand . in the case of inefficient ce ejection , i.e. low @xmath82 ,the binary has to spiral in much further during the ce phase before enough energy has been deposited in the envelope to make ejection possible , resulting in very short orbital periods . as @xmath82 is increased , the orbital period distribution shifts towards longer orbital periods .increasing the value of @xmath85 also increases the post - ce orbital periods , since it reduces the binding energy of the envelope .it also widens the period distribution very substantially .this reflects the relative importance of the ionization energy which varies significantly between a star of @xmath104 and @xmath105 . for a @xmath104 star ,the total binding energy of the envelope near the tip of the fgb is close to 0 ( in fact , it may be positive , see hpe ) , implying that the envelope is only very loosely bound and that very little spiral - in is required to eject it .on the other hand , the envelope of a @xmath105 star is much more tightly bound , leading to much smaller post - ce orbital periods of the systems .some of the curves ( e.g. the dotted ones ) have three peaks .the left one corresponds to giants with @xmath106 which have much more tightly bound envelopes than fgb stars with @xmath107 .since this requires much deeper spiral - in before the envelope can be ejected , it leads to the shortest orbital periods .the middle and the right peaks are the result of ce ejections for fgb stars with @xmath107 , where the middle peak corresponds to white dwarf primaries with @xmath108 and the right peaks to white dwarfs with @xmath109 . in cases where the ce ejection is not very efficient ,i.e. either @xmath82 or @xmath85 is small , the differences in the period distributions for the two sets of white dwarfs tends to be small and the latter two peaks tend to merge into one .the short ticks in these panels along the orbital period axis show the periods of sdb binaries with known orbital periods @xcite .as is clear by inspection , the simulated period distributions taken together cover the whole range of observed periods , although no single model on its own provides a perfect fit to the observed distribution .the best models require high values for both @xmath82 and @xmath85 in order to cover the whole range of orbital periods . inspection of panels ( a ) , ( b ) and ( c ) of figure [ sbps - p ] suggests that the dotted simulations with @xmath110 provide the best overall representation of the observed period distribution .such a combination of parameters is physically quite reasonable , since it suggests that both the ejection of the common envelope and the conversion of thermal ( and ionization ) energy are efficient processes , but are not perfect .this is consistent with theoretical ce ejection simulations @xcite and independent constraints from bps studies of other post - ce binaries with relatively long orbital periods ( e.g. certain symbiotic stars and barium stars ; see han et al . 1995 ) , which require fairly efficient ce ejection . in figure [ sbps - tg ], we present various simulated distributions of the best overall model , i.e. the model with @xmath111 and @xmath112 .panels ( a ) and ( b ) of figure [ sbps - tg ] are scatter diagrams in the @xmath7 @xmath3 and the h - r diagram , respectively .we assumed that the envelope masses are uniformly distributed between 0.0 and @xmath113 .this leads to the concentration of sdb stars in the upper - right part in panel ( a ) or the right part in panel ( b ) ( see panel ( b ) of figure [ sdb - evol ] ) .the bulk of the observed distribution of sdb stars in figure [ sbps - tg](a ) overlaps nicely with the simulated distribution .the systems that fall outside the theoretical region are the same as those already discussed in 3.2 .these may be more massive sdb stars or those with larger hydrogen - rich envelopes and may originate from some of the other sdb channels discussed in subsequent sections .there are no observational data points @xcite in the lower - right part of panel ( a ) due to observational selection effects ( this point will be addressed in detail in paper ii ) . figure [ sbps - mf]a shows the simulated distribution of the mass function for sdb stars and figure [ sbps - mf]b compares the mass function distribution for different masses of the sdb star . while this comparison should not be taken too literally ( since it relies on a very simplified and incomplete bps model ), it confirms that a model with a sdb mass of @xmath14 provides an overall good fit to the observed distribution , consistent with earlier findings ( e.g. heber 1986 ) .figure [ sbps - m ] gives the simulated distributions of masses of sdb stars .the distributions have three peaks a sharp major peak at @xmath75 , a secondary peak at @xmath114 , and a minor peak at @xmath115 .the major peak is caused by systems with a low - mass zams secondary where the ce is ejected near the tip of fgb ( see figure [ min - core ] ) .the secondary peak is due to the fact that the range in stellar radius which leads to ce ejection near the tip of the fgb and produces sdb stars is wider for @xmath116 than for @xmath117 ( see table [ min - core - tab ] ) .furthermore ce ejection with @xmath118 results in a low - mass sdb star which has a long core helium - burning lifetime .the minor peak contains systems which had a zams mass greater than the helium flash mass .the figure shows that almost all the simulated masses of sdb stars are less than @xmath119 , which means that the sparsely scattered dots at @xmath120 in figure [ sbps - tg ] correspond to the post - central - helium - burning phase of sdb stars . in order to test how the results depend on metallicity, we also performed a series of simulations representing a thick disc population with a metallicity @xmath63 @xcite .these results are shown in panel ( d ) of figure [ sbps - p ] ( for the same combinations of ce parameters as in 3.4.1 ) and are quite similar to the previous case .however , the middle and the right peaks discussed in 3.4.1 merge into a single peak .the reason for this convergence is that lower - metallicity giants are hotter and more compact , and therefore have more tightly bound envelopes .different wds then lead to smaller differences between orbital periods for the post - ce systems , causing a merging of the two peaks . in this sectionwe have shown that helium - burning sdb stars in compact binaries can be formed if the progenitor giants filled their roche lobe when they were close to the tip of the fgb .figure [ p - range ] shows the range of orbital periods at the beginning of mass transfer as a function of zams mass for which the core ( degenerate if @xmath107 for pop i or @xmath122 for @xmath63 ) will subsequently ignite helium .since this is a well - defined evolutionary channel , it provides an excellent diagnostic to test the criterion for the ejection of the ce . in figure [ p - tg ] we plot systems with determined orbital periods in the @xmath7 @xmath3 diagram where the size of the symbols indicates the orbital period. there may be a weak hint that the systems with longer orbital periods are less compact and cooler , which implies that they have bigger envelope masses .this would be consistent with simple expectations for ce ejection , since larger remnant envelopes may remain bound to the system if the envelope is ejected at a wider separation .on the other hand , a small fraction of sdb stars may result from ce ejections with @xmath123 for pop i or @xmath124 for @xmath63 . since the envelope of these more massive giantsare more tightly bound and the orbital energy available in the spiral - in phase is smaller ( due to the smaller orbital period of the pre - ce system ) , sdb stars produced from more massive giants tend to have very small orbital periods .most of these sdb stars have fairly low masses ( @xmath74 ) and are probably selected against in the observational sample ( see paper ii ) .our analysis , however , raises a few further issues .the first is the question whether the orbital parameters that lead to the formation of the compact sdb binary , in particular the orbital period range in figure [ p - range ] , are being realized by actual binaries ( note that we do not plot the figure for stars more massive than @xmath125 , as they hardly contribute to the formation of sdb stars due to their tightly bound envelopes ) .this depends on the previous binary evolutionary phases , which have not been modelled in this paper , but will be addressed in paper ii .a second problem is that we found that there is no single combination of the ce parameters @xmath82 and @xmath85 that produces a perfect fit to the observed period distribution for both metallicities considered .it is quite possible that the observed wide distribution is at least in part a consequence of a large variation of metallicities and possibly even a variation in the ce ejection parameters . considering how uncertain the theoretical modelling of the ce phase is at present, this would not be very surprising .finally we would like to point out an alternative possibility : in recent studies of the dynamics of the ce ejection process , ph.p .( see e.g. the discussion in podsiadlowski 2001 ) found that even in cases where the envelope is not ejected in the initial dynamical phase , it will always be ejected at a later stage if the spiralling - in star is relatively compact and penetrates into the initially radiative layer of the giant .an example for such a delayed dynamical ejection is shown in figure [ vrp ] , which represents a 1-dimensional , hydrodynamical simulation of the spiral - in of a @xmath126 compact star spiralling into the envelope of a 1.6@xmath43 giant .the initial spiral - in is very rapid ( see the thick solid curve ) , but slows down once the envelope has expanded significantly and the friction between the spiralling - in binary and the envelope has decreased ; in the subsequent slow spiral - in , the envelope is close to hydrostatic equilibrium .however , once the compact object reaches the initially radiative region , a dynamical instability ensues that it is likely to lead to the ejection of the envelope .while this result should only be considered tentative at the moment , we note that the conditions for such a delayed dynamical instability depend mainly on the structure of the giant and can not be described , not even in principle , by a simple @xmath19 ejection criterion .the characteristic period obtained in this case is of order @xmath127 , similar to the shortest periods of observed sdb binaries .if this is the process that leads to the formation of sdb binaries with the shortest orbital periods , one may expect to see some structure and possibly even some evidence for a bimodal orbital period distribution , for which there is no evidence at the moment .a second channel that can lead to the formation of a single sdb star involves the merger of two he white dwarfs @xcite . closehe wd binaries are formed as a result of one or two ce phases @xcite .if the orbital period is sufficiently short ( typically less than 6.76 hr for a @xmath126+@xmath126 pair to merge in 15gyrs ) , gravitational radiation will cause such a system to shrink until the lighter white dwarf fills its roche lobe at a typical period of @xmath128min .mass transfer will be dynamically unstable when the lighter white dwarf is larger than @xmath129 2/3 the mass of the more massive component @xcite .this leads to its dynamical disruption and the formation of an accretion disc surrounding the more massive white dwarf @xcite .while the subsequent evolution has not been modelled in any detail , it is probably reasonable to expect that a large part of this accretion disc will be accreted by the more massive component .this accretion will initially occur on a dynamical timescale , but , as the mass in the disc decreases and the disc expands , the accretion rate will decrease and be ultimately determined by the internal viscous processes that govern the evolution of the disc . as the mass of the white dwarf increases , there will be a point at which helium is ignited in a shell and the star will subsequently become a helium core - burning sdb star due to the inward propagation of these nuclear burning shells @xcite . unlike the other scenarios considered in this paper, this sdb star will be a single object ..the minimum merger masses for helium ignition [ cols= " < , < , < , < , < " , ] note - @xmath130 : zams mass of the primary ; @xmath131 : the surface mass of the primary at the onset of rlof ; @xmath132 : the mass of the companion ( wd / ns ) adopted ( close to the minimum mass for stable rlof according to the results in table [ stable - rlof ] ) ; @xmath133 : the core mass of the primary at the onset of rlof ( rlof before the core mass reaches @xmath133 will not result in a helium - burning sdb ) ; @xmath134 : the orbital period ( in days ) at the onset of rlof ; @xmath135 : the mass ratio at the onset of rlof ; @xmath136 : the orbital period of the sdb star formed ; @xmath137 : the mass of the sdb star formed ; @xmath138 : the mass ratio of the sdb binary .[ stable - rlof - sdb ] to determine the conditions for the formation of sdb stars through the stable rlof channel , we performed a series of binary stellar evolution calculations for mass donors with different masses on the zero - age main sequence ( zams ) , ranging from 0.8 to @xmath105 .for each mass , we systematically varied the mass of the core at the beginning of mass transfer and the mass of the companion star . in these calculations we assumed that mass transfer was completely non - conservative , and that all the mass that was lost from the system carried with it the orbital angular momentum of the accreting component ( as appears to be most appropriate if the accretor is a white dwarf ) . in the standard set of calculations , we included a reimers - type wind with @xmath139 ( see equ .[ reimers ] ) before the mass - transfer phase .we switched this wind off once the mass - transfer rate exceeded the value given by equ .[ reimers ] by a factor of 100 and did not include stellar - wind mass loss after the end of the mass - transfer phase . in each calculation, we checked first whether mass transfer was dynamically stable . in cases , where mass transfer is dynamically unstable, there is no solution for the mass - transfer rate , @xmath140 , for which the radius of the secondary can be equal to the roche - lobe radius ( see han , tout & eggleton 2000 for the treatment of the surface boundary condition ) . if mass transfer is stable , we continued mass transfer until the mass donor started to shrink below its roche lobe , terminating the mass - transfer phase .if , at this point , the mass of the h - exhausted core exceeded the appropriate minimum core mass for subsequent helium ignition ( as determined in 3.1 ) , it may appear as a helium core - burning sdb star .the results of these calculations are summarized in tables [ stable - rlof ] and [ stable - rlof - sdb ] , which show for each zams mass the core mass and the total mass of the mass donor at the beginning of mass transfer and the minimum mass of the secondary ( and critical mass ratio ) for which mass transfer is stable and leads to the formation of a helium - burning sdb star .these results demonstrate , as discussed above , that mass transfer is dynamically stable even if the mass donor is substantially more massive than the secondary and that sdb stars can form through this channel without any non - standard assumptions . for stars with zams masses@xmath141 , the secondary mass has to be larger than @xmath142 , which is similar to the maximum mass of a white dwarf , and hence does not correspond to realistic systems with white - dwarf accretors .note that , for each mass , the critical mass ratio tends to decrease for the initially more evolved systems , since for these the evolutionary timescale is shorter and hence the mass - transfer rate higher than for the less - evolved ones .this has the consequence that the core mass grows less during the mass - transfer phase .one may also notice that some of the behaviour in table [ stable - rlof ] is non - monotonic .the non - monotonic behaviour of @xmath143 at the lowest @xmath133 is caused by primaries with a core mass @xmath133 near the base of the fgb , where the core is not very degenerate and the envelope is not yet fully convective .the non - monotonic behaviour for @xmath144 stars with the largest @xmath133 is a consequence of its thin envelope mass .while these calculations show that sdb stars can form in wide binaries without any non - standard assumptions , it is quite plausible , perhaps even likely , that the wind mass - loss rate before the beginning of the rlof phase will be enhanced due to the tidal interaction with the companion ( as originally proposed by eggleton & tout 1989 ) .the main effect this has is to increase the overall parameter space for which this channel produces sdb stars in wide binaries .this is illustrated in figure [ min - wd ] , which shows the minimum mass of the accreting white dwarf as a function of the zams mass of the donor for the standard model used above and two sequences of calculations , where we assumed that the pre - rlof wind mass loss rate was enhanced by a factor of 4 and 8 , respectively ( i.e. we took @xmath145 and 2 ) .figure [ binary ] shows a representative binary calculation from one of the sequences in the previous section for a star with a zams mass of @xmath104 . at the beginning of mass transfer , the mass donor has a core mass of @xmath146 and a total mass of @xmath147 , and the mass of the companion star is @xmath148 . with these parameters , masstransfer starts at an orbital period of 348.4d .initially , mass transfer occurs on a thermal timescale and reaches a maximum of @xmath149 .after the mass ratio has been reversed and the star has regained thermal equilibrium , mass transfer settles to a rate of @xmath150 and gradually decreases as the secondary ascends the giant branch . once the mass in the h - rich envelope drops below @xmath151 , the secondary shrinks below the roche lobe and mass transfer stops .as the remnant envelope collapses , the secondary quickly moves across the h - r diagram and ultimately becomes a sdb star of @xmath152 in a wide binary with an orbital period of 948.9d . whether the secondary becomes a sdb star and its location both in the @xmath7 @xmath3 and the h - r diagram also depends on mass loss via a stellar wind after the rlof phase .for example , consider a binary with an orbital period of 606d consisting of a giant with a total mass of @xmath153 and a core mass of @xmath154 ( which corresponds to one of the sequences with a @xmath104 zams star ) in orbit with a @xmath155 white dwarf .if wind mass loss is switched off once the system has entered the rlof phase , the giant starts to ignite helium when its core mass reaches @xmath156 , but its total mass is still @xmath157 ( the system has an orbital period of 1306d at this point ) .if there is no further mass loss , the secondary will then settle on the normal horizontal branch , burn helium in the core and then ascend the asymptotic - giant branch ( agb ) ( this typically requires an envelope mass larger than @xmath158 ; see dorman , rood , oconnell 1993 ) . on the other hand ,if wind mass loss is not switched off during the rlof phase but continues at a rate of 1/4 of the reimers rate , the total mass is @xmath159 when helium is ignited in the core of mass @xmath156 . in this case it has the appearance of a sdb star lying in the upper - right corner of the @xmath7 @xmath3 diagram .as we have shown in this section , stable rlof provides a third channel for the formation of sdb stars .these will generally be in wide binaries with typical orbital periods of @xmath160d .the mass distribution is similar to the distribution in the ce channel .however , one might expect that the envelope masses could be systematically larger than in the ce ejection case , since the orbital period is much longer , which means that a larger envelope mass may remain bound to the degenerate core when the donor becomes detached .this would suggest that sdb stars formed through this channel would be less compact and cooler than their counterparts from the ce channel .the importance of this channel is difficult to assess .while standard assumptions lead to the formation of sdb stars , the allowed range for the binary parameters is probably quite small in this model , since it requires relatively massive white dwarfs , which are not very common .however , this parameter range could be dramatically increased if the stellar - wind mass - loss rate is significantly enhanced , e.g. due to the tidal interaction with the companion star @xcite . on the other hand ,observational surveys could in principle allow the calibration of such enhanced stellar winds .one may notice that we only deal with wd companions in this section , though stable rlof with ms companions can also produce sdb stars .observational surveys show that at least half of all sdb stars have cool ms companions ( see section 1 ) .this seems to be in contradiction with the observations by maxted et al . that the majority of sdb stars are short - period binaries with wd companions , as it implies a binary fraction larger than 1 .the contradiction is due to various observational selection effects and will be addressed in paper ii . in this section , we made the simplifying assumption that mass transfer to a wd companion is completely non - conservative in the stable rlof channel .this assumption is appropriate for low mass - transfer rates , where nova explosions are believed to be effective in expelling all of the transferred matter , but probably not for higher rates ( larger than @xmath161 ) , where the white dwarf may be able to accrete most of the transferred matter and burn it steadily ( as in supersoft x - ray binaries ) . on the other hand , for even higher rates the white dwarf will start to swell up @xcite and may then lose most of the transferred mass again , possibly in the form of an optically thick wind @xcite .clearly all of these effects need to be studied further and should be included in future studies .in this paper , we have demonstrated that the three binary evolution channels that have been proposed for the formation of sdb ( and related sdo / sdob ) stars may all contribute to the observed population . in the ce ejection channel , which may account for more than 2/3 of all sdb stars , dynamically unstable mass transfer near the tip of the fgb results in the formation of a ce and spiral - in phase , leaving a short - period binary after the envelope has been ejected .the system becomes a sdb binary if helium is ignited .using detailed stellar evolution calculations , we have determined how close to the tip of the fgb the progenitor has to be at the onset of rlof . using simplified binary population synthesis calculations ,we have been able to show that the ce ejection process has to be very efficient and that the ionization energy in the envelope has to be included in the ejection criterion in order to be able to explain the observed orbital period distribution . in the stable rlof channel ,the progenitor systems experiences stable mass transfer in which the giant is stripped off its envelope as a result of the mass transfer .if this occurs near the tip of the fgb , the remnant helium core will still ignite helium in the core and become a sdb star in a binary with a long orbital period and a fairly thick hydrogen - rich envelope as compared to the other channels . using detailed binary evolution calculations, we demonstrated that this channel usually requires a fairly massive white dwarf companion or enhanced stellar wind mass loss before the onset of rlof ( e.g. tidally enhanced winds ) .double he wds may coalesce due to gravitational wave radiation .when helium is ignited in the merger , a single sdb star is formed , and its hydrogen envelope is likely to be very thin . we have determined the conditions for which the merged system will be able to ignite helium . in a follow - up paperwe will implement these results in full binary population synthesis calculations to assess their relative importance and to allow direct comparison with observed subdwarf populations .we thank drs . a. lynas - gray and s. yi for many stimulating discussions .we are grateful to dr . o. pols , the referee , for his valuable suggestions which helped to improve the paper .zh thanks the department of astrophysics , oxford for its hospitality .this work was in part supported by a royal society uk - china joint project grant ( ph.p and z.h . ) , the chinese national science foundation under grant no . 19925312 , 10073009 and nkbrsf no .19990754 ( z.h . ) .alexander d.r ., ferguson j.w . , 1994a , in jorgensen u.g . ,, molecules in the stellar environment .springer - verlag , berlin p. 149alexander d.r ., ferguson j.w . , 1994b , apj , 437 , 879 allard f. , wesemael f. , fontaine g. , bergeron p. , lamontagne r. , 1994 , aj , 107 , 1565 andersen j. , 1991 , a&ar , 3 , 91 aznar cuadrado r. , jeffery c.s . , 2001 , a&a , 368 , 994 benz w. , cameron a.g.w . ,press w.h . , bowers r.l . , 1990 , apj , 348 , 647 bessell m.s ., brett j.m . , scholz m. , wood p.r . , 1989 , a&as , 77 , 1 brassard p. , fontaine g. , billres m. , charpinet s. , liebert j. , saffer r.a . , 2001 ,apj , 563 , 1013 brown t.m ., ferguson h.c . , davidsen a.f . , dorman b. , 1997 , apj , 482 , 685 cameron a.g.w . , iben i.jr . , 1986 , apj , 305 , 228 carraro g. , girardi l. , bressan a. , chiosi c. , 1996 , a&a , 305 , 849 castellani m. , castellani v. , 1993 , apj , 407 , 649 castellani v. , chieffi a. , pulone l. , tornambe a. , 1985 , apj , 296 , 204 caughlan g.r ., fowler w.a . , 1988 , at .data nucl .data tables , 40 , 284 caughlan g.r . , fowler w.a . ,harris m.j . ,zimmerman b.a . , 1985 , at .data nucl .data tables , 35 , 198 dcruz n.l . ,dorman b. , rood r.t . , oconnell r.w . , 1996 , apj , 466 , 359 dewi j. , tauris t. , 2000 , a&a , 360 , 1043 dorman b. , rood r.t . , oconnell r.w . , 1993 , apj , 419 , 596 downes r.a . , 1986 ,apjs , 61 , 569 eggen o.j . , 1985 ,aj , 90 , 333 eggleton p.p . , 1971 , mnras , 151 , 351 eggleton p.p . , 1972 , mnras , 156 , 361 eggleton p.p . , 1973 , mnras , 163 , 179 eggleton p.p ., tout c.a . , 1989 , in batten a.h . ,ed . , algols , kluwer , dordrecht , p. 164ferguson d.h . , green r.f . , liebert j. , 1984 , apj , 287 , 320 fitzpatrick e.l ., garmany c.d . , 1990 ,apj , 363 , 119 giannuzzi m.a . , 1981 , a&a , 103 , 111 gilmore g. , wyse r.f.g . , kuijken k. , 1989 , araa , 27 , 555 goldberg d. , mazeh t. , 1994 , a&a , 282 , 801 green e.m . , liebert j. , saffer r.a . , 2000 , astro - ph/0012246 green r.f . , schmidt m. , liebert j. , 1986 , apjs , 61 , 305 hachisu i. , kato m. , nomoto k. , 1996 , apj , 470 , l97 han z. , 1995 , ph.d . thesis ( cambridge ) han z. , 1998 , mnras , 296 , 1019 han z. , eggleton p.p . , podsiadlowski ph ., tout c.a ., 1995 , mnras , 277 , 1443 han z. , eggleton p.p . , podsiadlowski ph ., tout c.a . , webbink r.f ., 2001 , in podsiadlowski ph ., rappaport s. , king a.r ., dantona f. , burderi l. , eds , evolution of binary and multiple star systems , asp conf .ser . , vol .229 , p. 205han z. , podsiadlowski ph ., eggleton p.p . , 1994 , mnras , 270 , 121 ( hpe ) han z. , podsiadlowski ph . , eggleton p.p . , 1995 , mnras , 272 , 800 han z. , podsiadlowski ph . , maxted p.f.l . ,marsh t.r ., ivanova n. , 2002 , in preparation han z. , tout c.a ., eggleton p.p . , 2000 ,mnras , 319 , 215 han z. , webbink r.f . , 1999 , a&a , 349 , l17 heber u. , 1986 , a&a , 155 , 33 heber u. , moehler s. , napiwotzki r. , thejll p. , green e.m . , 2002 ,a&a , submitted ( astro - ph/0201096 ) hjellming m.s . , webbink r.f . , 1987 , apj , 318 , 794 humphreys r.m . , davidson k. , 1979 , apj , 232 , 409 iben i. jr . , 1973 , in schramm d.n ., arnett w.d ., explosive nucleosynthesis , univ . texas , austin , p. 115iben i. jr . , livio m. , 1993 , pasp , 105 , 1373 iben i. jr . , renzini a. , 1983 , ara&a , 21 , 271 iben i. jr . , tutukov a.v . , 1985 , apjs , 58 , 661 iben i. jr . , tutukov a.v . , 1986 , apj , 311 , 753 iben i. jr ., tutukov a.v . ,yungelson l.r . , 1997 , apj , 475 , 291 itoh n. , adachi t. , nakagawa m. , kohyama y. , munakata h. , 1989 , apj , 339 , 354 ( erratum 1990 , apj , 360 , 741 ) itoh n. , mutoh h. , hikita a. , kohyama y. , 1992 , apj , 395 , 622 , ( erratum 1993 , apj , 404 , 418 ) jeffery c.s . , pollacco d.l . , 1998 , mnras , 298 , 179 kilkenny d. , koen c. , jeffery j. , hill c.s ., odonoghue d. , 1999 , mnras , 310 , 1119 kippenhahn r. , weigert , a. , 1990 , stellar structure and evolution , springer , berlin koen c. , orosz j.a ., wade r.a . , 1998 , mnras , 300 , 695 lamers h.j.g.l.m . ,fitzpatrick e.l . , 1988 ,apj , 324 , 279 livio m. , pringle j.e . , saffer r.a . , 1992 , mnras , 257 , l15 maxted p.f.l . , moran c.k.j . , marsh t.r . , gatti a.a . , 2000 ,mnras , 311 , 877 maxted p.r.l . , marsh t.r . , north r.c . , 2000 ,mnras , 317 , l41 maxted p.f.l . , heber u. , marsh t.r ., north r.c . , 2001 ,mnras , 326 , 1391 mazeh t. , goldberg d. , duquennoy a. , mayor m. , 1992 , apj , 401 , 265 mengel j.g ., norris j. , gross p.g . , 1976 , apj , 204 , 488 miller g.e ., scalo j.m . , 1979 , apjs , 41 , 513 morales - rueda l. , maxted p.f.l . , marsh t.r ., heber u. , 2002 , mnras , submitted moran c. , maxted p. , marsh t.r . , saffer r.a . ,livio m. , 1999 , mnras , 304 , 535 nomoto k. , nariai k. , sugimoto d. , 1979 , pasj , 31 , 287 orosz j.a ., wade r.a . , 1999 ,mnras , 310 , 773 paczyski b. , 1976 , in eggleton p.p . , mitton s. , whelan j. , eds , structure and evolution of close binaries , kluwer , dordrecht , p. 75podsiadlowski ph . , 2001 , in podsiadlowski ph . , rappaport s. , king a.r . , dantona f. , burderi l. , eds , evolution of binary and multiple star systems , asp conf .ser . , vol .229 , p. 239podsiadlowski ph , han z. , 2002 , in preparation podsiadlowski ph ., hsu j.j.l . , joss p.c . ,ross r.r . , 1994 , in clegg r.e.s ., stevens i.r . ,meikle w.p.s ., eds , circumstellar media in the late stages of stellar evolution , cup , cambridge , p. 187podsiadlowski ph ., joss p.c , hsu j.j.l . , 1992 , apj , 391 , 246 podsiadlowski ph . , pringle j.e ., rees m.j . , 1991 , nat , 352 , 783 podsiadlowski ph . , rappaport s. , pfahl e. , 2002 , apj , 565 , 1107 pols o.r . ,, eggleton p.p ., han z. , 1995 , mnras , 274 , 964 pols o.r . , schrder k .-p . , hurley j.r ., tout c.a ., eggleton p.p . , 1998 ,mnras , 298 , 525 pols o.r ., tout c.a ., schrder k .-, eggleton p.p . ,manners j. , 1997 , mnras , 289 , 869 pringle j.e ., webbink r.f . , 1975 ,mnras , 172 , 493 reimers d. , 1975 , mem .lige 6e ser , 8 , 369 renzini a. , 1981 , in effects of mass loss on stellar evolution , ed . c. chiosi & r. stalio ( dordrecht : reidel ) , 319 rogers f.g . ,iglesias c.a . , 1992 ,apjs , 79 , 507 saffer r.a . ,bergeron p. , koester d. , liebert j. , 1994 , apj , 432 , 351 saffer r.a ., livio m. , yungelson l.r . , 1998 , apj , 502 , 394 saio h. , nomoto k. , 1998 , apj , 500 , 388 saio h. , jeffery c.s . , 2000 ,mnras , 313 , 671 sargent w.l.w . , searle l. , 1968 , apj , 152 , 443 schrder k .-, pols o.r . , eggleton p.p . , 1997 ,mnras , 285 , 696 shore s.n . , 1988 , in esa , a decade of uv astronomy with the iue satellite , vol . 1 , p67 soberman g.e . , phinney e.s ., van den heuvel e.p.j . , 1997 ,a&a , 327 , 620 sweigart a.v . , 1997 ,apj , 474 , l23 taam r.e . ,sandquist e.l . , 2000 ,ara&a , 38 , 113 tauris t.m . , savonije g.j . , 1999 ,a&a , 350 , 928 thejll p. , ulla a. , macdonald j. , 1995 , a&a , 303 , 773 tout c.a ., eggleton p.p . , 1988 ,apj , 334 , 357 tutukov a.v . ,yungelson l.r . , 1979 , acta astron . , 29 , 665 ulla a. , thejll p. , 1998 ,a&as , 132 , 1 ulmer a. , fitzpatrick e.l . , 1998 , apj , 504 , 200 vauclair g. , liebert j. , 1987 , in y. kondo ed . , exploring the universe with the iue satellite ( reidel , dordrecht)1 , p. 355webbink r.f . , 1984 ,apj , 277 , 355 webbink r. f. , 1988 , in mikoajewska j. , friedjung m. , kenyon s. j. , viotti r. , eds , the symbiotic phenomenon .kluwer , dordrecht , p. 311webbink r.f . ,iben i.jr . , 1992 , in a.g .davis philip , d.s .hayes and j.w .liebert , eds , iau colloq .95 , the second conference on faint blue stars , l. davis press , p.445 wood j.h . ,saffer r. , 1999 , mnras , 305 , 820 yi s. , demarque p. , oemler a. jr . , 1997 , apj , 486 , 201 yi s. , lee y. , woo j. , park j. , demarque p. , oemler a.jr . , 1999 , apj , 513 , 128

<S> subdwarf b ( sdb ) stars ( and related sdo / sdob stars ) are believed to be helium core - burning objects with very thin hydrogen - rich envelopes . in recent years </S><S> it has become increasingly clear from observational surveys that a large fraction of these objects are members of binary systems . to better understand their formation , we here present the results of a detailed investigation of the three main binary evolution channels that can lead to the formation of sdb stars : the common envelope ( ce ) ejection channel , the stable roche lobe overflow ( rlof ) channel and the double helium white dwarfs ( wds ) merger channel . </S><S> the ce ejection channel leads to the formation of sdb stars in short - period binaries with typical orbital periods between 0.1 and 10d , very thin hydrogen - rich envelopes and a mass distribution sharply peaked around @xmath0 . on the other hand , under the assumption that all mass transferred is soon lost </S><S> , the stable rlof channel produces sdb stars with similar masses but long orbital periods ( 400 1500d ) and with rather thick hydrogen - rich envelopes . </S><S> the merger channel gives rise to single sdb stars whose hydrogen - rich envelopes are extremely thin but which have a fairly wide distribution of masses ( 0.4 0.65@xmath1 ) . </S><S> we obtained the conditions for the formation of sdb stars from each of these channels using detailed stellar and binary evolution calculations where we modelled the detailed evolution of sdb stars and carried out simplified binary population synthesis simulations . </S><S> the observed period distribution of sdb stars in compact binaries strongly constrains the ce ejection parameters . </S><S> the best fits to the observations are obtained for very efficient ce ejection where the envelope ionization energy is included , consistent with previous results . </S><S> we also present the distribution of sdb stars in the @xmath2 - @xmath3 diagram , the hertzsprung - russell diagram and the distribution of mass functions . </S><S> binaries : close stars : ob subdwarfs stars : white dwarfs </S>

0711.3801

the may 12 1997 cme occurred during solar minimum in an isolated active region near disk center .since the solar magnetic field during this period had minimum complexity , this event is especially favorable for detailed study .therefore , it was selected by several working groups ( shine , cism and muri ) for modeling and analysis .several observational features of the event have been reported @xcite , which has provided a good base for creating a magnetohydrodynamic ( mhd ) model of the event .for the latter , we have used the following three - step approach .first , a potential magnetic field is determined from the observed photospheric distribution of the normal field component .second , the configuration is sheared quasi - statically by photospheric vortex motions applied to the opposite polarities of the active region ( ar ) .third , an appropriate tangential electric field is applied then to the central part of the ar to cancel half of the photospheric magnetic flux .this third step is motivated by the observed flux - canceling properties of the photospheric motions in the event @xcite .theoretically , the third step is intended first to form a flux rope and then to destabilize it by detaching the rope field from the photospheric one , which has eventually to produce its eruption into the corona and heliosphere .this complete modeling program has begun by our group and it will be described in forthcoming papers .preliminary results for these simulations have been presented by @xcite . in this first paper, we consider a simplified model of the magnetic field and its structure at each of the above three steps in order to identify the generic structural features of the configuration prior to its eruption . by the term generic " we mean those features of the structure that are likely to remain in more accurate models of the ar magnetic field ( ar8038 at n21w08 ) .observationally , this field had an apparently simple bipolar structure @xcite dominating over a more complicated but distant and weak surrounding field .as a starting point , it is numerically extrapolated in the potential approximation from the synthetic magnetogram obtained by merging an mdi magnetogram with the corresponding synoptic map . from the methodological point of view , it is sensible , however , to begin with a simpler model by approximating the numerical potential field by a sum of two components , an azimuthally averaged large - scale magnetic field plus a bipolar field of the active region itself .this model , comprising essentially two bipolar field components with different scales and orientations , has a deceptive simplicity .it turns out that the structure of such a field is nontrivial even at the initial current - free state , and it becomes even more complicated at the shearing and flux - canceling phases , features that have important implications for understanding the event .the analysis of this structure prior to eruption is a major goal of the present paper .the paper is organized in the following way : the construction of the field model is described in sections [ s : ipmf ] and [ s : mhd_ev ] ; in section [ s : flcsg ] , the so - called squashing factor is generalized to the case of spherical geometry in order to investigate in section [ s : sif ] the structure of the modeled field .section [ s : imp ] describes the relationship between field line connectivity and the current distribution .our results are summarized in section [ s : s ] .some details of the initial potential field model are provided in the appendix .although the computation of potential magnetic fields is routine , the investigation of the field structure and its impact on the subsequent mhd evolution is a novel approach . for this reason, we prefer to idealize our initial field @xmath0 as much as possible by keeping only the most essential properties of the potential field computed from magnetogram data .since our primary interest lies in exploring the eruption process , the initial field @xmath0 ought to incorporate at the very least the large - scale coronal field @xmath1 , which we call the background field , and the field @xmath2 of the active region itself .thus , the resulting field can be written as @xmath3 where @xmath4 and @xmath5 are harmonic potentials of the magnetic field .it should be noted that although both fields have a similar bipolar character , they have in general different orientations and length scales : the background field varies on the scale of the solar radius @xmath6 , which greatly exceeds the size of the active region @xmath7 . for the sake of simplicity , we also assume that @xmath1 and , in particular , its photospheric radial component @xmath8 , is axisymmetric .the latter can be approximated by averaging the observed @xmath9 over the longitude @xmath10 and then by smoothing and fitting it to the sum @xmath11 where the coefficients @xmath12 are determined by a least square minimization . imposing additionally the flux - conservation condition and sacrificing the accuracy of fitting at the poles , where the observational data are not reliable because of the projection effect, we have found that equation ( [ brbp ] ) provides at @xmath13 and @xmath14 quite a good approximation to the large - scale field of the sun ( figure [ f1 ] ) .+ under these assumptions , the potential of the background field can be expressed in terms of spherical functions as @xmath15 where @xmath16 is a legendre polynomial of @xmath17-th order .the terms with @xmath18 are omitted here because the total magnetic flux must be zero and @xmath4 is defined up to an arbitrary constant .the unknown coefficients @xmath19 in this expression are determined by linear algebraic equations derived from the boundary or asymptotic conditions .the first number @xmath20 of such equations can be obtained by equating the corresponding linear combinations of @xmath21 , @xmath22 , at @xmath23 in @xmath24 to either 0 or @xmath12 depending on @xmath17 ( see equation ( [ brbp ] ) ) : the first and second case corresponds to @xmath25 and @xmath26 , respectively .these equations are rather cumbersome , but they can easily be retrieved , so we do not present them here explicitly .the form of the remaining @xmath20 equations for @xmath19 depends on what kind of behavior at large radii @xmath27 we would like to have for @xmath1 .the simplest choice is to require @xmath28 , for which the corresponding coefficient equations are simply @xmath29 and this is one of the two cases that we have considered .however , for modeling the solar wind effect , the second choice of boundary condition , namely , @xmath30 with the source - surface radius @xmath31 equal , say , to @xmath32 , is also of interest . according to equation ( [ fbg ] ) , this yields @xmath33 which reduces to equation ( [ bc2_1 ] ) at @xmath34 , as expected .the magnetic field under this condition becomes radial at the sphere of radius @xmath31 , and the corresponding approach is called the potential field source surface ( pfss ) model @xcite .this type of solution models an open field structure sustained by the solar wind , which makes it possible to estimate its impact on the coupling of @xmath1 and @xmath2 . for the analysis of the field line structure in both configurations , it is useful also to have an explicit expression of the magnetic flux function @xmath35 , which in our case is @xmath36 \times \nonumber \\ & & \qquad\qquad \left[\cos\theta p_{n}(\cos\theta ) - p_{n+1}(\cos\theta ) \right ] .\label{psi}\end{aligned}\ ] ] the contours of @xmath37 in the planes @xmath38 define the field lines of @xmath1 , since @xmath39 we chose @xmath6 to be a unit length scale in the problem , so that the coefficients of both type of solutions are calculated from the above equations at @xmath40 . table [ t1 ] in appendix [ s : pars ] presents their numerical values and shows that the coefficients @xmath41 of these two solutions are not very different , and the coefficients @xmath42 of the second solution are rather small .this suggests that the solar wind does not much affect the magnetic field in the lower part of the corona , in agreement with our study of the entire magnetic field @xmath0 described below .it is also in agreement with the structure of the field lines plotted as equidistant contours of the flux function @xmath37 in the plane @xmath43 of a cartesian system of coordinates @xmath44 .figure [ f2 ] clearly demonstrates that both structures are nearly the same everywhere accept for @xmath45 at equatorial latitudes .as expected , the source - surface configuration has a null line along the equator of the sphere @xmath46 to accommodate the direction reversal of the open field .the normal component of the photospheric magnetic field in our ar has a well - pronounced bipolar character .it is natural therefore to approximate the corresponding potential field with the help of a fictitious subphotospheric dipole , whose potential is @xmath47 this analytical expression was used to fit the numerical potential field solution @xmath48 computed from the original magnetogram data .we have optimized the location of the dipole @xmath49 and its moment @xmath50 to minimize the difference between @xmath51 and @xmath52 .more precisely , @xmath49 and @xmath50 are fitted to minimize the sum of @xmath53 evaluated at 14 points which were randomly scattered over the ar on the sphere @xmath54 .the radius @xmath55 is arbitrary to a certain extent , and its value was chosen equal to @xmath56 to filter out magnetic inhomogeneities of the size @xmath57 in the lower part of the solar atmosphere . in spite of its simplicity ,the described analytical model provides a relatively good approximation of the numerical potential field at @xmath58 ( see figure [ f3 ] ) .the model becomes even more accurate at larger @xmath27 , because higher - order harmonics describing the small - scale inhomogeneities in the numerical solution decay faster with @xmath27 than the lower - order ones .table [ t2 ] in appendix [ s : pars ] presents the resulting optimized parameters of the dipole in spherical coordinates for both types of the above boundary conditions .it shows that the magnetic moments @xmath59 of these two types of solutions differ less than @xmath60 from each other , while the differences between other parameters are even smaller .this is consistent with the above conclusion that the solar wind does not greatly influence the magnetic structure in the low corona .the above potential field with the boundary condition @xmath61 is used as the initial configuration for our simplified mhd model of the ar evolution prior to eruption .this field is not intended to describe the coronal field at any instant of its evolution , but rather to serve as a convenient starting point for further modeling . to energize this configuration ,it is quasi - statically sheared by the photospheric flows that preserve the initial photospheric radial component of magnetic field .this step produces a stressed configuration with a nearly force - free magnetic field .its subsequent evolution is driven by imposing a special photospheric distribution of the tangential electric field that models the observed flux - cancellation process . the initial potential field @xmath0 is sheared by photospheric flows such that its footpoints are rotated in the flux spots of the ar in opposite directions by following approximately the contours of the initial photospheric distribution of the radial field component @xmath62 .the corresponding shearing flow is described by @xmath63 where @xmath64 stands for the gradient operator tangential to the photosphere .this velocity field vanishes at the pil defined by @xmath65 and preserves the initial distribution of @xmath66 , since @xmath67 .the function @xmath68 \label{f}\end{aligned}\ ] ] provides an oval mask concentrating the shear in the ar near the pil , where @xmath69 is an amplitude linearly ramping until @xmath70 is reached at @xmath71 .after that @xmath69 remains constant until @xmath72 , when it drops linearly to zero at @xmath73 and remains zero from there on .the time unit @xmath74 must be large enough to guarantee quasi - static evolution of the whole magnetic configuration .specifically , it was chosen to be @xmath75 times the alfvn time in the ar , and 4 times the alfvn time in the polar field region .the angle variables are measured hereafter in radians and defined as @xmath76 where @xmath77 is a tilt of the oval , @xmath78 and @xmath79 are the radius - vectors of the photospheric points with longitudes @xmath10 and @xmath80 and latitudes @xmath81 and @xmath82 , respectively .the parameters @xmath83 and @xmath84 controlling the sizes of the oval mask are chosen to be @xmath85 and @xmath86 . the resulting photospheric velocity field determined by equations ( [ v_sh])([delta ] )is shown for the described set of parameters in figure [ f4]a .the process of flux cancellation is modeled by imposing at the photosphere the following tangential electric field @xmath87 where the flux function @xmath88 is determined from the two - dimensional poisson equation on the sphere @xmath89 such that @xmath90 the choice of a spatially uniform factor @xmath91 provides a simple uniform reduction of @xmath92 . however , the flux function @xmath88 and the related electric field @xmath93 decay in this case too slowly with distance from the ar compared to the respective magnetic field .this leads unfortunately to the undesirable formation of a strong boundary layer covering a substantial area outside of the ar . to localize @xmath93 within the ar , the mask function @xmath94 \label{g}\end{aligned}\ ] ]is used with @xmath95 and @xmath96 representing an approximate position of the ar center , and @xmath97 determining the extent of the electric - field localization .the amplitude @xmath98 here is linearly increased from @xmath99 until some @xmath100 is reached at @xmath101 .after that , @xmath98 remains constant until @xmath102 , when it drops linearly to zero at @xmath103 and remains zero from there on .the constant @xmath104 is chosen from the requirement to reduce the ar flux approximately to a half of its initial value in the specified interval of time .the resulting reduction of the photospheric normal magnetic field in the described flux - cancellation model is depicted in figure [ f4]b .as we will see , our initial configuration has a simple topology such that no magnetic null points ( nps ) or bald patches ( bps ) are present in the modeling ar .bps are segments of the pil at which field lines are concave towards the corona @xcite .an exception is the above - mentioned null line at the source surface @xmath46 , which accommodates the direction reversal of the open field at the equator in the pfss model .this null line , however , is outside of the ar , so its structure may have only geometrical features relevant for understanding the eruption process under study .the features of major interest are quasi - separatrix layers ( qsls , @xcite ) and , particularly their combination in the form of two self - intersecting qsls , the so - called hyperbolic flux tubes ( hfts , @xcite ) .this is because hfts have properties favorable for developing current sheets and reconnection @xcite , even if the respective configuration is free of nps and bps .the properties of such structural features and their manifestation in observed events are recently reviewed in detail by @xcite .these features are determined with the help of a point - wise analysis of the magnetic field line mapping . for small ars , where the photosphere can be approximated by a plane ,it is convenient to define this mapping in a cartesian system of coordinates @xmath105 .such coordinates allow us not only to determine the location of the footpoints but also to measure the distances between them in the @xmath106 and @xmath107 directions .the mapping @xmath108 defines the connection between one footpoint @xmath105 and another @xmath109 by a magnetic field line .notice that this mapping is defined only for closed field lines , while its generalization for open field lines requires an introduction of an upper boundary surface ( see below ) .the local properties of the field line mapping are described by the jacobian matrix @xmath110 whose elements , however , are not invariant with respect to the direction of mapping .therefore , they can not be used directly for characterizing the field line connectivity and for detecting qsls in configurations .one can prove that a natural quantity which satisfies this requirement in the case of a plane photospheric boundary is expressed in terms of the elements of @xmath111 as @xcite @xmath112 where @xmath113 is the so - called norm squared @xcite and @xmath114 is the jacobian of the mapping .to explain the origin of @xmath115 , note that the jacobian matrix represents a linearized field line mapping in the vicinity of a given footpoint .therefore , any infinitesimal circle centered at such a point will be mapped along the neighboring field lines into an infinitesimal ellipse at the other footpoint .it can be shown that the value @xmath116 determines the aspect ratio of such an ellipse [ see the details in @xcite ] .it characterizes the degree of squashing of the corresponding infinitesimal magnetic flux tubes and turns simply into @xmath115 for @xmath117 . since qsls are determined by large values of @xmath115 , we will neglect the slight difference between @xmath115 and the above exact value of squashing degree by calling @xmath115 itself the squashing degree or squashing factor @xcite .another independent geometrical quantity invariant to the direction of mapping ( up to the sign ) is even simpler than @xmath115 .it determines the degree of expansion or contraction of infinitesimal flux tubes and , due to the flux conservation in them , can be expressed in terms of the ratio of the normal field components at the connected footpoints as @xcite @xmath118 here @xmath119 and @xmath120 are normal components of the magnetic field at the conjugate `` target '' and `` launch '' footpoints , @xmath121 and @xmath105 , respectively . in practice ,the numerical calculation of @xmath122 through this ratio is more precise than that given by equation ( [ dlt ] ) and therefore it should be used for computing @xmath115 in equation ( [ q ] ) as well .the quantities @xmath115 and @xmath123 provide a valuable tool for investigating the geometrical structure of coronal magnetic fields .for example , they made it possible to convincingly demonstrate that hfts are a generalization of separator field lines to the case of topologically simple magnetic configurations having no nps or bps @xcite .this is particularly interesting for our study , since hfts appear naturally in quadrupole configurations @xcite , which can be thought of as being the superposition of two nested dipole fields exactly the type of potential field considered in the previous section .such a field - line mapping technique can be generalized in covariant form to the case where the magnetic configuration is bounded by curved surfaces of arbitrary shape @xcite . herewe will restrict ourselves to spherical boundaries and use a much simpler heuristic approach to derive @xmath115 in a global spherical system of coordinates . notice first that our coronal magnetic field fills the volume between two spheres of radii @xmath6 and @xmath31 .for measuring distances between infinitesimally close footpoints , one needs to introduce two - dimensional euclidean bases , which are tangential to these boundaries at each of their points . in the traditional spherical system of coordinates with the latitude @xmath124 or @xmath125 at the poles of the sun , these must be pairs of orthonormal vectors @xmath126. then the infinitesimal increments of length along @xmath127 and @xmath128 in one of the two polarities are @xmath129 and @xmath130 , respectively .let @xmath131 be the vector function representing the field line mapping , then similar increments at the conjugate footpoint are @xmath132 and @xmath133 with either @xmath134 or @xmath135 depending on whether the conjugate footpoint is located at the photosphere or at the source surface , respectively .thus , for the elements of matrix ( [ d ] ) , we obtain in our case @xmath136 it is these expressions that must be used for computing norm ( [ n ] ) in squashing factor ( [ q ] ) generalized for the case of both open and closed magnetic fields in the coronal volume restricted by spherical boundaries .the value of @xmath122 as a part of the whole expression for @xmath115 is more accurately computed through the ratio of the normal ( radial ) field components , as already noted before .the above expressions for @xmath137 and @xmath138 have apparent singularities at the poles , where they actually reduce in the generic case to resolved indeterminacies with @xmath139 and @xmath140 proportional to @xmath141 .such indeterminacies are due to pole singularities of the global spherical system of coordinates rather than special properties of the field line mapping .computationally , however , this requires care in evaluating @xmath115 near the poles , for example , by excluding them from numerical grid .a more accurate approach is to use a covariant definition of @xmath115 with two overlapping regular coordinate charts covering the spherical boundaries @xcite .the photospheric distributions of @xmath115 and @xmath123 are a two - dimensional imprint of the three - dimensional ( 3d ) coronal field structure and so are very helpful for its analysis . since their expressions ( [ q])([cd ] ) are applicable to arbitrary field configurations , we have computed representative examples of the @xmath115 and @xmath123 distributions for each of the three stages of magnetic field described above . in general ,if the magnetic configuration has a separatrix surface ( ss ) due to the presence of an np in the corona or a bp at the pil , the field line mapping experiences a jump at the ss footprints and the corresponding jacobian matrix is not defined there .numerical derivatives , however , are estimated as a ratio of finite coordinates differences , so their absolute values are limited .thus , numerically , the @xmath115 factor is always defined at the ss footprints and its narrow spiky distribution there helps identify np sss or bp sss , if present .as expected from section [ s : ipmf ] , the structure of the initial potential field in the ar turns out to be rather similar for both our models .therefore we will describe further only the structure whose background field is determined in the framework of the pfss model .the computed @xmath115 distribution reveals two very sharp ridges with maxima @xmath142 , one for each magnetic polarity ( see figure [ f5]a , b ) .each ridge spreads from the respective bipole spot toward the pole region , curving in a sickle - like shape .they trace the footprints of the qsls which combine into one flux tube , called an hft .the cross - section of the hft varies in a rather nontrivial way : neglecting some curving and twisting of the hft , its cross - section can be represented as follows @xmath143{ii1 } } \end{picture}.\ ] ] this is possible because the cross - section shrinks in the longitudinal direction and stretches in the transverse direction when following the field lines .the dependence of such a variation on the arc length of the field lines has an exponential character . because of this , resolving the internal structure of our hft is numerically challenging . for this purpose ,a curvilinear and nonuniform adaptive mesh with a minimal cell - size of @xmath144 has been developed . to control the effect of truncation errors ,the computation of @xmath115 has been performed both with 10 and 12 significant digits by using the computer algebraic system maple .the resolved profile of @xmath115 across the ridge of the @xmath115 distribution in its highest part is shown in figure [ f6 ] .similar to the case of a quadrupole magnetic field @xcite , the @xmath115 distribution has two peaks , which are probably related to the presence of two very localized minima of @xmath145 at the hft footprints ( see below ) .therefore , strictly speaking , the @xmath115 distribution has two sharp ridges closely bordering one another in each of the strips .however , because this fine - scale feature of the hft structure is not significant for the results we report here , we neglect it in the subsequent discussion .there is also a strong similarity between @xmath123 distributions of the quadrupole @xcite and our configuration . in both cases , a very steep gradient of @xmath123 exists across the ridges of the @xmath115 distribution . on both sides of the ridges in the middle part of the strips, there are areas of negative @xmath123 ( blue colored in figure [ f5]c ) .if crossing the strips in these areas outward from the pil , the @xmath123 distribution with moderate negative values steepen first into a sharp front of highly negative values .it is followed then by a sharper front of positive values ( red colored but not resolved in figure [ f5]c ) , which finally drops back to moderate negative values . if moving along the strips from their middle to their ends , the areas and the fronts of negative @xmath123 gradually disappear , while the fronts of positive @xmath123 increase in amplitude and broaden up to the entire width of the strips . by definition ,positive and negative values of @xmath123 correspond to expansion and contraction , respectively , of elemental flux tubes .this behavior of the flux tubes is opposite at the conjugate footpoints : if a given flux tube expands toward one footpoint , it contracts toward the other .thus , the conjugate footpoints of the field lines are colored in figure [ f5]c in red and blue of equal intensity . in the @xmath115 distribution ( figure [ f5]b ) ,the conjugate footpoints are colored in grey of equal intensity . in combination with the pil and contours of the photospheric @xmath146 superimposed on such distributions , this provides a comprehensive graphical representation of the field line connectivity .one can see from this representation that most of the field lines in the hft connect the end and mid parts of the strips or , in other words , the areas of stronger and weaker magnetic field .the field lines are organized in the hft in such a way that changing footpoints across the middle part of a given ridge causes swiping of the conjugate footpoint along the other ridge between the ends of the strips . in the case of the quadrupole configuration in each polarity ,the ridges connect two magnetic flux concentrations . in our present model, they similarly connect the flux concentrations of the ar with the respective polar magnetic `` caps '' .the grey - scaled @xmath115 distribution and the red - blue scaled @xmath123 distribution can be superimposed on @xmath147-plane by using logical multiplication of pixel colors .both distributions can then be combined in one colored graph , allowing one to easily compare these different characteristics of the field line mapping .to more clearly illustrate the properties of @xmath115 and @xmath123 , we have mapped the resulting image on the spherical surface and superimposed several contours of the corresponding photospheric @xmath148 component on it .the resulting image shown in figure [ f7 ] summarizes the above discussed properties of the field line mapping in the hft . to better understand the hft structure , it is also very instructive to trace the field lines starting at the ridges of the @xmath115 distribution . in accordance with the above discussion, the conjugate footpoints are located not along the other ridge but rather across it and concentrated on a short arc passing through the point of a local maximum in @xmath115 .each of such two sets of field lines forms a mid - surface of the corresponding qsl , which is further called a quasi - separatrix surface ( qss ) .these two surfaces intersect along a field line that we label the quasi - separator ( figure [ f7 ] ) for the following two reasons .first , this field line is characterized by a maximum value of the squashing factor @xmath115 and it becomes a genuine separator field line in quadrupole configurations when their flux concentrations shrink to point sources @xcite .second , it is characteristic of the separator to pass through two magnetic null points of the configuration .the quasi - separator partly inherits this property in the sense that its footpoints are located near very localized minima of the magnetic field ( figure [ f5 ] ) . in general ,the presence of such minima in a given configuration is a good indication that the configuration also contains an hft .determining the minima of the photospheric distribution of @xmath145 is much easier than analyzing the magnetic connectivity in the above manner .therefore , an initial scanning of the photospheric field minima can provide a quick estimate of the presence of hfts in more complex configurations .it is necessary , however , to check additionally that the minima found are not due to the coronal magnetic nulls located right above the minima .this test can be done , for example , by plotting several iso - surfaces of @xmath149 to prove whether they do cover the minima and the gradient of @xmath145 has the proper direction . in the above considered configuration , such a test indeed confirms the absence of the coronal nulls above the photospheric field minima . the difference between the quadrupole case @xcite and our model is mainly in the type of magnetic flux distribution in one of the two bipoles of the configurations . in the quadrupole configuration ,the flux is concentrated in two spots of the same ar , while in our model , it is distributed over large solar hemispheres .such a large length scale leads to a significant increase in our model of maximum values of @xmath115 , which are four orders of magnitude higher than in the quadrupole field . in all other respects , the magnetic fluxes in both configurationsare partitioned by hfts in a rather similar way .figure [ f8 ] ( a)(c ) shows the photospheric @xmath115 distribution together with some of the new qsss developed in the configuration after it is subject to the shearing / twisting deformation described in section [ s : sh ] .it can be clearly seen from figure [ f8 ] that the @xmath115 distribution outside the ar changes very little after shearing , which implies a correspondingly small change in the large - scale hft initially present in the potential field .however , @xmath150 a circular vortex motion centered at some point @xmath151 can be defined as a differential rotation @xmath152 acting on a flux spot for time @xmath153 with angular velocity , say , @xmath154^{2 } \right\ } .\end{aligned}\ ] ] the size of the vortex ( @xmath155 ) corresponds here to the size of the flux spots . for simplicity ,let the vortices be the same in positive and negative polarities and centered at points @xmath156 and @xmath157 , respectively .assuming now that the configuration evolves according to ideal mhd , so that the field line connectivity is preserved , we can compose our field line mapping at time @xmath153 as @xmath158 this equation means that the conjugate footpoint is obtained in three steps : first , a given footpoint is tracked back in time by @xmath159 to get its initial location , from which @xmath160 provides then the initial location of the conjugate footpoint , whose current location is finally produced by @xmath161 .each of the three mappings @xmath161 , @xmath160 , and @xmath159 are incompressible , and so is their composition @xmath162 .this implies that @xmath163 in equation ( [ q ] ) and so the squashing factor @xmath115 coincides with @xmath164 defined by equation ( [ n ] ) , where the elements of the jacobian matrix can easily be obtained from the above definition of @xmath162 .figure [ f9 ] represents two calculated examples of @xmath115 distributions for symmetric and asymmetric cases , where @xmath157 and @xmath165 are equal and not equal , respectively .it is natural to call such cases correspondingly concentric and eccentric .one can see from these examples that even a relatively small `` eccentricity '' in the differential rotations at the opposite polarities causes a substantial effect .compared to the concentric case , @xmath115 increases here by an order of magnitude when each core of the vortices makes approximately one - half turn .the larger is the eccentricity in the vortices , the stronger is the squashing of elemental flux tubes at certain layers of the resulting twisted configuration .this has important implications for the current density distribution in the configuration under study , and also in a more general context , which is further discussed in section [ s : imp ] .figure [ f8 ] ( d)(f ) shows the @xmath115 distribution and qsss at the final phase of the computation , after flux cancellation has occurred at the pil .the small - scale hfts that are formed after the shearing phase have now increased in size .this occurs partly because of a noticeable migration of the conjugate footprints outward from the pil , which is accompanied by reconnection in the hfts due to a large enhancement of the current density in layer - like structures that will be discussed in section [ s : imp ] .our simulations confirm also that the photospheric flux - canceling process builds a twisted magnetic flux rope at the pil @xcite .the inverse - s - like footprint of the respective bp sss appears at this phase in the photospheric @xmath115 distribution by outlining the footprints of the flux rope [ figure [ f8](e - f ) ] in the way predicted earlier for a similar twisted structure by @xcite .here it is evident how universal our technique is it enables us to identify and locate all the structural elements of the magnetic field under study , including hfts and bp sss , as well as flux ropes . as a matter of fact, the very definition of a flux rope in a given asymmetric 3d configuration can be made precise via this technique .the flux rope is no more than a fraction of twisted field lines delimited from the surrounding configuration by bp sss and/or helical qsls [ see another example of the rope in @xcite ] .the presence of such delimiters " in the field configuration makes it possible to consider this fraction of field lines as a separate object , namely , a flux rope .unfortunately , the boundary conditions chosen for modeling the flux - canceling phase of this run did not produce an eruption , even after reducing the magnetic flux of the ar to half of its initial value .the fraction of the canceled flux that has been converted into the flux of the rope is not large enough to reach an eruptive unstable state in this particular computational run .so the evolution of the magnetic structure at the eruptive phase is still to be investigated , which we plan to study after clarifying the eruptive boundary conditions for the flux - canceling process .we expect that the evolution of the field structure prior to the eruption will remain similar because the hfts and bp ss we have found are stable structural features that will survive under a wide variety of conditions .one can anticipate also how the magnetic field structure will change during the initial stage of the eruption process . based on the results of @xcite, we expect that detachment of the unstable flux rope will cause first a bifurcation of the bp ss into a pair of bp sss intersecting along a generalized separator field line. these bifurcated bp sss will subsequently transform into an hft located below the rising flux rope , as demonstrated by @xcite for a similar equilibrium configuration .such a transformation , however , will be a highly dynamical process , pinching this new hft into a vertical current layer ( cl ) , similar to the one described in previous work @xcite . the magnetic reconnection in this cl ought to play a crucial role in making the event eruptive ( see sections [ s : imp ] and [ s : s ] ) .another important conclusion that follows from our example is that potential and non - potential magnetic fields with the same photospheric flux distributions may have very different structures in both a topological and geometrical sense .this example undermines therefore a simplifying assumption often used in the analysis of solar magnetic fields , namely that the topologies calculated in potential and non - potential approximations are similar . @xcitealso argued against this assumption by proposing a counterexample of a constant-@xmath166 force - free field that has a qualitatively different topology compared to the potential field with the same photospheric normal component of magnetic field .however , @xcite rightly argued against this counterexample by pointing out that it corresponds to a symmetric configuration , which is topologically unstable to small perturbations of the field .their consideration of asymmetric example supported the opposite conclusion that the potential field model generally provides a good approximation for the topology of the observed configurations .our configuration is also asymmetric , but it does not support this conclusion , at least for the ar , where it has essentially nonliner force - free field .this means that the results of topological analysis of realistic magnetic configurations based on the potential approximation can be misleading and should be handled with caution .photospheric shearing and twisting motions inflate the magnetic field of the ar , which presses up against the overlying large - scale hft . the hft footprints , however , being mostly outside of the applied photospheric vortices , are essentially not affected by these motions , so the line - tied hft is stretched and pinched to counter the magnetic pressure from the inflating ar field .this process implies an accumulation of the current density in the hft such that its x - type mid - cross - section is noticeably pinched in the vertical direction into a @xmath167-like shape with a relatively strong cl at the center ( figure [ f10 ] ) .the numerical grid used in our mhd code does not enable us to resolve the resulting thin cl and its four adjoining branches in its neighborhood .although the cell size of the grid in this region @xmath168 is several times smaller than the apparent thickness of the cl , it is still much larger than the characteristic thickness of the corresponding qsl ( see figure [ f6 ] ) , estimated as @xmath169 @xcite . nevertheless , the asymptotic tail distribution of such a cl is consistently retrieved by the code with an evident correlation between local maxima of squashing factor @xmath115 and parameter @xmath170 at both shearing / twisting and flux - canceling phases .a similar correlation has been found by @xcite in mhd simulations of the quadrupole magnetic configuration in cartesian geometry , where the corresponding hft has been perturbed by distorting its footprints rather than inflating the underlying structure as in our case .this is consistent with the structural similarity of our model and the quadrupole configuration as discussed in section [ s : sipf ] .one can expect therefore that this cl would lead to reconnection if eruptive conditions were met at the cancellation phase , to produce a rapidly rising flux rope that would push up more strongly against the overlying field structure .such a cl is an essential element of eruption models in quadrupole configurations with an emerging inner bipole without shear @xcite or with strong shear @xcite , or with an embedded flux rope @xcite , as in our case .we will argue in section as discussed in section [ s : s / t ] , two small - scale hfts are formed in the ar during the shearing / twisting phase and they survive during the flux - cancellation phase of our simulations [ panels ( a)(c ) and ( d)(f ) of figure [ f8 ] for the former and latter phases , respectively ] . in both phases , these hfts carry strong layer - like currents , as evidenced by the sharp angles at which the corresponding qsss intersect each other ( see figures [ f8]b and [ f8]e ) . in current - free hfts such angles would be close to right angles .figure [ f11 ] shows an evident correlation between local maxima of @xmath115 and @xmath166 and therefore confirms that thin cls are indeed formed along the hfts .one can also see from this figure that @xmath166 is positive in the cls and that they are located at the border of the twisted ar field , with negative @xmath166 in the bulk volume .so these cls carry a significant part of the return current that shields the twisted ar field from a nearly potential background field .their formation occurs in the coronal volume due to the interlocking of the field lines by a twisting pair of shearing motions eccentrically applied to the conjugate footpoints .in essence , this is another natural realization of parker s mechanism of current layer formations discussed by @xcite and @xcite for quadrupole configurations .this process has been called magnetic pinching . comparing the shearing / twisting and flux - cancellation phases [ panels ( a)(f ) and ( g)(n ) , respectively , in figure [ f11 ] ] , we see that the cls increase in size in the latter phase , where a noticeable migration of their footprints outward from the pil takes place . since the magnetic reynolds number based on the local velocities and length scale is not smaller than @xmath171 , this migration is likely because of the reconnection that occurs in the cls rather than due to magnetic diffusion at the photospheric boundary .such cls help redistribute magnetic field and current within the ar at the flux - cancellation phase .more detailed investigation of this process is required , especially for the case when eruption occurs .it is clear , however , that its significance extends far beyond this particular simulation .indeed , magnetic sunspots with twisting - type motions are quite natural and are often observed on the sun @xcite .for this reason , such motions have been considered for a long time as an in situ mechanism to build up magnetic energy in the solar corona .our example suggests , however , that the differential rotation of sunspots may also release this energy in the accompanying cls by magnetic reconnection triggered when the pinching of hfts reaches a critical level .it should also be noted that the largest values of @xmath166 are obtained in our simulations along the flux rope throughout its cross - section ( panels ( g)(n ) in figure [ f11 ] ) rather than in the cls .there are two reasons for this : first , the perturbation of the initial configuration by shearing / twisting motions may not be sufficiently high to form stronger cls , and second , these cls are not well - resolved in our simulations .the latter may also be the reason why a cl is not formed along the bp ss that wraps around the flux rope , although the formation of such a cl is generally expected in this configuration @xcite , even if it evolves quasi - statically .this is consistent with the numerical simulations of @xcite , where such a cl is formed along the bp ss only in the dynamical kink - unstable phase of the evolution of a similar configuration .an additional separate study is required to clarify these issues .we have developed a simplified model of the coronal magnetic field prior to the cme on may 12 1997 in the following three steps .first , this field is constructed by superimposing a global background field @xmath1 and a localized bipolar field @xmath2 to model the active region ( ar ) in the potential approximation .the background field @xmath1 is determined from the observed photospheric normal field averaged over the longitude of the sun .the ar field is modeled by a subphotospheric dipole whose parameters were optimized to fit the magnetic field obtained from an mdi magnetogram . to verify the impact of the solar wind on the field structure, @xmath1 is calculated for two different sets of boundary conditions . in the first case , @xmath1 is modeled as a closed potential field vanishing at infinity . in the second case , @xmath1 is constrained to be strictly radial and therefore open at the sphere of radius @xmath172 .the resulting field structures in the ar and its neighborhood are quite similar for both choices of boundary conditions , therefore only the first ( closed ) potential configuration has been used in the next step of mhd modeling . in this second step ,the potential field is quasi - statically sheared by photospheric vortex motions applied to the two flux concentrations of the ar .finally , the resulting force - free field is evolved by canceling the photospheric flux with the help of an appropriate tangential electric field applied to the central part of the ar . to investigate the structure of the modeled configurations , the field line mapping technique has been generalized to the case of spherical geometry . in our potential field configuration, this technique reveals a large - scale hyperbolic flux tube ( hft ) characterized by very large values of the squashing factor . as in quadrupole - type configurations ,such an hft consists of two intersecting quasi - separatrix layers ( qsls ) that provide an approximate partition of magnetic fluxes between four photospheric field sources . in our case ,the role of such sources is played by two flux concentrations of the ar and two photospheric polar regions .it is important that no other structural features , such as magnetic null points or `` bald patches '' , are found in the ar and its surrounding area of the initial potential configuration . in this respect ,our initial configuration drastically differs from that used in the breakout model @xcite , where the null points and separator are key topological elements of the structure . on the other hand ,our large - scale hft is a generalized analogue of the separator field line @xcite and is therefore a favorable site for current layer ( cl ) formation @xcite and magnetic reconnection @xcite .the accumulation of current density observed in the vicinity of this hft during the shearing / twisting phase of our simulations is in agreement with this point of view . the mid x - type cross - section of the hft pinches in the process of this accumulation into a @xmath167-like shape with a strong cl at the center .this structure is preserved during the cancellation phase when the flux rope is formed .had the eruption conditions been met in this phase , the flux rope that would form would most likely detach from the dense photosphere and rise by pushing the overlying field upward .this would cause an additional pinching of the cl to subsequently trigger reconnection within it .such a reconnecting cl is a basic element of the breakout model @xcite and we have demonstrated that it appears in the flux - cancellation model as well if the interaction of the ar with the global background field is included . with this generalization ,the only difference between the models is in their interpretation of the drivers of the eruption .the breakout model assumes that the driver is reconnection in the cl , allowing for an explosive expansion of the sheared arcade of the ar .the flux - cancellation model assumes instead that the driver is a rising flux rope unleashed from the strapping field of the ar .while the reconnecting cl is important in our generalized model , the latter mechanism appears to be more likely to drive the eruption .the following estimate of the magnetic fluxes in the interacting parts of our configuration suggests a picture of how this occurs . the flux associated with the background field is eight times larger than the initial flux in our ar .however , only certain fractions of these fluxes interact with each other .these fractions can be estimated from the geometrical properties of the quasi - separator shown in figure [ f7 ] . since the longitudinal extension of the quasi - separator is about 0.55 radians ( fig .[ f5]b ) , the fraction of the background magnetic flux overlying the quasi - separator is equal to @xmath173 . we must take into account also that only field components perpendicular to the quasi - separator can participate in reconnection ( reducing the fraction of the overlying flux from 0.7 to 0.4 ) , and that the flux of the ar is reduced to half of its initial value at the end of the flux - cancellation phase . with these corrections , we obtain the interacting fractions of the background and ar fluxes expressed by the ratio @xmath174 . in other words ,the interacting fluxes of the background and ar field are comparable .it implies that a substantial part of @xmath175 must be converted into the magnetic flux associated with the rising unstable flux rope in order to allow it to break through the overlying background field .such a conversion is likely provided by magnetic reconnection in the `` vertical '' cl below the rope , as observed in previous simulations by @xcite and @xcite .this process ought to replenish the magnetic flux in the rope that is gradually destroyed by reconnection in the horizontal " cl above the rising rope .we have also demonstrated that two small - scale hfts pinched into thin cls are formed by nonuniform vortex motions applied to the flux spots of the ar .the reconnection in these cls helps redistribute the magnetic flux between the ar and the surrounding field during the flux - cancellation phase .other possible implications of their presence should be explored in a more accurate model of the magnetic field evolution in this event . however, beyond its manifestation in this particular event , the discovered process of current layer formation by rotation of the flux spots has a broader significance .this process seems to be very important for understanding other eruptive / flaring events driven by rotating sunspots @xcite .our example shows that their rotation can serve as a mechanism of both the storage of free magnetic energy and its release via gradual formation of the above cls in the solar corona .more detailed studies are certainly required in this direction .we have also confirmed that the photospheric flux - canceling process builds a magnetic flux rope at the polarity inversion line @xcite .this flux rope appears as a well - defined element of the structure owing to the bald patch separatrix surface that wraps around the rope .the magnetic structure near the rope is topologically equivalent to the twisted configuration described by @xcite . in particular, the bald patch separatrix surface has a characteristic inverse - s - like shape .this type of sigmoid corresponds to the negative sign of the current helicity in the flux rope , which is in accordance with observations @xcite .titov is thankful to bernhard kliem for several useful discussions on magnetic eruptions .this research was supported by nasa and the center for integrated space weather modeling ( an nsf science and technology center ) .table [ t1 ] presents the coefficients of spherical harmonics for potential background magnetic fields @xmath1 with two different types of boundary conditions . in both cases @xmath176is prescribed at @xmath177 by defining the neumann boundary condition ( [ bc1 ] ) for the harmonic potential @xmath4 . in the first case, the additional asymptotic condition at infinity @xmath178 is required to provide a standard exterior neumann problem for the laplace equation with the spherical boundary @xmath89 . in the second case ,the dirichlet boundary condition @xmath179 at @xmath180 is imposed instead to define a mixed boundary value problem for @xmath4 in the coronal volume @xmath181 , which is a so - called pfss model .table [ t2 ] presents the optimized parameters of the magnetic dipole in spherical coordinates to model the active - region field @xmath2 through the gradient of the respective dipole potential ( [ far ] ) .rrrrr 1 & @xmath182 & & @xmath183 & @xmath184 + 2 & @xmath185 & & @xmath186 & @xmath187 + 3 & @xmath188 & & @xmath189 & @xmath190 + 4 & @xmath191 & & @xmath192 & @xmath193 + 5 & @xmath194 & & @xmath195 & @xmath196 + 6 & @xmath197 & & @xmath198 & @xmath199 + 7 & @xmath200 & & @xmath201 & @xmath202 + 8 & @xmath203 & & @xmath204 & @xmath205 + 9 & @xmath206 & & @xmath207 & @xmath208 + 10 & @xmath209 & & @xmath210 & @xmath211 + 11 & @xmath212 & & @xmath213 & @xmath214 + rcc @xmath215 & @xmath216 & @xmath217 + @xmath218 & @xmath219 & @xmath220 + @xmath221 & @xmath222 & @xmath223 + @xmath59 & @xmath224 & @xmath225 + @xmath226 & @xmath227 & @xmath228 + @xmath229 & @xmath230 & @xmath231 +

<S> a simple model of the coronal magnetic field prior to the cme eruption on may 12 1997 is developed . </S><S> first , the magnetic field is constructed by superimposing a large - scale background field and a localized bipolar field to model the active region ( ar ) in the current - free approximation . </S><S> the background field is determined from the normal component of the observed photospheric magnetic field averaged over the longitude of the sun . </S><S> the ar field is modeled with the help of a subphotospheric dipole whose strength , location , and orientation are optimized to fit the magnetic field obtained from an mdi magnetogram . </S><S> second , this potential configuration is quasi - statically sheared by photospheric vortex motions applied to two flux concentrations of the ar . </S><S> third , the resulting force - free field is then evolved by canceling the photospheric magnetic flux with the help of an appropriate tangential electric field applied to the central part of the ar . to understand the structure of the modeled configuration , we use the field line mapping technique by generalizing it to spherical geometry . </S><S> it is demonstrated that the initial potential configuration contains a hyperbolic flux tube ( hft ) which is a union of two intersecting quasi - separatrix layers . </S><S> this hft provides a partition of the closed magnetic flux between the ar and the global solar magnetic field . </S><S> such a partition is approximate since the entire flux distribution is perfectly continuous . </S><S> the vortex motions applied to the ar interlock the field lines in the coronal volume to form additionally two new hfts pinched into thin current layers . </S><S> reconnection in these current layers helps to redistribute the magnetic flux and current within the ar in the flux - cancellation phase . in this phase </S><S> , a magnetic flux rope is formed together with a bald patch separatrix surface wrapping around the rope . </S><S> other important implications of the identified structural features of the modeled configuration are also discussed . </S>

1502.03802

the current state - of - art coding method like hevc @xcite usually relies on a block based hybrid scheme . a block in a video frame is first predicted through the best inter- or intra - predcition , and the prediction error ( residual )is then passed into a transform coder , where a fixed orthogonal transform such as dct is used .the resulting transform coefficients are then quantized and the non - zero dct coefficients , as well as the prediction candidate locations ( e.g. the motion vectors in the case of inter - prediction ) and/or modes are finally coded using an entropy coder . in @xcite, we proposed a new video coding scheme that directly represents each block in a video frame using a self - adaptive dictionary consisting of all possible temporal prediction candidate blocks over a search range in the previous frame . here in this work ,we propose an extension of that by including a second stage that codes the residual after approximating the current block using the chosen atoms in the first stage .the rationale for this second stage is that we have identified the self - adaptive dictionary loses its representation efficiency after first few chosen atoms .this is because the atoms in the self - dictionary may be highly correlated and concentrated in a sub - space .we code the residual after the second stage using a non - redundant orthonormal dictionary that spans the null - space of the sub - space spanned by chosen atoms in the first stage . in our current implementation, we derive this second dictionary by orthonormalizing all dct transform basis blocks with respect to the chosen atoms in the first stage .we would like to emphasize that the proposed two - stage framework differs significantly from the state - of - art video coding standards . specifically , although our self - adaptive dictionary consists of conventional motion search candidate blocks and we use a linear combination of a few motion search candidates for the first stage representation ( which we term as atoms as in sparse recovery literature ) , we allow any combination of possible candidates and the weights are solved through an optimization problem ; we do not restrict the number of chosen atoms to a pre - set number , but minimizing this number through minimizing the l0-norm of the weights . by comparison, state - of - art coding standards like hevc only permit a fixed linear combination of several adjacent motion candidates through the use of fractional - pel motion compensation , where the weights are implicitly determined by the interpolation filter for deriving fractional - pel motion vectors .besides , our second stage orthonormalizes the fixed dct basis with respect to the block - dependent first stage chosen atoms .effectively , the orthonormalization in the second stage adaptively alters the dct basis to span over the null space of the first stage reconstruction and hence is more efficient in representing the residuals .to the contrary , the transform coding stage in all current video coding standards is fixed .several research groups have attempted using redundant dictionaries for block - based image and video coding , including @xcite . inall reported dictionary - based video coders , the dictionary atoms are used to represent the motion - compensated residual block for intra / inter - frame video coding .therefore , they are very different from what is proposed here .the work in @xcite codes each frame in the intra - mode , with a dictionary that is updated in real time based on the previously coded frames .although such online adaptation can yield a dictionary that matches with the video content very well , it is computationally very demanding. furthermore , the dictionary is not block adaptive .the proposed framework uses a self - adaptive dictionary that only depends on the block location , without requiring realtime design / redesign of the dictionary .two major challenges have to be addressed for making the two - stage framework efficient .the first is the fact that the redundant dictionary is highly correlated . directly quantizing the coefficients with the highly correlated atomsis inefficient . in this work , as our original work in @xcite , we recursively orthonormalizes the chosen atoms during the iteration of omp .however we have developed an improved omp algorithm to solve the sparse recovery problem , which can yields a sparser representation than the original omp method @xcite .the coefficients associated with the orthonormalized chosen atoms can be quantized independently without losing efficiency .the second challenge is how to determine the switch point from the first stage to the second stage . in this work, we switch to the second stage when the first stage self - adaptive dictionary begins to slowly correct the residual , which indicates that the remaining atoms in the self - adaptive dictionary have limited representation power for the remaining residual . in the remainder of this paper , we describe the proposed two - stage framework in sec .[ sec : spdcp ] ; we present the entropy coder design in sec .[ sec : ent ] ; in sec .[ sec : res ] we show the rd performance and comparisons .we conclude the paper in sec .[ sec : con ] .the proposed two - stage framework is illustrated in fig .[ fig : framework ] . we describe the different components in more details below .the first stage is similar to the one reported in @xcite . for each mean - removed blockrepresented as a vector @xmath0 , denote the dictionary by @xmath1 , which consists of the mean - removed vector representation of all possible motion search candidate blocks corresponding to integer motion vectors in a preset motion search range , and the coefficients by @xmath2 .we find the sparse set of coefficients by solving the following l0-norm minimization problem .@xmath3 we set @xmath4 to be slightly larger than the expected distortion introduced by the quantization stepsize @xmath5 , i.e. @xmath6 .because of the non - convexity of l0-norm , greedy algorithm is often used to solve this problem .one prominent algorithm is the orthogonal matching pursuit ( omp ) . in the original omp algorithm@xcite , at every iteration , the correlation between the residual and the remaining atoms are evaluated , and the best correlated atom is chosen ; a least square step is then used to update all the coefficients , where the coefficients are related to the original chosen atoms .the least squares update step effectively makes the residual to lie in the null space of the chosen atoms . in @xcite , we implemented the original omp algorithm with embedded orthonormalization of chosen atoms so that we could solve the sparse reconstruction and atom orthonromalization simultaneously .this is accomplished by orthonormalizing the latest chosen atom with respect to all previously chosen atoms using a gram - schmidt procedure , and finding the coefficient associated with the orthonormalized latest chosen atom by a simple inner product . in this work ,we use an improved version of omp , denoted by eomp @xcite . in each iteration , instead of evaluating the correlation between the residual and the remaining atoms , we first update the remaining atoms by orthonormalizing them to all previously chosen orthonormalized atoms . this orthonormalization can be done efficiently through a recursive one - step orthonormalization procedure .let @xmath7 denote the @xmath8-th remaining atoms , @xmath9 the last chosen atom , the one - step orthonormalization is given by the following : @xmath10 the correlation between the residual and orthonormalized remaining atoms are then evaluated and the atom with the largest correlation ( i.e. the magnitude of the inner product ) is chosen . note that this chosen atom is already orthonormalized and the coefficient corresponding to this atom has already been calculated . by recursively updating the remaining atoms , we have shown in @xcite that this improved eomp has better sparsity recovery capability than the original omp algorithm . the early termination constraint in eq .[ eq : l0 ] is expressed in the terms of residual norm reduction ratio threshold @xmath11 , a threshold on the ratio between the residual norm difference of the current and last iterations and the residual norm of the last iteration .this constraint is specially designed to terminate the omp algorithm before reaching the fidelity constraint if newly chosen atom does not decrease the residual norm significantly , which usually signifies the low correlation of the residual and the remaining atoms and the loss of the representation power of the remaining atoms in the self - adaptive dictionary .obviously , the choice of @xmath11 will affect the overall coding efficiency and must be chosen appropriately .once all the orthonormalized atoms denoted by a matrix @xmath12 and their corresponding coefficients @xmath13 are found , we apply uniform quantization with deadzone to each coefficient with the same stepsize @xmath5 , with the deadzone @xmath14 , following the hevc standard recommendation for inter - coded block . denoting the quantized coefficients by @xmath15 , the first stage reconstructed blockcan be represented by @xmath16 , and the residual is @xmath17 .we start our discussion with a plot demonstrating the motivation for designing the second stage . in fig .[ fig:2stageprof ] , the continuous black curve shows the residual norm versus number of chosen atoms ( @xmath18 ) from a particular block in sequence _ touchdownpass _ , when only self - adaptive first stage is used . the beginning part of the curve shows a very fast decay of the residual norm , which indicates the effectiveness of the self - adaptive dictionary .however , after few iterations , the decaying rate significantly slows down for each newly chosen atom .this suggests that the remaining atoms after a few iteration may lie largely in the same subsapce as the chosen atoms and can not represent efficiently the residual , which is orthogonal to this subspace .in other words , the self - adaptive dictionary may not span over the full space @xmath19 and it is very inefficient to reduce the residual after a few iterations .in addition to the continuous black curve , we also show two red curves originating from two specific @xmath18s .they represent the residual norm reductions using dct basis had the first stage stopped at that @xmath18 . in the left red curve , it is apparent that stopping first stage too early yields suboptimal result , i.e. at that stage , self - adaptive dictionary is still more efficient .interestingly , the right red curve shows an opposite trend , that using dct basis becomes more efficient in reducing the residual norm compared to continuing with the self - adaptive dictionary .we would like to note that the horizontal axis of fig .[ fig:2stageprof ] is in terms of @xmath18 rather then the bit rate used to represent @xmath18 coefficients . considering the huge size difference between the self - adaptive dictionary ( in our experiments , the number of atoms is 2304 ) and dct dictionary ( 256 for @xmath20 block size ) , greater rate savings will be obtained by switching to dct at the starting point of the right red curve . , comparing using the self - adaptive atoms alone , vs. using two stage . ] now we give the description of the second stage coder . given the residual @xmath21 , the dct basis vectors are first orthonormalized with respect to the first stage chosen atoms .the coefficients of the residual with respect to these altered dct basis vectors are then derived simply by performing inner product of the residual vector with each altered dct basis vector .these coefficients @xmath22 are quantized with quantization stepsize @xmath23 using the same deadzone quantizer .the residual @xmath21 is therefore reconstructed as @xmath24 , where @xmath25 contains the altered dct basis vectors and @xmath26 are the quantized second stage coefficients .finally the reconstructed block @xmath27 is given by @xmath28 .the information we need to code in the first stage consists of mainly three parts : 1 ) the location of the chosen atoms ; 2 ) the order of the chosen atoms , since decoder needs such information to perform the orthonormalization using the same order as the encoder ; and 3 ) the levels of each quantized coefficients . the second stage codes the residual after the first stage using an approach similar to the residual coding method of hevc @xcite using dct , but with a fixed transform block size equal to the block size .for the first stage , we have developed a new content adaptive arithmetic entropy coder inspired from the cabac design in hevc @xcite .we would like to emphasize that one major difference between the proposed first stage coder and the cabac coder in hevc is that our coder needs to code the order of chosen atoms to enable the decoder to perform the same orthonormalization . as with cabac in hevc, we define two probability update modes , the normal mode when the probability for coding each bin is updated based on the past data ; and the bypass mode when such probability is fixed .we indicate which atoms are chosen by a 2d binary significance map .for example , if the search range is @xmath29 , the map size is @xmath30 , with the top - left element corresponding to the candidate block that is shifted from the current block position by @xmath31 .an element is assigned `` 1 '' if the corresponding atom is chosen , and `` 0 '' otherwise .we code the significance map using a three - layer multi - resolution approach . for the search range of @xmath32 , the bottom , middle , and top layer have sizes of @xmath30 , @xmath33 , and @xmath34 , respectively , as shown in fig .[ fig:3layers ] .each element in an upper layer corresponds to a @xmath35 region in the next layer .the encoding process starts from the top layer .every time when a `` 1 '' is coded , the subsequent @xmath35 block should be encoded until the bottom layer is reached .every pixel in the top @xmath34 layer is coded directly , with a position dependent probability . in the subsequent layers , for every @xmath35 block that corresponds to a `` 1 '' in the upper layer ,the encoder first specifies how many `` 1 ' 's are in this block .the coder then specifies the position of the first `` 1 '' the @xmath35 block in a forward horizontal scanning order ( i.e. row by row , from left to right ) using 4 bits . for the remaining `` 1 ''s in this block , the proposed coder forms a run - length representation , which is binarized using truncated unary code and entropy coded using a bitstream using the normal mode .this coding scheme is developed based on the observation that the first stage chosen atom locations are usually sparse and separated .a chosen atom in a particular location often indicates that its nearby atoms are unlikely to be chosen .therefore , in each @xmath35 block , there are usually only one atom chosen .we code the orders of the chosen atoms after encoding their locations . because we can deduce the total number of chosen atoms , i.e. @xmath18 , from the decoded significance map , we adaptively determine the total bits needed to represent the order values .for example , if 3 atoms are chosen , we will form a sequence of 3 values , each represented by 2 bits .after binarization , a bit - plane coding method is employed to code each bit plane of the @xmath18 order values .the proposed entropy coder encodes the run - lengths of `` 1 ' 's in each bit - plane and different bit - planes are encoded separately .furthermore , given the non - repeatability of the order values , the number of `` 1 ' 's is deterministic in each bit - plane .we make use of this property to further reduce the bits spent for coding .coefficient levels are encoded in a reversed chosen order , which has been specified in the previous step .we split the levels into the absolute values and signs , and code them separately . for the signs ,the bypass mode is always used . for absolute values ,predictive coding is used .the last chosen atom s level is predictively encoded first . for remaining levels but the first chosen atom , the coder always uses the previously encoded level as prediction .the first chosen atom s level however , is directly binarized by truncated unary and exponential golomb code .this is because we have observed a different distribution for the first chosen atoms .finally all binarized absolute values are coded using the normal mode and written to the bitstream .we show the coding performance of the proposed coder and other four comparison coders including the one - stage only baseline coder in @xcite for two test sequences , _ football _ , and _city_. the _ football _ sequence has a frame size @xmath36 and the _ city _ sequence has a frame size @xmath37 . in the proposed two - stage framework, there are a total of three parameters controlling the final rate .two are the quantization stepsizes for first stage and second stage , @xmath5 and @xmath23 , respectively ; and the third is the residual norm reduction ratio threshold @xmath11 for switching from the first stage . at present time , we used an exhaustive search to determine the optimal @xmath5 , @xmath23 , and @xmath11 for each target rate in each sequence .we have found that for a given @xmath23 , optimal @xmath5 is usually same or slightly smaller than @xmath23 , and @xmath11 increases with @xmath5 . for both the one - stage and two - stage coders , we use the eomp proposed in @xcite to derive the first stage atom coefficients . for hevc, we restrict the prediction and transform unit up to @xmath20 , but allow the smaller prediction and transform sizes . for h.264 , we compare both a version with cabac and all other advanced coding tools and subpartitions enabled ( referred as h264cabac ) , as well as the baseline cavlc with only @xmath20 block size , denoted by h264cavlc .figure [ fig : rd ] shows the rate - distortion curves of encoding one p - frame ( based on a previous frame coded as an intra - frame using h.264 ) for proposed and comparison methods . for both sequences ,the two - stage coder significantly improved upon the one - stage coder .it also outperforms both h.264 options significantly .more importantly , it provides substantial gain over hevc in the low rate range , although it fails to compete with hevc in the high rate region .we believe one reason our two - stage coder becomes less efficient than hevc at higher rates is because our second stage coding using a fixed transform size , whereas hevc allows variable transform sizes as well as transform skip .figure [ fig : rd10frm ] shows the rate - distortion curves for encoding ten p - frames , using ippp structure .the rate and psnr are averaged from all ten p - frames .our proposed coder can still outperform both h.264 options , and is competitive with hevc in the low to intermediate rate range ( in fact the proposed coder is up to 0.5db better for _ football _ at low rate range ) .the significant gain that was achieved for coding the first p - frame at low rates vanished unfortunately .we suspect that this is because our coder has not implemented in - loop filters such as deblocking filter and sample adaptive offset filter .those filters are known to mitigate the error propagation effect . because our first stage dictionary consists of blocks from the previously decoded frame ,its representation power reduces when the previous frame has large quantization error .effective in - loop filtering is likely to suppress this efficiency loss .besides , our proposed coder only uses inter - prediction candidates in the self - adaptive first - stage dictionary .we speculate the absence of intra - modes may also contribute to the loss of efficiency in coding subsequent frames .in this work we propose a two - stage video encoder framework , which uses a self - adaptive redundant dictionary in the first stage and uses dct bases ( after orthonormalization with the first stage chosen atoms ) in the second stage .we use a residual norm reduction ratio threshold to switch from the first to the second stage .we further propose a complete context adaptive binary entropy coding method to efficiently code the location , order and the coefficient values of the chosen atoms of the first stage .the location and values of non - zero coefficients in the second stage are coded following the same transform coding method of hevc for coding the prediction residual , but using a fixed transform size of @xmath20 .the two - stage coder achieved around 1db psnr improvement over our previous one - stage approach in @xcite .its performance is generally competitive with hevc , with slightly better performance at low rates , and worse performance at very high rates .note that our present coder uses a fixed block size of @xmath20 , has only inter - prediction candidates in the self - adaptive dictionary , and has not implemented in - loop filtering .the fact that it is competitive with hevc is quite promising , considering that hevc allows variable block size both for prediction and for coding of prediction residual , adpative switching between inter- and intra - prediction , and incorporates sophisticated in - loop filtering .there are many aspects of the present coder that can be improved .adaptive switching from the first stage to the second stage on a block - basis is likely to yield significant gain .incorporation of in - loop filtering and intra - prediction may also bring significant improvements .

<S> in this work , we propose a two - stage video coding framework , as an extension of our previous one - stage framework in @xcite . </S><S> the two - stage frameworks consists two different dictionaries . </S><S> specifically , the first stage directly finds the sparse representation of a block with a self - adaptive dictionary consisting of all possible inter - prediction candidates by solving an l0-norm minimization problem using an improved orthogonal matching pursuit with embedded orthonormalization ( eomp ) algorithm , and the second stage codes the residual using dct dictionary adaptively orthonormalized to the subspace spanned by the first stage atoms . the transition of the first stage and the second stage is determined based on both stages quantization stepsizes and a threshold . </S><S> we further propose a complete context adaptive entropy coder to efficiently code the locations and the coefficients of chosen first stage atoms . </S><S> simulation results show that the proposed coder significantly improves the rd performance over our previous one - stage coder . </S><S> more importantly , the two - stage coder , using a fixed block size and inter - prediction only , outperforms the h.264 coder ( x264 ) and is competitive with the hevc reference coder ( hm ) over a large rate range . </S>

1503.02779

hamming space @xmath18 of binary @xmath0-strings , equipped with the hamming distance is one of the classical objects studied in combinatorics .its properties that received significant attention are the maximal packing densities , covering numbers , isoperimetric inequalities , list - decoding properties , etc . in this paperwe are interested in studying metric properties of maps @xmath19 between hamming spaces of different dimensions .indeed , frequently one is interested in embedding @xmath18 into @xmath20 `` expansively '' , i.e. so that points that were far apart in @xmath18 remain far apart in @xmath20 .two immediate examples of such maps are : the error - correcting codes with rate @xmath21 and minimum distance @xmath9 satisfy @xmath22 where here and below @xmath23 is the hamming weight of the vector .another example is the repetition coding with @xmath24 mapping @xmath25 into @xmath26 repetitions of @xmath25 .this map satisfies : @xmath27 with these two examples in mind , we introduce the main concept of this paper .[ def : ab ] a map @xmath28 is called an @xmath29-map ( or simply an @xmath2-map ) if @xmath3 and @xmath4 are integers and for all @xmath30 we have either @xmath31 where @xmath18 is the hamming space of dimension @xmath0 over the binary field .we next define the hamming graphs @xmath7 for integer @xmath32 $ ] as follows : @xmath33 by @xmath34 and @xmath35 we denote the vertices of @xmath36 , the edges of @xmath36 and the cardinality of the maximal independent set of @xmath36 .all graphs in this paper are simple ( without self - loops and multiple edges ) . by @xmath37we denote the ( simple ) graph obtained by complementing @xmath38 and deleting self - loops .the relevance of hamming graphs to this paper comes from the simple observation : @xmath39 where @xmath40 denotes the existence of a graph homomorphism ( see section [ sec : hom ] for definition ) .this paper focuses on proving negative results showing impossibility of certain parameters @xmath41 .note that there are a variety of methods that we can use to disprove existence of graph homomorphisms .for example , by computing the shortest odd cycle we can prove @xmath42 in this paper , however , we are interested in the methods that provide some useful information in the asymptotic regime of @xmath43 , @xmath44 and fixed @xmath2 .our original motivation for definition [ def : ab ] was the following .suppose the map @xmath19 is used to protect the @xmath0 data bits against noise .if the points @xmath45 are far apart but @xmath24 and @xmath46 are close , i.e. if a map fails to satisfy ( [ eq : def_ab ] ) , then @xmath24 may be confused with @xmath46 in a noisy environment . consequently , this would lead to a severe discrepancy if @xmath47 is reported instead of @xmath25 .below we briefly discuss how @xmath2-property relates to some previously studied concepts .first , a @xmath48-map is simply an error - correcting code of rate @xmath21 and minimum distance @xmath49 .thus , @xmath2-condition is a _ relaxation of the minimum - distance property : _ the separation of @xmath49 is only guaranteed for data vectors @xmath50 that were @xmath51 apart to start with .practically , data may have some structure guaranteeing some separation between feasible data - vectors ( e.g. if @xmath25 is english test , changing one letter is unlikely to result in a grammatically correct phrase ) .second , in the inverse problem of reconstructing @xmath25 from a noisy version @xmath52 , one may proceed by computing a pre - image of the hamming ball of radius @xmath53 around @xmath54 . then the @xmath2-condition guarantees that the points in the pre - image will all be close to each other .third , an @xmath2-map can be used to convert a code with ( normalized ) minimal distance @xmath55 to a code of minimal distance @xmath56 at the expense of losing a factor @xmath57 in rate .this observation leads , on one hand , to a non - trivial bound on achievable parameters @xmath2 , see ( [ eq : ccb ] ) below . on the other hand, it also suggests that @xmath2-maps could be employed for adapting properties of a fixed mother code to the changing noise environment .fourth , an @xmath2-map with @xmath58 can be seen as a type of hashing in which one wants the hashes of dissimilar strings to be also dissimilar .in fact , the @xmath2-condition is weakening of the locality - sensitive hashing condition @xcite .finally , relaxation of the minimum - distance property taken in definition [ def : ab ] may be motivated by availability of the redundancy in the @xmath0-bit data . in information theory transmitting such data across a noisy channel is known as the joint source - channel coding ( jscc ) problem .combinatorial variation , cf .@xcite , can be stated as follows : say that @xmath28 is a @xmath59-jscc if there exists a decoder map @xmath60 with the property@xmath61 the operational meaning is that a @xmath59-jscc reduces the ( adversarial ) noise of strength @xmath62 in @xmath1-space to ( adversarial ) noise of strength @xmath63 in @xmath0-space .a special case of @xmath64 was introduced by spielman @xcite under the name of _ error - reducing _ codes .the connection to def .[ def : ab ] comes from the simple observation : @xmath65 thus , every impossibility result for @xmath2-maps implies impossibility results for @xmath59-jscc and spielman s error - reducing codes .for @xmath5 the best known bound to date is due to mceliece et al @xcite .it says that any set @xmath66 with @xmath67 for all @xmath68 satisfies @xmath69 where @xmath70 for @xmath71 and for @xmath72 : @xmath73 where @xmath74 and the minimum is taken over all @xmath75 satisfying @xmath76 for distances @xmath77 the solution is given by @xmath78 and @xmath79 has a simpler expression : @xmath80 thus from ( [ eq : mrrw ] ) we get @xmath81 a natural question is whether going from @xmath5 to @xmath82 may enable larger rates @xmath83 . the first impulse could be that the answer is negative .indeed , note that for @xmath84 there is @xmath85 points @xmath47 s.t .thus it may seem that for @xmath84 this relaxation yields no improvements ( asymptotically ) compared to @xmath5 .this observation is incorrect for two reasons .first , we do not require @xmath87 to be injective thus although all points @xmath46 are far from @xmath24 , they may not all be distinct .second , even though each @xmath25 has many @xmath47 satisfying @xmath86 , we in fact need a collection @xmath88 s.t .@xmath86 for all _ pairs _ @xmath89 .only then we may conclude that @xmath90 is code in @xmath20 with large minimal distance .thus , @xmath91 needs to be an independent set in @xmath92 .how large can it be ? to that end , we recall turan s theorem , cf .* theorem iv.6 ) : @xmath93 counting the number of edges of the graph @xmath92 via stirling s formula we get @xmath94 .therefore , @xmath95 consequently , if an @xmath41-map exists then comparing ( [ eq : mrrw ] ) and ( [ eq : gv1 ] ) we get @xmath96 one natural way to improve the bound would be to notice that graphs @xmath97 have a lot of extra structure and perhaps simplistic estimate ( [ eq : gv1 ] ) via turan s theorem can be improved . unfortunately , despite decades of work the lower bound ( [ eq : gv1 ] ) , known as the gilbert - varshamov bound , is asymptotically the best known .( for non - binary alphabets , however , better bounds exist @xcite . ) instead , the next theorem improves ( [ eq : ccb ] ) by establishing how another graph - function ( the @xmath12-function , see ( [ eq : ts_max ] ) below ) behaves under graph homomorphisms , and then applying known results on @xmath12-function for hamming graphs established by samorodnitsky @xcite and mceliece et al .@xcite .[ th : main1 ] for every @xmath98 there exist a sequence @xmath99 s.t .if an @xmath100-map exists with @xmath101 and @xmath102 then @xmath103 where @xmath104 for @xmath84 and zero otherwise . the bound ( [ eq : ccsam ] ) is better for @xmath105 , while ( [ eq : it ] ) is better for @xmath106 .see section [ sec : disc ] for evaluations .note that by virtue of relying only on the number of edges in @xmath107 the bound in ( [ eq : ccb ] ) is robust in the sense that whenever @xmath41 violate ( [ eq : ccb ] ) , there will be great many pairs of @xmath50 that violate ( [ eq : def_ab ] ) . here is a similar strengthening of theorem [ th : main1 ] .[ th : main2 ] for every @xmath98 there exist a sequence @xmath99 with the following property . for every map @xmath28 , every @xmath88 of size @xmath108 andevery @xmath109 $ ] satisfying @xmath110 there exists a pair @xmath89 such that @xmath111 in particular , there are at least @xmath112 un - ordered pairs @xmath113 satisfying ( [ eq : tm2_2a ] ) .next we consider an improved bound for the case of @xmath13 .notice that by plotkin bound ( * ? ? ?* chapter 2.2 ) we have @xmath114 in particular , @xmath115 does not contain @xmath116 whenever @xmath117 .therefore , any graph @xmath36 which contains @xmath116 can not map into @xmath115 .for example : @xmath118 the following elaborates on this idea : [ th : main3 ] for every @xmath119 there exists @xmath120 such that if there exists an @xmath121 map with @xmath122 and @xmath123 $ ] then @xmath124 furthermore , for any map @xmath125 , any @xmath126 and @xmath123 $ ] and any set @xmath88 of size @xmath127 there exists a pair of points @xmath89 satisfying ( [ eq : tm2_2a ] ) . considering the argument preceding the theorem , it should not be so surprising that the relation between @xmath128 and @xmath129 in ( [ eq : tm3 ] ) is independent of the rate @xmath130 .the significance of ( [ eq : tm3 ] ) is that for the case of @xmath131 this bound is ( asymptotically ) optimal , as the example of the repetition map ( [ eq : rep ] ) clearly shows . for linear @xmath2-mapsthe result was shown in ( * ? ? ?* theorem 8) by studying properties of the generator matrix . when applied to linear maps @xmath132theorems [ th : main2 ] and [ th : main3 ] have the following geometric interpretations : [ th : cor2 ] for every @xmath98 there exists a sequence @xmath133 with the following property . fix any two lists of ( possibly repeated ) points @xmath134 and @xmath135 in projective space @xmath136 s.t . that they are not all contained in a codimension 1 hyperplane .fix any @xmath137 $ ] s.t .@xmath138 or @xmath139 there exists a hyperplane @xmath140 of codimension 1 in @xmath141 such that @xmath142 [ th : cor3 ] for every @xmath119 there exists @xmath143 with the following property .fix any two lists of ( possibly repeated ) points @xmath134 and @xmath135 in projective space @xmath136 s.t . that they are not all contained in a codimension 1 hyperplane .fix any @xmath126 and @xmath144 $ ] s.t .@xmath145 then there exists a hyperplane @xmath140 of codimension 1 in @xmath141 satisfying ( [ eq : ab_hlin ] ) .note that by identifying homogeneous coordinates with affine coordinates we can establish set - isomorphism @xmath136 and @xmath146 .thus , previous corollaries can be equivalently restated in terms of @xmath147 .for example , for any @xmath98 , all @xmath0 sufficiently large and all @xmath1 : _ fix some basis of @xmath18 and arbitrary non - zero points @xmath148 .then there exists a @xmath149-subspace containing @xmath150 @xmath151-points and @xmath152 basis vectors . _ note that this is a manifestly @xmath17-property since over large fields one could select @xmath151-points ( when @xmath153 ) so that no @xmath154 of them are contained in a @xmath149-subspace . the rest of the paper is organized as follows : section [ sec : hom ] proves a few results on graph homomorphisms . in section[ sec : hamming ] these results are applied to prove theorems [ th : main1]-[th : main3 ] and corollaries [ th : cor2]-[th : cor3 ] .we conclude in section [ sec : disc ] with discussion , numerical evaluations and some open problems .let us introduce notation to be used in the remainder of the paper : @xmath155 where @xmath156 is a positive - semidefinite matrix of order @xmath157 , @xmath158 is an all - one matrix of the same size , @xmath159 denotes the maximal eigenvalue and @xmath160 denotes a subset @xmath161 of the entries of matrix @xmath156 , so that @xmath162 . @xmath163 and @xmath164are the schrijver and lovsz @xmath12-functions , respectively for @xmath164 and any of @xmath165 , @xmath166 or @xmath167 for @xmath163 . ] .we recall a few properties of the @xmath12-function ( one may consult @xcite for more ) : * both @xmath12-functions are typically used to upper bound the independence number of a graph : @xmath168 * @xmath169 , while yielding a looser bound on @xmath170 , is multiplicative under strong product of graphs is a simple graph with vertex set given by @xmath171 and edges @xmath172 . ] as shown in @xcite : @xmath173 * for a vertex transitive graphs , we also have reciprocity @xcite : @xmath174 our main technical contribution in this section is the following partial generalization of ( [ eq : lova_1])- ( [ eq : lova_2 ] ) to @xmath175 : [ th : sprod ] let @xmath36 be vertex transitive , then @xmath176 proof is given at the end of this section . we next discuss its application to existence of graph homomorphisms . the graph homomorphism @xmath177 is a map of vertices of @xmath178 to vertices of @xmath179 such that endpoints of each edge of @xmath178 map to the endpoints of some edge in @xmath179 .if there exists any graph homomorphism between @xmath178 and @xmath179 we will write @xmath180 .the problem of finding @xmath181 is known as @xmath179-coloring problem . for establishing properties of graph homomorphismsit is convenient to introduce _ homomorphic product _@xcite which corresponds to @xmath182 .] : graph @xmath183 is a simple graph with vertices @xmath184 and @xmath185 if @xmath186 or @xmath187 .from ( [ eq : lova_0 ] ) and definition of @xmath183 we have : @xmath188 and @xmath189 we overview some of the well - known tools for proving @xmath190 : * ( no - homomorphism lemma @xcite ) if @xmath180 and @xmath179 is vertex transitive then @xmath191 * ( monotonicity of @xmath192 ) if @xmath180 then @xmath193 * ( monotonicity of @xmath194 ) if @xmath180 then @xmath195 * ( homomorphic product ) if @xmath180 then @xmath196 note that ( [ eq : cdp1])- ( [ eq : cdp4 ] ) give necessary conditions for @xmath180 .although , generally not tight , these conditions can be understood as elegant relaxations ( semi - definite , fractional , quantum etc ) of the graph homomorphism problem , cf .@xcite .inequalities ( [ eq : nohom])- ( [ eq : cdp4 ] ) are useful for showing @xmath190 . if @xmath190 it is natural to ask for a quantity measuring to what extent @xmath178 fails to homomorphically map into @xmath179 .one such quantity is @xmath197 , since @xmath198 indeed , by construction any independent set @xmath91 in @xmath183 has at most one point in each fiber @xmath199 and thus projection @xmath200 onto @xmath178 always yields an induced subgraph @xmath201 satisfying @xmath202 . with ( [ eq : alpha_int ] ) in mind , the next set of results will allow us to assess the degree of failure of @xmath190. if @xmath178 is vertex transitive , then @xmath203 if @xmath179 is vertex transitive , then @xmath204 one may view ( [ eq : alpha_b1])- ( [ eq : thetal_b1 ] ) as a quantitative version of criteria ( [ eq : cdp0])- ( [ eq : cdp2 ] ) and ( [ eq : alpha_b2])- ( [ eq : thetal_b2 ] ) as a quantitative version of no - homomorphism lemma ( [ eq : nohom ] ) . note also that the right - most versions of ( [ eq : thetal_b1 ] ) and ( [ eq : thetal_b2 ] ) hold without any transitivity assumptions ( * ? ? ?* theorem 17 ) : for any @xmath205 @xmath206 the proof relies on the following simple observation : the strong product @xmath207 of @xmath178 and @xmath208 is a subgraph of @xmath209 . thus by edge - monotonicity : @xmath210 from here the results on @xmath211 and @xmath175 follow from lemma [ th : sprod ] with @xmath212 , @xmath213 ( for ( [ eq : alpha_b1 ] ) and ( [ eq : thetas_b1 ] ) ) or @xmath214 and @xmath215 ( for ( [ eq : alpha_b2 ] ) and ( [ eq : thetas_b2 ] ) ) .for @xmath216 the equality parts of ( [ eq : thetal_b1 ] ) and ( [ eq : thetal_b2 ] ) follow from the results of lovsz ( [ eq : lova_1 ] ) and ( [ eq : lova_2 ] ) .one of the classically useful methods in coding theory is the elias - bassalygo reduction : from a given code in @xmath20 one selects a large subcode sitting on a hamming sphere of a given radius .one then bounds minimum distance ( or other ) parameters for the packing problem in the johnson graph @xmath217 .it so happens that taking a simple dual certificate for @xmath218 and transporting the bound back to the full space results in excellent bounds , which are hard ( but possible see rodemich theorem in @xcite ) to obtain by direct sdp methods in the full space .succinctly , we may summarize this as follows : if @xmath219 is an induced subgraph of a vertex transitive @xmath36 then @xmath220 here is a version of the similar method for the graph - homomorphism problem and for the problem of finding independent sets in @xmath221 : [ th : elbas_prod ] let @xmath36 be a vertex transitive graph and @xmath219 its induced subgraph. then @xmath222 and same for @xmath216 .let @xmath223 be the group of automorphisms of @xmath36 .the action of @xmath223 naturally extends to the action on @xmath221 via : @xmath224 let @xmath91 be the maximal independent set of @xmath221 .consider the chain : @xmath225 where ( [ eq : eb1 ] ) follows since each @xmath226 is an independent set of @xmath227 , ( [ eq : eb2 ] ) is obvious , and ( [ eq : eb3 ] ) is because by the transitivity of the action of @xmath223 : @xmath228 .clearly , ( [ eq : eb3 ] ) is equivalent to ( [ eq : elbas_alpha ] ) . for ( [ eq : elbas_theta ] ) let @xmath229 be the maximizer in ( [ eq : ts_max ] ) .symmetrizing over @xmath223 if necessary we may assume that @xmath230 last equation also implies that the subspace spanned by vectors @xmath231 is an eigenspace of @xmath156 . here and below @xmath232 are all - one vectors of dimensions @xmath157 and @xmath233 respectively .and @xmath234 is a zero / one vector of dimension @xmath157 having ones in coordinates corresponding to vertices in @xmath219 .set @xmath235 one easily verifies that @xmath236 is a feasible choice for the primal program ( [ eq : ts_max ] ) for @xmath237 . to compute @xmath238 we notice that @xmath239 where @xmath240 is a standard inner product on @xmath241finally , observe that orthogonal decomposition @xmath242 remains orthogonal after application of @xmath156 , cf .( [ eq : ebt_1 ] ) .therefore , we get by positivity @xmath243 that @xmath244 which together with ( [ eq : ebt_2 ] ) completes the proof of ( [ eq : elbas_theta ] ) .[ th : elbas ] let @xmath245 and @xmath246 be induced subgraphs of @xmath178 and @xmath179 , respectively .if @xmath178 is vertex transitive then @xmath247 if @xmath179 is vertex transitive then @xmath248 we will give an explicit proof by exhibiting a choice of @xmath249 in ( [ eq : ts_min ] ) for computing @xmath250 .let @xmath156 be the optimal ( primal ) solution of ( [ eq : ts_max ] ) for @xmath251 and let @xmath252 be the optimal ( dual ) solution of ( [ eq : ts_min ] ) for @xmath253 .we know : @xmath254 and also from the vertex - transitivity of @xmath36 without loss of generalitywe may assume that @xmath255 where @xmath256 is an all - one vector .we now define and assuming @xmath257 we could write ( [ eq : ts_q3 ] ) as @xmath258 , which is more natural . ]@xmath259 where as before @xmath158 denotes the square matrix of all ones ( of different dimension depending on context ) and @xmath260 we will prove that @xmath261 is a feasible choice in the ( dual ) problem ( [ eq : ts_min ] ) for @xmath250 . then we can conclude that since @xmath243 and @xmath262 that @xmath263 since by construction @xmath264 ( recall that @xmath156 and @xmath158 commute ) .thus , @xmath265 proving ( [ eq : thetas_b1 ] ) . to verify that @xmath261 is feasible dual assignment , we need to show @xmath266 which follows since @xmath267 consists of all self - loops and edges connecting pairs that are non - adjacent ( and non - identical ) in either @xmath36 or @xmath140-coordinate . to verify ( [ eq : ts_q2 ] ) we recall that @xmath156 and @xmath252 satisfy @xmath268 then verification proceeds in a straightforward manner .for example , in the first case in ( [ eq : ts_q2 ] ) we have @xmath269 because @xmath270 and by ( [ eq : ts_q2a ] ) .the two remaining cases are checked similarly .before going into details of the proofs , we make a clarifying remark . our main goal is to improve the simple bound ( [ eq : ccb ] ) , which ( we remind ) was obtained by noticing that independent sets of @xmath271 transform under @xmath2-maps into independent sets of @xmath272 .the improvement comes by noticing that @xmath2-maps also transform any matrix @xmath156 in ( [ eq : ts_max ] ) feasible for @xmath271 into a matrix @xmath273 feasible for @xmath272 .good feasible matrices for @xmath271 were found previously in @xcite .we proceed to formal details . clearly , it is sufficient to prove theorem [ th : main2 ] .we quote the following results of kleitman @xcite , mceliece et al @xcite and the joint lower bound of samorodnitsky @xcite and navon - samorodnitsky @xcite : @xmath274 where the remainder term @xmath275 uniformly on compacts in @xmath276 $ ] .take @xmath277 and @xmath278 .define @xmath279}\delta_m(\lambda)$ ] .assume that ( [ eq : tm2_1 ] ) holds .consider arbitrary @xmath19 .notice that if @xmath91 is a set which does not contain any pair satisfying ( [ eq : tm2_2a ] ) , then the set @xmath280 is an independent set of @xmath281 .( this is also clear from ( [ eq : alpha_int ] ) as @xmath87 defines a homomorphism @xmath282 if @xmath91 is viewed as induced subgraph of @xmath107 . )thus it is sufficient to show @xmath283 this follows from the following chain : @xmath284 where ( [ eq : tm22 ] ) is from ( [ eq : thetas_b1 ] ) , ( [ eq : tm23 ] ) is from ( [ eq : tmrrw ] ) and ( [ eq : tsam ] ) and ( [ eq : tm24 ] ) is from ( [ eq : tm2_1 ] ) .if instead of ( [ eq : tm2_1 ] ) the pair @xmath2 satisfies ( [ eq : tm2_2 ] ) then the argument is the same except in ( [ eq : tm22 ] ) we should apply ( [ eq : alpha_b2 ] ) and ( [ eq : tkl ] ) to get : we have@xmath285 therefore here we could still operate with @xmath175 only and apply ( [ eq : thetas_b2 ] ) .we chose to use @xmath211 s because kleitman s theorem ( [ eq : tkl ] ) has explicit non - asymptotic form and thus for finite @xmath286 results in a better bound . ] @xmath287 and the rest of the proof is the same .finally , to show the statement about the number of pairs satisfying ( [ eq : tm2_2a ] ) define a graph @xmath36 with vertices @xmath18 and @xmath288 if ( [ eq : tm2_2a ] ) holds .we have already shown @xmath289 then from turan s theorem we have @xmath290 ( after enlarging @xmath291 slightly ) .the argument follows step by step the proof of theorem [ th : main3 ] except that at ( [ eq : tm22 ] ) we use the ( almost ) exact value of @xmath292 for @xmath293 found in the lemma below .[ th : plot ] for any @xmath294 there exists @xmath295 s.t .@xmath296 furthermore , @xmath297 uniformly on compacts of @xmath298 .we need to introduce the standard definitions from linear programming bounds in coding theory , cf .any polynomial @xmath299 $ ] of degree @xmath300 can be represented as @xmath301 where krawtchouk polynomials are defined as @xmath302 and , for example , @xmath303 , @xmath304 .it is a standard result ( * ? ? ?* theorem 3 ) that for the hamming graphs the semidefinite program ( [ eq : ts_max ] ) becomes a linear program .we put it here in the following form : @xmath305\cap\mathbb{z } , f(x ) \ge 0 , x\in[0,n]\cap \mathbb{z}\right\ } \label{eq : ts_ham1}\\ & = \min\left\ { 2^n { g(0)\over \hat g(0 ) } : \hat g \ge 0 , g(x)\le 0 , x \in[d+1 , n]\cap \mathbb{z } , \hat g(0)>0 \right\ } \label{eq : ts_ham2}\end{aligned}\ ] ] where @xmath87 and @xmath306 are polynomials of degree at most @xmath1 .the upper bound ( [ eq : p2 ] ) is a standard plotkin bound ( see ( * ? ? ?4 ) ) : taking @xmath307 we notice that @xmath308 thus @xmath309 and we get for @xmath310 @xmath311 for the lower bound ( [ eq : p2 ] ) we assume that @xmath310 , @xmath312\ ] ] and @xmath1 is sufficiently large ( for all small @xmath1 we may take @xmath313 ) .we first consider the case of @xmath9 odd .consider the polynomial @xmath314 with coefficients given by @xmath315 where @xmath316 is to be determined . to compute values of this polynomial ,we employ the orthogonality relation for krawtchouk polynomials , cf .? * ( 34 ) ) : @xmath317\cap \mathbb{z}\ ] ] from here we get @xmath318\cap \mathbb{z}$}\ , . \end{cases}\ ] ] we note that this @xmath24 was guessed by studying levenshtein s codes that attain plotkin bound ( * ? ? ?* chapter 2 , theorem 8) . to verify that @xmath24 is a ( asymptotically ! ) feasible solution of ( [ eq : ts_ham1 ] ) we need to check @xmath319 .first , let @xmath320 and notice that ( * ? ? ?* ( 31)-(32 ) ) @xmath321 therefore , since @xmath322 is even it is sufficient to verify @xmath323 for all @xmath324\cap\mathbb{z}$ ] . for @xmath325this is obvious , for @xmath326 we have ( * ? ? ?* ( 13 ) ) @xmath327 and thus taking @xmath328 makes ( [ eq : p3 ] ) hold at @xmath329 .it is known that @xmath330 has @xmath331 real roots with the smallest root @xmath332 satisfying ( * ? ? ? * ( 71 ) ) @xmath333 therefore , polynomial @xmath330 is decreasing on @xmath334 $ ] and hence ( [ eq : p3 ] ) must also hold for all odd @xmath335 $ ] ( for even @xmath336 , inequality ( [ eq : p3 ] ) holds just by considering the signs ) . in view of ( [ eq : p4 ] )we only need to show ( [ eq : p3 ] ) for @xmath337 , where @xmath338 in this range , we will show a stronger bound @xmath339 the following bound is well known ( * ? ? ?* ( 87 ) ) in ( [ eq : p1 ] ) , we could use the non - asymptotic bound in ( * ? ? ?* lemma 4 ) , which also holds for @xmath340 .] : @xmath341 note that by the constraint ( [ eq : p6 ] ) @xmath342 is bounded away from @xmath343 and thus we can estimate @xmath344 for some sequence @xmath345 that only depends on @xmath346 . thus comparing the exponents on both sides of ( [ eq : p3 ] ) we see that it will hold provided that @xmath347 but notice that by ( [ eq : p5 ] ) the exponent in parenthesis is exactly the gap between the gilbert - varshamov bound @xmath348 and the first linear programming bound @xmath349 , cf .( [ eq : rlp1 ] ) .there exists @xmath350 separating these two bounds for all @xmath351 s in ( [ eq : p6 ] ) .thus , the right - hand side of ( [ eq : p7 ] ) is exponentially decreasing @xmath352 and hence for sufficiently large @xmath1 it must hold .this completes the proof that @xmath24 in ( [ eq : p8 ] ) is a feasible choice in ( [ eq : ts_ham1 ] ) .therefore , we have shown that for all @xmath1 sufficiently large @xmath353 if @xmath9 is odd and @xmath354 if @xmath9 is even .assume to the contrary that one found @xmath355 and @xmath356 s.t .there is no hyperplane satisfying ( [ eq : ab_hlin ] ) .then as explained in section [ sec : lincode ] below ( see ( [ eq : ab_linx ] ) ) , there is an independent set of size @xmath357 in @xmath358 . by inspecting the proofs of theorem [ th : main2 ] and[ th : main3 ] we notice that they prove three different upper bounds on @xmath359 that are equal to exponentiation of the left - hand sides of ( [ eq : cora ] ) , ( [ eq : corb ] ) and ( [ eq : corc ] ) respectively .therefore , @xmath331 can not satisfy any of ( [ eq : cora ] ) , ( [ eq : corb ] ) or ( [ eq : corc ] ) a contradiction .in this section we evaluate our bounds . we consider the asymptotic setting @xmath43 and @xmath360 where @xmath361 is fixed . in fig .[ fig : eval1 ] ( @xmath362 ) and fig .[ fig : eval2 ] ( @xmath363 ) we plot the various bounds on the region of asymptotically feasible pairs @xmath2 : * for @xmath362 the lower bound for @xmath364 is ( [ eq : ccsam ] ) from theorem [ th : main1 ] ; for @xmath365 is theorem [ th : main3 ] . * for @xmath363the lower bound ( for all @xmath128 ) is ( [ eq : it ] ) . in this casethe other two bounds , ( [ eq : ccsam ] ) and ( [ eq : tm3 ] ) , are strictly worse . * for @xmath362the straight dashed line denotes performance of the repetition map ( [ eq : rep ] ) . * for @xmath366the straight dashed line denotes performance of the _ majority - vote _ map .namely @xmath367 gives a majority vote for every one of @xmath368-bit blocks .it is clear that for all @xmath50 we have : @xmath369 indeed any pair of @xmath368-bit strings for which majority - vote agrees can be at most hamming distance @xmath370 away ( as @xmath371 and @xmath372 ) .* finally , the curved dashed line corresponds to the _ separation map _ defined as follows .fix @xmath373 and cover @xmath18 with balls of radius @xmath374 .it is sufficient to have @xmath375 such balls .also consider a packing of balls of radius @xmath53 in @xmath20 . by gilbert - varshamovbound we know that we can select at least @xmath376 such balls .thus whenever @xmath377 we can construct the map @xmath19 that maps every point inside an @xmath374 ball to a center of the corresponding packing ball .clearly , such map will be an @xmath2 map .thus , asymptotically all pairs of @xmath2 s.t .@xmath378 and @xmath379 are achievable . note that as @xmath380 the bound ( [ eq : ccsam ] ) becomes : @xmath381 this is a significant improvement over what the simple bound ( [ eq : ccb ] ) yields : @xmath382 in particular , ( [ eq : ccsam_x ] ) has finite slope at @xmath383 and furthermore as @xmath384 the slope - discontinuity at @xmath385 , see fig .[ fig : eval1 ] , disappears .this last effect is a consequence of the navon - samorodnitsky @xcite part of the bound ( [ eq : rsam_def ] ) .one of the more interesting conclusions that we can draw from our bounds is the following .it is well known that there is only finitely many balls of radius @xmath386 that can be packed inside @xmath20 without overlapping ( plotkin bound ) . however ,if one allows these balls to cover each point with multiplicity at most 3 then it is possible to pack exponentially many balls @xcite .thus , if one is allowed to decode into lists of size 3 , it is possible to withstand adversarial noise of weight @xmath387 while still having non - zero communication rate . by setting @xmath388 and applying theorem [ th : main3 ]we figure out , however , that no matter how the balls are labeled by @xmath0-bit strings , at least one ball of radius @xmath389 will contain a pair of points whose labels differ in at least @xmath390 positions .so although list - decoding allows one to overcome the @xmath391 barrier , there is no hope ( in the worst case ) to recover any information bits from the labels .this is only true beyond the radius @xmath391 , since of course , below @xmath391 one can use codes with list-1 . loosely speaking, we have a `` phase - transition '' in the communication problem at noise - level @xmath391 .it is possible to write a linear program for @xmath392 similarly to the standard delsarte s program ( [ eq : ts_ham1])- ( [ eq : ts_ham2 ] ) . to that end , for an arbitrary polynomial @xmath393 of degree at most @xmath0 in @xmath25 and at most @xmath1 in @xmath54 we can write it as @xmath394 where @xmath395 and @xmath396 are krawtchouk polynomials ( [ eq : krawt ] ) . with this definition of the fouriertransform @xmath397 we have : @xmath398 where @xmath399 are bi - variate polynomials of degree at most @xmath400 , and @xmath401\cap \mathbb{z } , y\in [ 0,n]\cap \mathbb{z } , ( x=0 , y\neq 0)\mbox{~or~}(x>\alpha k , y \le \beta n)\right\}\\ \mathcal{d}^c & \eqdef \left\{(x , y ) : x \in [ 0,k]\cap \mathbb{z } , y\in [ 0,n]\cap \mathbb{z } , ( 0<x\le \alpha k)\mbox{~or~}(x\neq 0 , y>\beta n ) \right\ } \end{aligned}\ ] ] the bound used in theorem [ th : main2 ] states @xmath402 this bound corresponds to the following choice of @xmath403 in ( [ eq : tsh_ham2 ] ) : @xmath404 where @xmath405 is the samorodnitsky assignment @xcite in the primal for @xmath406 and @xmath407 is a standard choice of mceliece - rodemich - rumsey - welch @xcite in the dual for @xmath408 . in fact , any primal @xmath409 and any dual @xmath410 give a candidate for @xmath403 .thus , we have @xmath411 optimizing over all @xmath409 and @xmath410 we get @xmath412 ( this , of course , is exactly how the bound was obtained in lemma [ th : sprod ] . ) we were not able to find any choice of @xmath403 in the dual problem ( [ eq : tsh_ham2 ] ) that beats the product @xmath413 .this seems to be the most natural direction for improvement .another open problem is to find an upper bound on @xmath414 that does not follow from an upper bound on @xmath415 or to prove that these are asymptotically equivalent . by looking at fig .[ fig : eval1 ] we can see that relaxation of the minimum - distance property , cf .[ def : ab ] , that we consider in this paper allows one to have non - zero rate even at `` minimum distance '' @xmath126 .however , in this case @xmath416 ( theorem [ th : main3 ] ) and furthermore repetition map ( [ eq : rep ] ) is optimal ( provided @xmath417 ) .this raises a number of questions : * can one show that any @xmath2-map in high-@xmath129 regime is structured similarly to a repetition map ?* for the case when @xmath418 ( e.g. @xmath419 ) , how do we asymptotically achieve @xmath420 ? * the corresponding situation with majority - vote maps is even worse , here we need @xmath130 be an odd integer .what is the counterpart for even @xmath130 ?* finally , how do we smoothly interpolate between the `` non - smooth '' separation construction ( that is not even injective ) and the repetition map ? in fact the last question was our main practical motivation for looking into the concept of @xmath2-maps .we do not have any good candidates at this point . a natural approach to construct good @xmath2 mapsis to restrict to linear maps @xmath19 .a linear @xmath87 is @xmath2 if @xmath421 instead of working with this condition , we get a more invariant notion by considering the graph of @xmath87 that is just a linear subspace of @xmath422 .different conditions can be stated on @xmath423 that will ensure that @xmath423 defines an @xmath2-map , or that it gives an independent set in @xmath424 , or even an independent set in @xmath425 .we state these conditions for a general field @xmath426 and also in a geometric language of @xcite .the extension of the concept of an @xmath2-map , cf . definition [ def : ab ] , and hamming graphs @xmath427 , cf .( [ eq : hnd_def ] ) , to arbitrary field @xmath426 is obvious .suppose that we have ( not necessarily distinct ) points @xmath428 where @xmath141 is a projective space of dimension @xmath429 over the field @xmath426 .for every codimension 1 hyperplane @xmath140 define @xmath430 by writing these two collections of points in homogeneous coordinates we get a @xmath431 matrix over @xmath426 , whose row - span gives the linear subspace @xmath432 .we then have the following statements : 1 .if points @xmath433 , j\in[n]\}$ ] are not contained in any ( codimension 1 ) hyperplane @xmath434 and satisfy @xmath435 then @xmath436 2 .if points @xmath437\}$ ] are not contained in any ( codimension 1 ) hyperplane @xmath438 and @xmath439 then @xmath440 ( note that assumption implies @xmath441 here ) .if in addition to previous assumption we also have @xmath442 , i.e. points @xmath437\}$ ] span @xmath443 , then there exists a linear @xmath2-map and thus @xmath444 note that if @xmath445 span @xmath443 and @xmath5 then condition ( [ eq : ab_linx2 ] ) simply states that one can not include more than @xmath446 points from @xmath447\}$ ] into any hyperplane i.e. a standard geometric condition for @xmath448_q$]-systems , cf .similarly to how @xmath448_q$ ] systems exactly correspond to @xmath448_q$ ] linear codes , existence of points @xmath449 satisfying ( [ eq : ab_linx ] ) and assumptions in item 3 is _ equivalent _ to existence of an @xmath426-linear @xmath121-map . as an example , consider @xmath450 .we will construct a linear map @xmath451 by selecting seven points on the binary projective plane @xmath452 : @xmath453 are any points spanning @xmath454 , @xmath455 and finally put @xmath456 to be the only point not contained in any of the lines @xmath457 .see fig .[ fig : fano2 ] for an illustration .it is easy to see that condition ( [ eq : ab_linx2 ] ) holds with @xmath458 and @xmath459 , thus @xmath460 computing ( [ eq : tsh_ham2 ] ) with smaller @xmath128 and larger @xmath129 shows that the code of fig . [fig : fano2 ] is optimal .-map by selecting seven points on the binary projective plane .this material is based upon work supported by the national science foundation under grant no ccf-13 - 18620 .author is grateful to the support of simons institute for the theory of computing ( uc berkeley ) , at which this work was finished .discussions with prof .a. mazumdar , a. samorodnitsky and participants of the program on information theory at the simons institute were helpful .t. cubitt , l. mancinska , d. roberson , s. severini , d. stahlke , a. winter , bounds on entanglement assisted source - channel coding via the lovsz @xmath461 number and its variants , ieee trans .theory ( 11 ) ( 2014 ) 73307344 .a. samorodnitsky , http://www.cs.huji.ac.il/~salex/papers/old_sq_measure.ps[extremal properties of solutions for delsarte s linear program ] , preprint .

<S> a mapping of @xmath0-bit strings into @xmath1-bit strings is called an @xmath2-map if @xmath0-bit strings which are more than @xmath3 apart are mapped to @xmath1-bit strings that are more than @xmath4 apart . </S><S> this is a relaxation of the classical problem of constructing error - correcting codes , which corresponds to @xmath5 . </S><S> existence of an @xmath2-map is equivalent to existence of a graph homomorphism @xmath6 , where @xmath7 is a hamming graph with vertex set @xmath8 and edges connecting vertices differing in @xmath9 or fewer entries . </S><S> this paper proves impossibility results on achievable parameters @xmath2 in the regime of @xmath10 with a fixed ratio @xmath11 . </S><S> this is done by developing a general criterion for existence of graph - homomorphism based on the semi - definite relaxation of the independence number of a graph ( known as the schrijver s @xmath12-function ) . </S><S> the criterion is then evaluated using some known and some new results from coding theory concerning the @xmath12-function of hamming graphs . as an example </S><S> , it is shown that if @xmath13 and @xmath14 integer , the @xmath15-fold repetition map achieving @xmath16 is asymptotically optimal . finally , constraints on configurations of points and hyperplanes in projective spaces over @xmath17 are derived . </S><S> error - correcting codes , graph homomorphism , schrijver s @xmath12-function , projective geometry over @xmath17 </S>

hep-ph9807358

parton distributions were first introduced by feynman as the phenomenological quantities describing the properties of the nucleon manifest in high - energy scattering@xcite .when a nucleon travels near the speed of light , it can be viewed as consisting of a beam of non - interacting partons that are massless and collinear .parton distributions are the densities of partons as a function of the fraction @xmath0 of the nucleon momentum they carry . with the advent of quantum chromodynamics ( qcd ) ,feynman s parton model , modulo logarithms , can be justified in a rigorous field theoretical approach @xcite .meanwhile , parton distributions can be expressed as the matrix elements of the bilocal light - ray operators in the ground state of the nucleon @xcite . only in the light - like gauge , @xmath1 , and light - cone quantization ,does their physical interpretation become apparent .when studying the physics of parton distributions themselves , one of course can no longer regard the nucleon as consisting of free partons .calculating these distributions requires a detailed knowledge of the nucleon wave function in quark and gluon degrees of freedom , and undoubtedly is one of the most challenging problems in nonperturbative qcd . at present , the only fundamental approach we know is the lattice qcd method @xcite . on the phenomenological side , however , the unpolarized parton distributions have been extracted with good precision from various high - energy scattering data taken in the last thirty years @xcite .the empirical parton distributions not only provide the necessary input for calculating high - energy scattering cross sections , e.g. , of top - quark production , but also contain valuable information about how the nucleon is made of its constituents . recently , there has been much theoretical activity surrounding the so - called off - forward parton ( or non - forward , off - diagonal ) distributions ( ofpd s ) .the new distributions generalize those introduced by feynman : they characterize certain properties of the nucleon exhibited in a class of high - energy scattering ; they reflect the low - energy internal structure of the particle . however , there are also important differences between the off - forward and forward distributions .the ofpd s in general can not be regarded as particle densities , but rather their physical interpretation is given in terms of probability amplitude .as we shall see , they do have simple physical significance in light - cone coordinates ( or the infinite momentum frame ) .the most natural appearance of the ofpd s is in _ non - forward _ high - energy processes . as far as we know ,such processes were first considered by watanabe @xcite and by bartels and m. loewe @xcite . in ref .@xcite , a general virtual - photon compton scattering off a nucleon target was studied in operator product expansion ( ope ) , an integral representation of the amplitude was obtained , and the off - forward matrix elements of twist - two operators were defined .the study , however , mentioned little about physical motivations and the practicality of the process ; in particular , significance of the integral representation was obscure . in ref .@xcite , diffractive production of @xmath2 and photons was studied in perturbative qcd and the properties of the non - diagonal gluon ladder were investigated . however , no notion of an off - forward distribution of the nucleon target was apparent in the paper . following @xcite , a set of off - forward evolution kernels were presented by gribov , levin , and ryskin @xcite .these kernels were considered as those for the scattering amplitude .the role of off - forward distributions in hard diffractive processes had remained obscured in relatively recent studies . in ref .@xcite , diffractive electroproduction of @xmath3 was analyzed by ryskin in the small @xmath0 region .the production amplitude was related to the usual feynman gluon distribution .the off - forwardness was taken into account by a phenomenological form factor . in refs .@xcite , ryskin s analysis was generalized to light vector - meson production in which the meson structure is accounted for by the leading - twist light - cone wave function .to the author s knowledge , one of first examples of ofpd s was introduced by dittes et al .@xcite , and was called `` interpolating function . ''ironically , the main motivation for introducing such an object was not from studying any off - forward process ; rather , it was considered as an example of the physical quantities whose evolution kernel interpolates between the two well - known cases : the dglap and erbl evolutions @xcite .a serious study of the ofpd in a physical process was made by d. mller et al .@xcite in which they showed that it contributes to the general two virtual - photon compton scattering the same process considered earlier by watanabe .they also derived the evolution equations for the ofpd .note that in ref .@xcite a general off - forward amplitude ( or single - particle green s function ) was introduced in studying hadron helicity flip in hard scattering . however , because the correlation of two quark fields in the object is off the light cone , it does not have a simple parton interpretation as the ofpd s do .the recent wave of interest in the ofpd s was generated from two new developments .the first development is the recognition of the importance of the ofpd s in describing the internal structure of the nucleon @xcite . in studying the spin structure, it was found that the fractions of the nucleon spin carried by quarks and gluons can be determined by the form factors of the corresponding energy - momentum tensors .these form factors can be obtained from the moments of the ofpd s which were independently introduced in @xcite . in the same paper , deeply virtual compton scattering ( dvcs )was proposed as a practical means to measure the ofpd s . in dvcs ,a highly virtual photon scatters off a quark in a nucleon target , which then emits a real photon and returns to form a recoil nucleon .although the two - photon process studied earlier by watanabe and mller et al .contains dvcs as its special kinematic limit , the latter has important experimental advantages and a special theoretical status . indeed , dvcs is more delicate to treat theoretically because of an extra light - like vector among the external momenta .the second development is the realization that hard diffractive electroproduction of vector mesons requires the use of ofpd s @xcite .the first serious studies in this direction were made by radyushkin @xcite , hoodbhoy @xcite , and collins et al .@xcite . in ref .@xcite , an off - forward gluon distribution - most important in the low @xmath0 region - was introduced and used to calculate vector meson production in deep - inelastic electron scattering .the result is free of the assumptions made in ref .@xcite about the dominance of the absorptive contribution and the use of the forward gluon distribution .hoodbhoy @xcite used the off - forward gluon distribution to calculate heavy quarkonium production investigated earlier by ryskin @xcite .he obtained a similar result as radyushkin s . in a comprehensive paper , collins , frankfurt , andstrikman showed that a factorization theorem exists for general hard diffractive meson production in deep - inelastic scattering @xcite .the factorization theorem explicitly involves the ofpd s .recently , many works have been produced in studying off - forward parton distributions , deeply virtual compton scattering , and diffractive vector meson production .it is the goal of this paper to review these interesting developments . in sectionii , off - forward distributions are introduced from different perspectives and various definitions in the literature are compared . in section iii , we make a digression to discuss the spin structure of the nucleon . the role of the ofpd s in this interesting subject is exposed . in section iv , we summarize our present knowledge about the ofpd s at hadron mass scales and their evolution to high - energy scales . in sectionv , we consider processes in which the ofpd s may actually be measured : deeply virtual compton scattering and hard diffractive vector - meson production .section vi contains the summary and outlook .since the off - forward distributions contain rich information about the nucleon structure , it is instructive to examine their definitions and interpretations from different angles . in this section, we will first introduce the off - forward distributions from the form factors of twist - two operators and then we explore their significance in light - cone coordinates .finally we compare the ofpd s with other equivalent definitions in the literature . from the point view of the low - energy nucleon structure , it is perhaps most interesting to consider the off - forward distributions as the generating functions for the form factors of the so - called twist - two operators .recall that the matrix elements of the electromagnetic current in the equal momentum states are entirely determined by symmetry , whereas those in the unequal momentum states define the ( dirac and pauli ) form factors which contain such interesting information as charge radius and magnetic moment of the nucleon .the following tower of twist - two operators represents a generalization of the electromagnetic current @xmath4 where all indices are symmetrized and traceless ( indicated by @xmath5 ) and @xmath6 technically , these operators transform like @xmath7 under lorentz transformations .our interest in such operators is not accidental .they appear , for instance , in the operator production expansion of the two electromangetic currents in deep - inelastic scattering @xcite .thus although these generalized currents do not couple directly to any known fundamental interactions , they can nonetheless be studied indirectly in hard scattering processes . since the operators for @xmath8 are not related to any symmetry in the qcd lagrangian, their matrix elements between the equal momentum states , @xmath9 contain valuable dynamical information about the internal structure of the nucleon . in this paper, we use the covariant normalization @xmath10 for the nucleon state .the @xmath11 dependence of the above matrix elements signifies the dependence on renormalization scale and scheme @xcite .the quark distribution @xmath12 introduced by feynman has a simple connection to the above matrix elements : @xmath13 where @xmath14 is chosen to have support in @xmath15 . for @xmath16 , @xmath12 is the just density of quarks which carry the @xmath0 fraction of the parent nucleon momentum .the density of antiquarks is customarily denoted as @xmath17 , which in the above notation is @xmath18 @xcite . just like the form factors of the electromagnetic current ,additional information about nucleon structure can be found in the form factors of the twist - two operators when the matrix elements are taken between the states of unequal momenta .using lorentz symmetry and parity and time reversal invariance , one can write down all possible form factors of the spin-@xmath19 operator @xmath20 } a_{qn,2i}(t ) \delta^{\mu_2}\cdots \delta^{\mu_{2i+1 } } \overline{p}^{\mu_{2i+2}}\cdots \overline{p}^{\mu_n ) } \nonumber \\ & & + ~ { \overline u}(p'){\sigma^{(\mu_1\alpha } i\delta_\alpha \over 2m}u(p ) \sum_{i=0}^{[{n-1\over 2 } ] } b_{qn,2i}(t ) \delta^{\mu_2}\cdots \delta^{\mu_{2i+1 } } \overline{p}^{\mu_{2i+2}}\cdots \overline{p}^{\mu_n ) } \nonumber \\ & & + ~ c_{qn}(t ) ~{\rm mod}(n+1,2)~{1\over m}\bar u(p ' ) u(p ) \delta^{(\mu_1 } \cdots \delta^{\mu_n ) } \ , \label{form}\end{aligned}\ ] ] where @xmath21 and @xmath22 are the dirac spinors and @xmath23 is 1 when @xmath19 even , 0 when @xmath19 odd .the four - momentum transfer is denoted by @xmath24 and its invariant by @xmath25 .the average nucleon momentum @xmath26 is also a useful variable . for @xmath27 ,even or odd , there are @xmath28 form factors .@xmath29 is present only when @xmath19 is even .as the forward matrix elements @xmath30 , the above form factors can be used to define a new type of parton distributions _ off - forward distributions_. to accomplish this , we introduce a light - like vector @xmath31 , which is conjugate to @xmath32 in the sense that @xmath33 .write @xmath34 , where @xmath35 and @xmath36 is another light - like vector .contracting both sides of eq .( [ form ] ) with @xmath37 , we have @xmath38 where @xmath39 } a_{qn,2i}(t ) ( -2\xi)^{2i } + { \rm mod}(n+1,2)~ c_{qn}(t ) ( -2\xi)^n \ , \nonumber \\ e_{qn } ( \xi , t ) & = & \sum_{i=0}^{[{n-1\over 2 } ] } b_{qn,2i}(t ) ( -2\xi)^{2i } - { \rm mod}(n+1,2)~ c_{qn}(t ) ( -2\xi)^n \ . \end{aligned}\ ] ] here we have defined a new variable @xmath40 . clearly , the @xmath41 dependence of @xmath42 and @xmath43 helps to distinguish among the different form factors of the same operator .now one can introduce the off - forward parton distributions ( ofpd s ) @xmath44 and @xmath45 through their moments @xmath46 since all form factors are real , the new distributions are consequently real .moreover , because of time - reversal and hermiticity , they are _ even _ functions of @xmath41 . for simplicity ,we have not shown here the renormalization scale dependence .frequently we will also omit the @xmath47 dependence .the new distributions are more complicated than the feynman parton distributions because of their dependence on the momentum transfer @xmath48 . as such ,the ofpd s contain two more scalar variables besides the feynman variable @xmath0 .the variable @xmath47 is the usual @xmath47-channel invariant which is always present in a form factor .the @xmath41 variable is a natural product of marrying the concepts of the feynman distribution and form factor : the former requires the presence of a prefered momentum @xmath49 along which the partons are predominantly moving and the latter requires a four - momentum transfer @xmath48 ; @xmath41 is just a scalar product of the two momenta .some insights about the dependence of the ofpd s on these two extra variables will be offered later in section iv.a . in qcd , there are five additional towers of twist - two operators besides that in eq .( [ o1 ] ) : @xmath50 the corresponding off - forward distributions can be labelled by @xmath51 , @xmath52 , @xmath53 , @xmath54 , @xmath55 , @xmath56 , @xmath57 , @xmath58 , and @xmath59 , @xmath60 , respectively .the physical significance of parton distributions in high - energy processes is apparent only in light - cone coordinates and light - cone gauge . to see this , we sum up all the local twist - two operators into a light - cone bilocal operator and express the parton distributions in terms of the latter , @xmath61 the light - cone bilocal operator ( or light - ray operator ) arises frequently in hard scattering processes in which partons propagate along the light - cone .in fact , the taylor - expansion of this operator along the light - cone leads us immediately to the twist - two operators considered in the last subsection .the parton distributions are most naturally defined in terms of the matrix elements of the bilocal operator . in this context , the feynman @xmath0 is just the fourier variable of the light - cone distance .the lorentz structures in the second line in the above equation are independent and complete . in ordinary coodinates ,the correlation of two quark fields in the light - ray operator is in both space and time . as such, the physical interpretation is quite complicated .however , in light - cone coordinates , the physical content of the bilocal operator becomes transparent . to see this ,let us recall that the light - cone coordinates are defined as , @xmath62 in this new system of coordinates , the role of time is played by @xmath63 . in the same fashion , introduce the light - cone dirac matrices @xmath64 and the projection operators , @xmath65 .a dirac field @xmath66 can be decomposed into a sum of @xmath67 , the independent degrees of freedom ( good " ) , and @xmath68 , the dependent degrees of freedom ( bad " ) . in light - cone quantization ,the independent components of the dirac field obey the following commutation relation @xcite @xmath69_+|_{x^+=y^+ } = { 1\over \sqrt{2 } } p_{+\alpha\beta}\delta(x^--y^-)\delta^2(x_\perp - y_\perp ) \ , \ ] ] where @xmath70_+$ ] denotes anticommutation .the above relation can be solved by a fourier expansion @xmath71 and similarly for @xmath72 .the quark ( antiquark ) creation and annihilation operators , @xmath73 ( @xmath74 and @xmath75 ( @xmath76 , obey the commutation relation @xmath77_+ & = & 2k^+ \delta_{\lambda\lambda'}(2\pi)^3 \delta(k^+-k'^+ ) \delta^2(k_\perp - k_\perp ' ) \ , \nonumber \\\left(~\left[d_\lambda(k^+,\vec{k}_\perp ) , d^\dagger_{\lambda'}({k'}^+ , \vec{k'}_\perp ) \right]_+ \right . &= & \left .2k^+ \delta_{\lambda\lambda'}(2\pi)^3 \delta(k^+-k'^+ ) \delta^2(k_\perp - k_\perp ' ) \ .\right ) \end{aligned}\ ] ] in the presence of a gauge field , the above is unchanged if the light - cone gauge , @xmath1 , is chosen .the same gauge choice eliminates the gauge link in the light - ray operator .[ fig1 ] now we can look at the physical content of the off - forward distributions in the light - cone coordinates and gauge .since they are even in @xmath41 , we first restrict ourselves to @xmath78 . substituting the light - cone fock expansion into eq .( [ string ] ) , we have @xmath79 where @xmath80 is a volume factor .the distribution has different physical interpretations in the three different regions . in the region @xmath81 , it is the amplitude for taking a quark of momentum @xmath82 out of the nucleon , changing its momentum to @xmath83 , and inserting it back to form a recoiled nucleon . in the region @xmath84 , it is the amplitude for taking out a quark and antiquark pair with momentum @xmath85 . finally ,in the region @xmath86 , we have the same situation as in the first , except the quark is replaced by an antiquark .the first and third regions are similar to those present in the ordinary parton distributions and hence will be called the dglap region . the middle region is similar to that in a meson amplitude and hence will be called the erbl region . the @xmath41 symmetry is an important property of the ofpd s . to illustrate the parton content of this symmetry , one can work out the equivalent of eq .( [ fock ] ) by assuming @xmath41 is negative .the result is best shown in fig .1 , where the quark and antiquark creations and annihilations are indicated explicitly . from eq . ( [ fock ] ), the off - forward distribution is very much like a quark - nucleon scattering amplitude .use @xmath87 to label the helicity amplitudes , where @xmath88 are the helicities of the initial and final nucleons and @xmath89 are those of the created and annihilated quarks .time - reversal and parity invariance provide the constraints @xmath90 and @xmath91 .therefore , there are total of six independent helicity amplitudes which imply four additional twist - two off - forward distributions @xcite @xmath92 } \over m } u(p ) \ . \end{aligned}\ ] ] their parton interpretation is similar to that of @xmath93 , except that for @xmath94 there is a sign depending on the quark helicity and for @xmath95 there is a quark helicity flip @xcite .off - forward gluon distributions can be introduced in the similar fashion .consider first the gluon - nucleon helicity amplitudes .symmetries limit the number of independent ones to six , and hence there are six leading - twist gluon distributions . herewe quote only the helicity - independent ones : @xmath96 where the gauge link is not shown .because of the bose symmetry , @xmath97. the interested reader can find the other twist - two gluon distributions in @xcite .in light - front coordinates , only @xmath98 are the independent degrees of freedom .they obey the commutation relation @xcite @xmath99|_{x^+=y^+ } = { 1\over 2 } \delta_{ij } \delta^2(x_\perp - y_\perp ) \epsilon(x^--y^-)\ ] ] where @xmath100 .the commutation relation can be solved by the following expansion @xmath101 where @xmath102 and @xmath103 are the gluon creation and annihilation operators . using this expansion , the physical content of the gluon distributionscan be worked out immediately @xmath104 in the region @xmath81 , it is the amplitude for taking a gluon of momentum @xmath82 out of the nucleon , changing its momentum to @xmath83 , and inserting it back to form a recoiled nucleon . in the region @xmath105 , it is the amplitude for taking out a gluon pair with momentum @xmath85 . in an actual calculation of the distributions , the vacuum subtraction in matrix elementsis always implied . as such , it is convenient to insert a @xmath106 product between the two fields at light - cone separations .the difference between a @xmath106 product and an ordinary product is just a constant @xmath107_+ \ .\ ] ] the constant does not contribute to any physical matrix element .therefore , a light - cone @xmath106 product contains only dispersive contributions .the notion of ofpd s we discussed above are essentially that of refs . @xcite and @xcite .the variables @xmath0 and @xmath41 here are the same as the variables @xmath47 and @xmath108 in @xcite . other names( non - forward distributions , off - diagonal distributions , double distributions ) and notations exist in the literature .radyushkin introduced the _ non - forward distributions _@xmath109 in refs .@xcite . to see their relationship to the ofpd s , one can write eq .( [ string ] ) for an arbitrary light - like @xmath19 @xmath110 one can also shift the argument of the string operator using heisenberg s equations of motion .now if one chooses @xmath111 the variables are related according to @xmath112 the matrix element of the string operator becomes @xmath113 compared with eqs .( 9.1 ) and ( 9.2 ) in ref .@xcite , we have @xmath114 the advantages of @xmath115 over @xmath109 are : 1 ) it treats the initial and final nucleon symmetrically and thus @xmath41 symmetry is obvious ; 2 ) it has a simple connection with local hermitian operators .however , the @xmath116 variable in @xmath109 has the advantage of measuring one of the initial parton momenta according to the longitudinal momentum of the initial nucleon , analogous to the forward case .collins et al .introduced the off - diagonal distributions @xmath117 @xcite .they also chose the @xmath19 vector conjugate to the initial state momentum .the relations of their variables to ours are , @xmath118 here @xmath119 is the same as radyuskin s @xmath116 .the off - diagonal distributions are related to the ofpd s by @xmath120 on the other hand , for gluons , one has @xmath121 again , the @xmath41 symmetry is not obvious in these off - diagonal distributions .radyushkin also introduced what he called double distributions @xmath122 , where @xmath0 and @xmath123 are the fractions of the initial nucleon momentum @xmath124 and the momentum transfer @xmath48 carried by the active parton @xcite .it is interesting from the point of view of parametrizing the momentum flow out of a hadron vertex in a @xmath41-independent way .however , the physics of a double distribution is equivalent to that of the corresponding off - forward distribution .one of the principal motivations for studying off - forward distributions is that they provide a new class of observables for the internal structure of the nucleon .we believe that the physics of strong interactions is described by the fundamental theory quantum chromodynamics ( qcd ). however , qcd in the low - energy region is strongly interacting and highly relativistic , involving an infinite number of degrees of freedom .no theory of this type has ever been solved before with a clear physical insight .the only available tool now for solving qcd bound states is lattice qcd @xcite . at this stage, the approach has not yet been developed to a degree of sophistication so that one can reliably calculate fundamental observables with respectable precisions .therefore , any experimental information on the structure of the nucleon is valuable in helping to understand how nature constructed this particle of extreme importance to our existence . since gell - mann and zweig s quark model, many nucleon models have been proposed to correlate the experimental data from low and high energy probes .although these models are not precisely qcd , they have been quite successful phenomenologically .however , the ultimate question confronting models is how the degrees of freedom used in them are related to those in qcd .this question is of high importance because experimental probes do not couple to model degrees of freedom . in practice ,simple assumptions have been used in comparing model predictions and data , e.g. , the constituent quarks are identified as qcd quarks at low - energy scales although a precise notation of the latter is unclear .these assumptions have been seriously challenged in 1987 when the european muon collaboration ( emc ) measured , with an unprecedented precision , the proton s spin - dependent structure function @xmath125 in polarized deep - inelastic scattering . combining their data with the hyperon beta decay rates , augmented with the assumption of the flavor su(3 ) symmetry , emc extracted the fraction of the nucleon spin carried in the spin of quarks : @xmath126 this result is in flat contradiction with the cherished quark model prediction @xmath127 .this so - called spin crisis " has touched off an explosive activity in both theoretical and experimental communities .after ten more years of careful experimental studies , the original emc result has essentially been confirmed , although the small @xmath0 contribution to @xmath128 remains largely uncertain .a recent global analysis of the data can be found , for instance , in ref .@xcite .many theoretical solutions " to the spin crisis " have been proposed in the literature .apart from doubts about the experimental data and the interpretation of them , there have been serious attempts in explaining the emc result from the fundamental theory .for instance , lattice qcd calculations have produced numbers close to the emc result @xcite , although the systematic effects of quenched approximation , finite lattice sizes , and large quark masses remain uncertain . despite these, more fundamental approaches have not yet offered a complete picture about the spin structure of the nucleon . as to the naive quark model prediction, the key question is what has been alluded to abovel : namely , whether the quark model result can be directly compared with the experimental data , or whether the constituent quarks are the same as the qcd quarks probed in high - energy scattering .no convincing argument has been offered towards a positive answer .more recently , progress has been made in attempting to understand how qcd builds up the nucleon spin .the progress potentially has an important impact on other theoretical studies , and ultimately on our basic knowledge of the internal structure of the nucleon . on the one hand, it suggests a thorough examination of the spin structure in a fundamental way . on the other hand , they called for understanding the role of the gluon degrees of freedom in model building .as it turns out , a key element in unraveling the nucleon spin is the off - forward parton distributions .the history began with identifying the sources of the qcd angular momentum .jaffe and manohar wrote down the complete angular momentum operator in a free - field - theory form which contains naturally the quark and gluon spin and orbital contributions @xcite .the importance of the quark orbital angular momentum in qcd has been recognized before by sehgal @xcite and ratcliffe @xcite in different contexts .the emc result indicates , among others , that a large fraction of the nucleon spin is carried by other sources of angular momentum . in order to associate these contributions to possible observables ,the qcd angular momentum operator has been rewritten in a gauge - invariant form @xcite , @xmath129 where @xmath130 \ , \nonumber \\ \vec{j}_g & = & \int d^3x ~\vec{x } \times ( \vec{e } \times \vec{b } ) \ .\label{ang}\end{aligned}\ ] ] the quark and gluon parts of the angular momentum are generated from the quark and gluon momentum densities @xmath131 and @xmath132 , respectively .@xmath133 is the dirac spin - matrix and the corresponding term is clearly the quark spin contribution .@xmath134 is the covariant derivative and the associated term is the gauge - invariant quark orbital contribution . with the above expression, one can easily construct a sum rule for the spin of the nucleon . consider a nucleon moving in the @xmath135-direction , and polarized in the helicity eigenstate @xmath136 .the total helicity can be evaluated as an expectation value of @xmath137 in the nucleon state @xmath138 where the three terms denote the matrix elements of three parts of the angular momentum operator in eq .( [ ang ] ) .the physical significance of each term is obvious , modulo the scale and scheme dependence indicated by @xmath11 .the scale dependence in @xmath139 is generated from the u(1 ) axial anomaly , and there have been attempts to remove the scale dependence by subtracting a gluon contribution @xcite .unfortunately , such a subtraction is by no means unique .here we adopt the standard definition of @xmath140 as the matrix element of the multiplicatively renormalized quark spin operator .note that the individual term in the above equation is independent of the momentum of the nucleon .in particular , it applies when the nucleon is travelling with the speed of light ( the infinite momentum frame ) @xcite .the scale dependence of the quark and gluon contributions can be calculated in perturbative qcd . by studying renormalization of the nonlocal operators, one can show @xcite @xmath141 as @xmath142 , there exists a fixed point solution @xmath143 thus as the nucleon is probed at an infinitely small distance scale , approximately one - half of the spin is carried by gluons .a similar result has been obtained by gross and wilczek in 1974 for the quark and gluon contributions to the momentum of the nucleon @xcite .strictly speaking , these results reveal little about the nonperturbative structure of bound states .however , experimentally it was found that about half of the nucleon momentum is carried by gluons already at quite low energy scales @xcite .thus the gluon degrees of freedom not only play a key role in perturbative qcd , but also are a major component of nonperturbative states , as they should be . an interesting question is then , how much of the nucleon spin is carried by the gluons at low energy scales ?a solid answer from the fundamental theory is not yet available .recently , balitsky and this author have made an estimate using the qcd sum rule approach @xcite .we found @xmath144 which yields approximately 0.25 .if this calculation indicates anything of the truth , the spin structure of the nucleon roughly looks like @xmath145 a quark model calculation of @xmath146 has also been made recently by barone et al . @xcite .the result also indicates @xmath147 at low - energy scales . by examining carefully the definition of @xmath148 @xmath149one realizes that they can be extracted from the form factors of the quark and gluon parts of the qcd energy - momentum tensor @xmath150 .specializing eq .( [ form ] ) to ( @xmath151 ) @xmath152 u(p)\ .\ ] ] taking the forward limit in the @xmath153 component and integrating over 3-space , one finds that @xmath154 give the momentum fractions of the nucleon carried by quarks and gluons ( @xmath155 ) . on the other hand , substituting the above into the nucleon matrix element of eq .( [ matrix ] ) , one finds @xcite @xmath156 \ .\end{aligned}\ ] ] therefore , the matrix elements of the energy - momentum tensor provide the fractions of the nucleon spin carried by quarks and gluons .there is an analogy for this . if one knows the dirac and pauli form factors of the electromagnetic current , @xmath157 and @xmath158 , the magnetic moment of the nucleon , defined as the matrix element of ( 1/2)@xmath159 , is @xmath160 . since the quark and gluon energy - momentum tensors are just the twist - two , spin - two , parton helicity - independent operators , we immediately have the following sum rule from the off - forward distributions @xmath161= a_q(t ) + b_q(t ) \ , \end{aligned}\ ] ] where the @xmath41 dependence , or @xmath162 contamination , drops out . extrapolating the sum rule to @xmath163 , the total quark ( and hence quark orbital ) contribution to the nucleon spinis obtained .a similar sum rule exists for gluons .thus a deep understanding of the spin structure of the nucleon may be achieved by measuring the ofpd s from high - energy experiments .comparing with the ordinary parton distributions , off - forward distributions depend on two additional variables : the @xmath47-channel longitudinal momentum fraction @xmath41 and the invariant mass @xmath47 . because of these , the ofpd s contain much more information and are necessarily more difficult to model in practice . at present, there is little experimental data to constrain these distributions . in this section, we will describe a few of the recent studies of the distributions at low - energy scales where they reflect directly the nonperturbative structure of the nucleon .we will also discuss their evolution equations which have been computed to leading logarithmic order in perturbation theory .some qualitative features of the evolution to an asymptotically large energy scale will be highlighted .a few rigorous results about the ofpd s are known .first of all , in the limit of @xmath164 and @xmath165 , they reduce to the ordinary parton distributions . for instance , @xmath166 where @xmath167 and @xmath168 are the unpolarized and polarized quark densities .similar equations hold for gluon distributions . for practical purposes , in the kinematic region where@xmath169 an off - forward distribution may be approximated by the corresponding forward one .the first condition , @xmath170 , is crucial - otherwise there is a significant form - factor suppression which can not be neglected at any @xmath0 and @xmath41 . for a given @xmath47 ,@xmath41 is restricted to @xmath171 therefore , when @xmath172 is small , @xmath41 is automatically limited and there is in fact a large region of @xmath0 where the forward approximation holds .the first moments of the off - forward distributions are constrained by the form factors of the electromagnetic and axial currents . indeed , by integrating over @xmath0, we have @xcite @xmath173 the @xmath47 dependence of the form factors are characterized by hadron mass scales .therefore , it is reasonable to speculate that similar mass scales control the @xmath47 dependence of the off - forward distributions . in the small @xmath0 region, it has been argued that the @xmath47 dependence corresponds to a mass scale @xmath174 or something close to the inverse slope of the elastic @xmath175 cross section , @xmath176 gev@xmath177 @xcite .martin and ryskin made a very important observation that one can obtain an upper bound on the off - forward distributions by using the schwarz inequality @xcite .for instance , consider the quark distribution in eq .( [ fock ] ) and take a ket @xmath178 . since @xmath179 , we have the following constraint @xmath180 for @xmath81 , where @xmath181 and @xmath182 which is different from that in eq .( [ x2 ] ) .similarly , one has @xmath183 for @xmath86 . in the middle region , however , the vacuum subtraction ruins a useful constraint .an upper bound for the off - forward gluon distribution can be derived in a similar way : @xmath184 for @xmath81 . because of the different definitions of @xmath185 , the above inequality is slightly different from martin and ryskin s .however , they are practically the same in the small @xmath0 region .it must be cautioned that the above inequalities are derived by disregarding renormalization of the distributions , and hence are correct in the leading - logarithmic approximation only .a first model calculation of the quark off - forward distributions was reported in ref .the calculation was done in the mit bag model , whose parameters are adjusted so that the electromagnetic form factor and the feynman parton distributions are best produced .the shape of the distributions as a function of @xmath0 are rather similar at different @xmath47 and @xmath41 : peaked around @xmath186 @xmath187 0.4 , and tapering off rather quickly as @xmath188 , 0 and negative , where the calculation can not be trusted .the @xmath47 dependence of the energy - momentum form factors is controlled by a mass parameter between 0.5 and 1 gev@xmath177 .the @xmath41 dependence of the distributions is surprisingly weak ; a clear physical reason is unknown .the off - forward quark distributions were also studied in the chiral quark - soliton model by petrov et al .in contrast to the bag model results , the chiral soliton model yields a rather strong @xmath41 dependence of the off - forward distribution .the model also predicts qualitatively different behaviors in the regions @xmath189 and @xmath190 , with discontinuities at @xmath191 .although the discontinuities are believed to be an artifact of the calculation , the qualitatively different features in different regions are in line with the interpretation of the ofpd s . from eq .( [ fock ] ) , we understand that the off - forward distributions have similar physics as the forward distributions in @xmath189 .when @xmath190 , however , the distributions reflect @xmath47-channel exchanges of meson and glueball quantum numbers .therefore , a physics - oriented modelling in this region may consider the convolution of the meson cloud of a nucleon with the leading light - cone wave function of the mesons .models can not be used yet to calculate off - forward gluon distributions in the small @xmath0 region , which are phenomenologically important @xcite .it is believed in the literature that the distributions in this region and negligible @xmath47 are independent of @xmath41 @xcite . as we mentioned earlier , this might be correct as long as @xmath192 .for @xmath193 , however , the best argument so far is offered by martin and ryskin @xcite . from the diffractive scattering phenomenology, they argue that the distribution @xmath194 at small @xmath0 should be bounded below by the forward distribution at @xmath195 , @xmath196 combining this with the upper bound in eq .( [ gmr ] ) , they concluded phenomenologically when @xmath197 , @xmath198 at @xmath199 gev@xmath177 for the grv gluon distribution @xcite and at @xmath200 gev@xmath177 for the mrs(r2 ) distribution @xcite .parton distributions are renormalization scale and scheme dependent because the defining operators are .thus , a natural starting point for discussing scale evolution is to consider renormalization of twist - two operators .however , it is well known that the parton evolution has a simple physical interpretation in light - cone coordinates and gauge , as exemplified by the dglap evolution kernels .therefore , we shall also consider the off - forward parton evolution kernels .renormalization of the twist - two operators appearing in feynman parton distributions is a special example of a more general case . for a given twist - two operator of spin @xmath19 shown in eq .( [ o1 ] ) , apart from its mixing with the gluon operator of same quantum numbers it also mixes with the following operators with total derivatives , @xmath201 consideration of such mixing is already necessary in studying the evolution of the leading - twist meson wave functions .the answer was first obtained by efremov and radyushkin @xcite , and by brodsky and lepage @xcite .actually , at the leading logarithmic order the answer may be guessed from the naive conformal symmetry which is broken by quantum effects only at the next - to - leading - logarithmic order @xcite .the following combination of the twist - two operators furnishes a representation of the special conformal symmetry group , @xmath202 where the @xmath203 are the gegenbauer polynomials of order 3/2 .its evolution takes exactly the diagonal form of that for the operator without the total derivatives @xcite @xmath204 where @xmath205 \ , \ ] ] and @xmath206 .@xmath207 for the su(3 ) color group and @xmath208 is the number of light quark flavors . for the leading - twist pion wave function ,the above result leads to a general expansion in terms of gegenbauer polynomials .generalization to the singlet case is in principle straightforward . for the twist - two operators with natural parity ,the evolution was first worked out by chase @xcite . for those with unnatural parity, it was worked out by ohrndorf @xcite , shifman and vysotsky @xcite , and baier and grozin @xcite .the conclusion of these studies is that the gluon towers of operators evolve in the leading - logarithmic approximation exactly like the forward case if they are combined according to the gegenbauer polynomials of order 5/2 .these results were confirmed later in a more elegant form of light - ray or string operators by geyer et al .@xcite , and by balitsky and braun @xcite . in light of these developments , evolution of off - forward distributionscan simply be worked out by converting the anomalous dimensions of the composite operators into evolution kernels . however , as in the case of the dglap evolution @xcite , one can obtain the same results more directly through studying parton splitting processes . a first such study for the nonforward , helicity - independent parton evolution was made by bartels and loewe in the small @xmath0 limit @xcite .gribov et al .presented the complete leading - order kernels including the singlet mixing @xcite .mller et al . found the non - singlet evolution kernel by generalizing the brodsky - lepage result @xcite .more recently , the issue has been reexamined from different angles by radyushkin @xcite , ji @xcite , chen @xcite , frankfurt et .@xcite , blmlein et al .@xcite , and martin and ryskin @xcite . as an example of the off - forward evolution, we cite below the helicity - independent result in ref .@xcite .according to eq .( [ fock ] ) , it is natural that the distributions in three different kinematic regions evolve differently . in the region @xmath81 where we have a quark creation and annihilation ,the evolution equation is , @xmath209 where @xmath210 \ .\ ] ] the parton splitting function is @xmath211 the end - point singularity is cancelled by the divergent integrals in @xmath212 . obviously , when @xmath213 , the splitting function becomes the usual dglap evolution kernel @xcite .for @xmath214 where we have creation or annihilation of a quark - antiquark pair , the evolution takes the form , @xmath215 f_{ns}(y,\xi , q^2 ) \ , \ ] ] where @xmath216 when @xmath217 , the kernel reduces to the brodsky and lepage kernel @xcite . for @xmath86 where we have antiquark creation and annihilation, the evolution takes the same form as eq .( [ ap ] ) , apart from the replacement @xmath218 . the helicity - dependent off - forward distributions were introduced in ref .@xcite and their evolution was also studied in that paper .subsequently , the kernels have been confirmed by the studies of blmlein et al .@xcite and balitsky and radyushkin @xcite .the helicity - flip off - forward quark distributions were introduced by collins et al .@xcite , and their evolution was studied by belitsky and mller @xcite .the helicity - flip gluon distributions were defined by hoodbhoy and ji @xcite who also derived the leading - logarithmic evolution .it is worthwhile to point out that the combinations of the helicity - flip quark and gluon operators according to the gegenbauer polynomials of order 3/2 and 5/2 respectively also form representations of the naive conformal symmetry group and hence have diagonal evolutions at the leading - logarithmic level . as in the forward case, the off - forward parton evolution equations can be solved in many different ways .a few approaches have already been suggested in the literature @xcite .one important aspect of the evolution is the asymptotic form of the off - forward distributions .radyushkin found @xcite ( also @xcite ) for the non - singlet quark distributions , @xmath219 where @xmath220 is the appropriate electromagnetic form factor .this asymptotic form has interesting implications . when @xmath213 , the asymptotic form is a @xmath221-function in @xmath0 .this reminds us of the well - known fact that as @xmath222 , parton distributions migrate to smaller and smaller @xmath0 . for @xmath217 ,the evolution is that of a meson wave function with an asymptotic limit @xmath223 , which has a broad shape in @xmath0 .therefore , it is quite plausible that for an intermediate @xmath41 , one would see the migration of the distribution from large to small @xmath224 if @xmath189 .however , in the middle region @xmath190 , the total strength increases and at the same time , the distribution evens out toward @xmath223 .these features of evolution were observed in the numerical calculations by belitsky et al .@xcite and mankiewicz et al .@xcite .the evolution effects for the off - forward gluon distribution at small @xmath0 are especially interesting from the phenomenological point of view .these effects have been studied by frankfurt et al .@xcite , and martin and ryskin @xcite . in ref .@xcite , they assumed negligible off - diagonal effects at @xmath225 gev , and then evolved the off - diagonal distribution to @xmath226 , 17 , 41 , and 110 gev at @xmath227 , and @xmath228 .for @xmath229 to 1 , they found that the ratio of off - diagonal to diagonal distributions ranges from 1.2 to 1.7 .they pointed out that for a fixed @xmath230 , the off - diagonal distribution at large @xmath231 is obtained from that in the much larger @xmath0 region , however , with the same difference @xmath230 where the off - diagonal effect is very small .therefore at very small @xmath0 , unless @xmath231 is very close to hadron mass scales , the off - diagonal effects come mostly from the perturbative evolution .these effects are less than a factor of 2 for the kinematics region considered .similar studies have also been made in ref .the qualitative conclusion is the same although quantitatively there are differences .martin and ryskin pointed out that the off - diagonal evolution happens faster than the diagonal one because the difference in the splitting functions is positive .furthermore , the difference is only important when @xmath231 is close to the final target scale and @xmath0 becomes comparable to the size of @xmath41 .eventually , usefulness of the ofpd s depends on whether they can actually be measured in any experiment . the simplest , and possibly the most promising , type of experiments are hard electro- or deep - inelastic production of photon , mesons , and perhaps even lepton pairs . in this section, we will consider separately two experiments : deeply virtual compton scattering ( dvcs ) in which a real photon is produced , and diffractive meson production .there are practical advantages and disadvantages from both processes .real photon production is , in a sense , cleaner but the cross section is reduced by an additional power of @xmath232 .moreover , the dvcs amplitude interferes with the bethe - heitler one .meson production may be easier to detect , however , it has a twist suppression , @xmath233 .in addition , the theoretical cross section depends on the unknown light - cone meson wave function . at present , there exist already some data from hera on meson production @xcite .more experiments will be done in the future at hera , cern , and jefferson lab .deeply virtual compton scattering was proposed in ref .@xcite as a practical way to measure the off - forward distributions .the result was confirmed in ref .@xcite using the double distributions .the general two - photon processes had actually been studied earlier by watanabe @xcite and mller et al .@xcite , and later by chen @xcite .recently , there have been a number of new results obtained about this process , which will be described below .consider the virtual photon scattering shown in fig .2a , where the momenta of the incoming ( outgoing ) photon and nucleon are @xmath234 and @xmath235 , respectively .the compton amplitude is defined as @xmath236 where @xmath237 . in the bjorken limit ,@xmath238 and @xmath239 and their ratio remains finite , the scattering is dominated by the single quark process shown in fig .2b in which a quark absorbs the virtual photon , immediately radiates a real one , and falls back to form the recoiled nucleon . in the process, the initial and final photon helicities remain the same .the leading - order compton amplitude is @xcite @xmath240 where @xmath19 and @xmath36 are the conjugate light - cone vectors defined according to the collinear direction of @xmath241 and @xmath242 , and @xmath243 is the metric tensor in the transverse space .@xmath41 is related to the bjorken variable @xmath244 by @xmath245 .[ fig2 ] the same dvcs final state can also be produced through the bethe - heitler process in which the initial or final state electron ( or muon ) radiates a real photon and at the same time scatters elastically off the target nucleon .this process can be calculated as accurately as the data on the nucleon elastic form factors .physically , the bethe - heitler process is a background to dvcs and can overshadow the latter signal if the former is too large . on the other hand , with an appropriate size of bethe - heitler amplitude, one may hope to measure its interference with the dvcs amplitude which then can be directly extracted from data .the complete leading - order dvcs cross sections and the interference with the bethe - heitler amplitude can be found in ref .@xcite . recently , guichon made some interesting estimates of the cross sections at compass and jefferson lab energies @xcite .he found that while at low scattering energy , the bethe - heitler background is quite large , it drops significantly at high energies when the virtual photon flux is large .in fact , if one is focused on production of real photons along the direction of the virtual photon , the dvcs cross section is quite significant at compass energy , and certainly far above the bethe - heitler background . as far as studying the orbital angular momentum of quarksis concerned , this is the most interesting kinematic region .guichon s estimate is encouraging for a real experiment .the one - loop corrections to dvcs have recently been studied by osborne and ji@xcite .they have also been studied by mller @xcite and by belitsky and mller @xcite using the constraints from conformal symmetry .results of these studies have been confirmed by mankiewicz et al . @xcite . besides their obvious use in precision analysis ,the results indicate that dvcs is factorizable at the next - to - leading order .a higher - order calculation of the coefficient functions in the renormalon - chain approximation has been done by belitsky and schfer @xcite .an all order proof of the dvcs factorization was first given by radyushkin @xcite and recently in a different perspective by osborne and ji @xcite , and collins and freund @xcite .the simple dvcs scattering mechanism can be tested through scaling relations and selection rules , as in the case of deep - inelastic scattering . in ref .@xcite , diehl et al . considered the possibility of using @xmath246 dependence of the cross section to test the single quark scattering picture , where @xmath246 is the angle between the lepton and hadron scattering planes .they observed that through qcd radiative corrections , the photon helicity - flip amplitude may contribute to dvcs .a recent study by hoodbhoy and ji showed that this indeed happens @xcite .the key element here is the helicity - flip off - forward gluon distributions , which decouple in the forward limit due to angular momentum conservation .they pointed out that a nonvanishing leading - twist photon helicity - flip amplitude signals unequivocally the presence of gluons in the nucleon .there are other processes that are as clean as the dvcs , but more difficult to measure experimentally .for instance , @xmath2 production was considered in ref .since @xmath2 s are signaled through lepton pairs , one can also consider lepton - pair production through intermediate time - like virtual photons .theoretical treatment of these is exactly the same as for dvcs .heavy quarkonium production was first considered by ryskin as a way to measure the feynman gluon distributions at small @xmath0 @xcite .a leading - order diagram for the process is shown in fig 3a , where the virtual photon fluctuates into a @xmath247 pair which subsequently scatters off the nucleon target through two - gluon exchanges . in the process , the pair transfers away certain amount of its longitudinal momentum and reduces its invariant mass to that of a @xmath248 .the cross section is , @xmath249 ^ 2\ ] ] where @xmath250 , @xmath251 is the @xmath3 mass , and @xmath252 is the decay width into the lepton pair .the equation was derived in the kinematic limit @xmath253 and the fermi motion of the quarks in the meson was neglected .two other important approximations were used in the derivation .first , the contribution from the real part of the amplitude is neglected , which may be justifiable at small @xmath0 .second , the off - forward distributions are identified with the forward ones . as we have discussed in iv.a , even for small @xmath47 ,the approximation is good only if the off - diagonal evolution effects can be neglected .a more refined treatment of the process and comparison with hera data can be found in ref .@xcite .[ fig3 ] the above result was extended to the case of light vector - meson production by brodsky et al . , who considered the effects of meson structure in perturbative qcd @xcite .they found a similar cross section , @xmath254 where the dependence on the meson structure is in the parameter @xmath255 and @xmath256 is the leading - twist light - cone wave function .evidently , the above formula reduces to ryskin s in the heavy - quark limit @xmath257 .more discussions on vector meson production along this line , including comparison with recent data , can be found in ref .@xcite .radyushkin was the first to calculate the amplitude of hard diffractive electroproduction in terms of off - forward gluon distributions @xcite .with the virtual photon and vector meson both polarized longitudinally , his result corresponds to @xmath258 where again @xmath259 .the above formula is valid for any @xmath260 and @xmath47 smaller or around hadron mass scales .hoodbhoy has also studied the effects of the off - forward distributions in the case of @xmath3 production @xcite .he found that ryskin s result needs to be modified in a similar way once the off - forward effects becomes important .collins , frankfurt , and strikman have made an extensive analysis of the high - order qcd contributions to hard diffractive meson production @xcite . using the standard toolsdescribed in @xcite , they showed that the leading - twist contributions are factorizable to all orders in perturbation theory .the factorization theorem provides a rigorous qcd basis for calculating the process in terms of the off - forward distributions .in fact , the paper outlined a recipe for calculating a whole class of processes in perturbation theory .mankiewicz , piller and weigl recently calculated the neutral pseudoscalar and vector meson production cross sections in terms of the non - forward quark and gluon distributions @xcite .shown in fig .3b is a feynman diagram in which meson production depends on a quark distribution .they also investigated charged particle production in which the initial nucleon has a different charge from the final one @xcite .the charge - changing off - forward distributions can be obtained from the charge - conserving ones using isospin symmetry .the result can be generalized to the octet baryons if the su(3 ) flavor symmetry is good .one can also consider production of heavy - flavored baryons from the nucleon target . unlike the heavy - quark distributions in light hadrons ,the corresponding off - forward distributions are not suppressed by heavy - quark masses .the off - forward parton distributions are a new type of nucleon observables which have recently generated much interest in hadron physics community . on the one hand , the ofpd s contain all the form factors of the twist - two operators which generalize the well - known electroweak currents .the form factors of the spin - two , twist - two operators ( energy - momentum tensor ) are especially interesting in unravelling the spin structure of the nucleon . on the other hand , they summarize succinctly the leading structural dependence in a class of high - energy processes involving the nucleon .the partonic interpretation of the ofpd s is quite simple it represents parton - nucleon scattering amplitudes on the light - cone .recent studies of the ofpd s can largely be classified into three categories : modelling the new distributions , studying their scale dependences , and investigating processes that are sensitive to them . in terms of model studies ,our accomplishments so far are rather limited . even though there are bag and chiral soliton model calculations , these calculations need much improvement in the future .in particular , the physics in the @xmath190 region has not been entirely captured in these calculations .the models are yet to be extended to including gluon degrees of freedom which are known to be important even at low energy scales .indeed , a vanishing helicity - flip gluon distribution at low - energy scales will remain zero at other scales because it has no mixing with any quark distribution .a number of rigorous constraints discussed in section iv.a can be helpful in modelling : apart from the special limits and @xmath261 moments , there exist useful upper bounds for the ofpd s .there have been many studies in perturbative evolution . at the leading - logarithmic level ,evolution equations now have been obtained for all leading - twist ofpd s . atnext - to - leading order , one needs two - loop anomalous dimensions which exist only for the spin - independent non - singlet operators @xcite .calculations of additional two - loop anomalous dimensions will be needed as data on the ofpd s become available .recent works of mller and belitsky illustrate the power of exploiting the anomalous conformal symmetry to obtain these anomalous dimensions @xcite .numerical studies of evolution effects have just begun .qualitatively we can understand the evolution effects from its asymptotic behavior .quantitatively , one needs to establish high - precision , high - speed evolution codes which allow efficient data fitting in the future .the leading - order cross sections for dvcs and meson production can be found in the literature .for dvcs , the cross section has also been computed to the next - to - leading order in @xmath262 . a similar study for meson productioncan be made , and will be necessary once precision data become available .higher - twist effects are important at low @xmath263 and must be investigated thoroughly . generally speaking , they are more complex to study than that in deep - inelastic scattering because of the extra photon or meson , and the extra soft scale @xmath47 .additional processes may be explored to probe the ofpd s , including ones in which the final state baryon may contain strange or heavy flavors .polarization observables need to be seriously investigated .a solid experimental confirmation of the scattering mechanisms involving the ofpd s is itself very significant .it involves testing scaling relations and selection rules .finally , we notice that the ofpd s are functions of multi - variables . as such, it would be difficult to get a complete picture of them in a short time period and in a single type of experiments .moreover , experimental cross sections can not be used to extract the distributions directly ; they usually involve integrals over parton momenta .therefore , it will be necessary to parametrize the ofpd s and fit the parameters to data . in finding appropriate ways to make parametrizations and fits, model studies will play important role , unless of course lattice calculations can make a significant breakthrough in the near future . *note added : * after this paper was submitted for publication , the author has noticed some new works in the literature which are relevant to the subject of this paper . in ref .@xcite , radyushkin studied both hadron form factors and wide - angle compton scattering using nonforward parton densities . in ref .@xcite , diffractive dijet photoproduction was proposed as a probe of the off - diagonal gluon distribution . in ref .@xcite , certain positivity constraints for off - forward parton distributions were derived . in a different line of development ,the orbital angular momentum distribution in the nucleon was investigated @xcite . in the last two yearsi have benefited considerably from discussions surrounding the topic of this article with s. brodsky , j. collins , p. hoodbhoy , a. mueller , j. osborne , a. radyushkin , g. sterman , m. strikman , and others , and i wish to thank them all .i also would like to thank a. martin for his encouragement in writing this topical review , k. golec - biernat , a. radyushkin and l. robinette for a critical reading of the manuscript , and j. osborne for drawing the figures .this work is supported in part by funds provided by the u.s .department of energy ( d.o.e . ) under cooperative agreement doe - fg02 - 93er-40762 .dittes and a. v. radyushkin , phys . lett .* 134b * ( 1984 ) 359 ; m. h. sarmadi , _ ibid . _ * 143b * ( 1984 ) 471 ; g. r. katz , phys . rev .* d31 * ( 1985 ) 652 ; s. v. mikhailov and a. v. radyushkin , nucl .b254 * ( 1985 ) 89 .

<S> recently , there have been some interesting developments involving off - forward parton distributions of the nucleon , deeply virtual compton scattering , and hard diffractive vector - meson production . </S><S> these developments are triggered by the realization that the off - forward distributions contain information about the internal spin structure of the nucleon and that diffractive electroproduction of vector mesons depends on these unconventional distributions . </S><S> this paper gives a brief overview of the recent developments . </S>

1502.02914

the high energy stereoscopic system ( h.e.s.s . ) is located near the gamsberg mountain in namibia at an altitude of 1800 meters .detects @xmath1-ray photons by recording the cherenkov light produced by the electromagnetic shower resulting from the interaction of the photons with the atmosphere .the phase i of the project was completed in december 2003 . at this timethe array was made of four 12-meters telescopes .each telescope has a camera composed of 960 photo - multipliers ( pmts ) and works in stereoscopic mode .the h.e.s.s .phase i array has a field of view of 5 degrees , an angular resolution of 0.1 degree for an energy threshold down to 100 gev .the array is taking data for more than 10 years now and has increased the catalogue of sources detected in the very high energy range ( vhe , @xmath2100 gev ) and our knowledge of the field .has discovered more than 80 objects ( galactic or extragalactic , fig .[ source ] ) , performed a deep galactic plane survey , dark matter searches , multiwavelength campaigns with other instruments and studies of the extragalactic background light with blazars .one of the aims of h.e.s.s . is the detection of gamma - ray bursts ( grbs ) . while no grb has been detected so far in the vhe range ,more than 20 follow - up have been performed @xcite .also monitors variable and bright objects and responds to target of opportunity in order to better know the mechanisms that produce the variability of blazars .the experiment entered into its phase ii with the addition of a fifth telescope ( named ct5 ) placed in the middle of the array .the dish is 32.6 meters by 24.3 meters , equivalent to 28-meters circular dish for a focal length of 36 meters .the telescope is equipped with a alt - az mount @xcite .the camera composed of 2048 pmts for a total weight of @xmath3 3 tons , can record 3600 images per second @xcite and is mounted on an auto - focus system .a picture of this telescope is show on figure [ ct5 ] .the field of view of this telescope is 3.5 degrees for an angular resolution from @xmath3 0.4 degree to less than 0.1 .the energy coverage of the h.e.s.s .experiment is then extended down to energies of a few tens of gev .characteristics of the ct5 telescope are summarized in table [ tablect5 ] .phase ii is the first hybrid array of cherenkov telescopes and is designed to work in different configurations .data can be taken by ct5 only in the so - called _mono mode_. the _ hybrid mode _ involves all five telescopes for a better sensitivity in the entire energy range .stereoscopic observations with only the four 12-meters telescopes are still possible .the ability to split the array in 2 ( ct5 mono mode + the four 12 meters telescopes ) allows to increase the observation time which is rather low for such an experiment ( @xmath3 1000 hours per year ) ..characteristics of the ct5 telescope . [ cols="<,^ " , ] [ table2155 ] pks 2155 - 304 is naturally a good target for calibration purposes and has been observed for a total live time of 42.9 hours in 2013 .the analysis yields an excess of more the 4400 events for a detection at a 29.7 @xmath4 level ( table [ table2155 ] ) .excess map of the field of view and sed are shown on figure [ pksmap ] .the 2fgl and 1fhl are also used for comparison . a large part of the data were taken during a multiwavelength campaign with nustar and _ fermi_. these data were used to build the most precise sed of this object to date : nustar extended the x - ray spectrum up to 79 kev and _ fermi _ pass 8 data were used .more details and results are given in @xcite and in a forthcoming publication .one of the main goals of the h.e.s.s .phase ii is to study the variability on short time - scales .figure [ sens ] presents the differential sensitivity of ct5 and _ fermi _ as a function of time . below @xmath5 seconds, ct5 is clearly much more sensitive that the lat above 25 gev .this opens a new window for the detection of grbs in this energy range .an alert system has been developed to reply to gcn alerts and an automatic re - pointing procedure is in place in case of such an alert .ct5 can , in such case , be on target within a minute @xcite allowing prompt observations .the h.e.s.s . array is taking data now since more than 10 years and has allowed many discoveries . with the advent of the phase ii of the experiment, a new window has been opened in the 30 gev-100 gev energy range . after a commissioning phase , ct5 is now running running in a normal operation mode .the support of the namibian authorities and of the university of namibia in facilitating the construction and operation of h.e.s.s .is gratefully acknowledged , as is the support by the german ministry for education and research ( bmbf ) , the max planck society , the french ministry for research , the cnrs - in2p3 and the astroparticle interdisciplinary programme of the cnrs , the u.k .science and technology facilities council ( stfc ) , the ipnp of the charles university , the polish ministry of science and higher education , the south african department of science and technology and national research foundation , and by the university of namibia .we appreciate the excellent work of the technical support staff in berlin , durham , hamburg , heidelberg , palaiseau , paris , saclay , and in namibia in the construction and operation of the equipment . the _ fermi_-lat collaboration acknowledges support for lat development , operation and data analysis from nasa and doe ( united states ) , cea / irfu and in2p3/cnrs ( france ) , asi and infn ( italy ) , mext , kek , and jaxa ( japan ) , and the k.a .wallenberg foundation , the swedish research council and the national space board ( sweden ) .science analysis support in the operations phase from inaf ( italy ) and cnes ( france ) is also gratefully acknowledged .aharonian , f. , akhperjanian , a. g. , bazer - bachi , a. r. , et al .2005 , a&a , 442 , 895 aharonian , f. , akhperjanian , a. g. , barres de almeida , u. , et al . 2008 , a&a , 477 , 481 aharonian , f. , akhperjanian , a. g. , barres de almeida , u. , et al .2009 , a&a , 495 , 505 aharonian , f. , akhperjanian , a. g. , anton , g. , et al .2009 , apjl , 696 , l150 abramowski , a. , acero , f. , et al .2010 , a&a , 520 , aa83 aharonian , f. , et al .2007 , apjl , 664 , l71 aharonian , f. , et al .2006 , a&a , 457 , 899 abramowski , a. , aharonian , f. , ait benkhali , f. , et al . 2015 ,arxiv:1501.05087 bolmont , j. , corona , p. , gauron , p. , et al .2014 , nuclear instruments and methods in physics research a , 761 , 46 de naurois , m. & rolland , l. 2009 , astroparticle physics , 32 , 231 hofverberg , p. , kankanyan , r. , panter , m. , et al .2013 , arxiv:1307.4550 kastendieck , m. a. , ashley , m. c. b. , & horns , d. 2011 , a&a , 531 , aa123 nolan , p. l. , abdo , a. a. , ackermann , m. , et al .2012 , apjs , 199 , 31 sanchez d.a .et al , probe of lorentz invariance violation effects and determination of the distance of pg 1553 + 113 , these proceedings .sanchez d.a .et al , multiwavelength campaign on the hbl pks 2155 - 304 : a new insight on its spectral energy distribution , these proceedings .

<S> the number of extragalactic sources detected at very hight energy ( vhe , e@xmath0100gev ) has dramatically increased during the past years to reach more than fifty . </S><S> the high energy stereoscopic system ( h.e.s.s . ) had observed the sky for more than 10 years now and discovered about twenty objects . with the advent of the fifth 28 meters telescope , the h.e.s.s . </S><S> energy range extends down to 30 gev . </S><S> when h.e.s.s . </S><S> data are combined with the data of the fermi large area telescope , the covered energy range is of several decades allowing an unprecedented description of the spectrum of extragalactic objects . in this talk , a review of the extragalactic sources studied with h.e.s.s . </S><S> will be given together with first h.e.s.s . </S><S> phase ii results on extragalactic sources . </S>

cond-mat0610010

although there is a large volume of literature devoted to the effects of quenched disorder , there has been relatively little study on how it influences the behavior of systems whose pure versions undergo a first - order phase transition .this question was first addressed by imry and wortis @xcite who showed that inhomogeneities may cause local variations of the transition temperature inside the sample .provided that the cost in interface energy is not great , bubbles of the wrong " phase are formed , eventually leading to a substantial rounding of the transition . a theorem due to aizenman and wehr @xcite shows that in less than two dimensions there can be no phase coexistence at the transition , and therefore no latent heat , in a system with quenched random impurities .therefore these systems are expected to always exhibit a continuous transition .the influence of quenched impurities coupling to the local energy density has been extensively studied by cardy @xcite .a mapping to the random field ising model , whose renormalization group flows are well known , addresses the question of what happens in higher than two dimensions where the aizenman - wehr theorem is not applicable .it was found that , depending on the specific values of parameters such as the laten heat and the surface tension , the phase transition can either be first or second order .more conclusive analytic results were obtained in an ising model with discrete order parameter @xcite .the pure system exhibits a fluctuation - driven first order transition , where the mean field theory predicts a continuous transition but fluctuation effects make it discontinuous . quenched randomness eventually drives the transition to become continuous in two dimensions , in accordance with the aizenman - wehr theorem , but this may or may not happen in higher dimensions .the majority of studies in this field are carried out for spin - glass or analogous systems , e.g. with a frustrated dipolar ordering @xcite . however , in such systems the experimental work is difficult and results are often indirect . in contrast , the quadrupolar orientational ordering of nematic liquid crystals offers the easy experimental access to thermodynamic and structural features of phase transitions . the classical work on frustrated nematic liquids by bellini et al .@xcite has generated a large interest in studies of liquid crystals in random environments , such as porous silica gel .one must appreciate , however , that the characteristic length scale of such disorder is much greater than the coherence length of the nematic order parameter and thus the theoretical concept of a continuous coarse - grained random field @xcite is difficult to sustain .quenched disorder is intrinsically present in nematic elastomers as a direct result of their synthesis @xcite .sources of disorder are introduced by crosslinking a liquid - crystalline polymer melt . in the simplest situation, the crosslinking takes place in the isotropic phase , in which case the local anisotropy axis of each crosslinking moiety is randomly oriented .once the polymer network is formed , the configuration of the crosslinks remains quenched , that is , it does not change with time and temperature .unless special precautions are taken during network fabrication , the low temperature ( ordered ) phase of nematic elastomers is always an equilibrium polydomain director texture @xcite .this is in marked contrast with a kinetic `` polydomain '' texture often referred to as schlieren texture @xcite , which is the consequence of nucleation and growth mismatch in a system undergoing the first - order transition @xcite .the equilibrium polydomain structure of nematic elastomers is reversible with changing temperature and is characterized by the uniform non - zero order parameter , but the highly non - uniform orientation of the principal axis of nematic director @xmath1 .correlations between directors decay rapidly and eventually vanish at distances much larger than @xmath2 , the correlation length or domain size .this is in agreement with the general result that quenched impurities destroy long - range order , first shown by larkin @xcite and then generalized by imry and ma @xcite .there is a full analogy with a corresponding dipolar system named `` random anisotropy magnets '' @xcite .in fact , all other ( e.g. smectic @xcite ) liquid crystal elastomers follow the same pattern of forming the equilibrium textures with a characteristic length scale often referred to as the domain size .this length scale in typical nematic elastomers is of the order of microns @xcite , therefore light passing through the sample is multiply scattered on birefringent domains with randomly oriented optical axis @xcite ( see @xcite for a brief review of experimental facts in this area ) . as a result sucha sample is completely transparent at high temperatures , but becomes opaque below its nematic - to - isotropic transition temperature @xmath3 .it should be noted that the polydomain texture is the thermodynamically stable low - temperature phase .applying an adequately strong aligning stress @xcite or increasing the temperature above @xmath3 @xcite destroys the polydomain texture , by aligning the local axis of each domain , or removing the optical contrast between them .however , once the stress is removed ( or the temperature lowered ) , the elastomer returns to its previous state and the average domain size is found to be reversible during this stress ( or temperature ) cycling .experimental investigations in nematic elastomers support the theoretical ideas of a continuous nematic - isotropic transition , rather than the discontinuous first - order transition expected by the landau - de gennes mean field theory based on the symmetry of quadrupolar ordering in three dimensions .the fact that polydomain nematic textures are optically opaque creates a practical problem when attempting to experimentally determine the local order parameter @xmath0 of the mesogenic units .it is impossible to use birefringence , dichroism or xray measurements , the methods that have made this task simple in aligned liquid - crystal systems .nuclear magnetic resonance ( nmr ) provides perhaps the only opportunity of order detection , by tracking the bias in orientational motion of selected chemical bonds and providing a unique @xmath0-signature even in the orientationally averaged case .it was used to measure @xmath4 of both a nematic polymer melt and its corresponding crosslinks network @xcite .the addition of crosslinks was shown to make the nematic transition smooth , as well as to slightly reduce order .an indirect alternative to nmr comes from applying an external field to align the domains .birefringence can then be measured , as long as the system has passed the critical point of the polydomain - monodomain transition @xcite .the order parameter can thus be found for a series of decreasing applied fields .although the zero - stress limit is not accessible , it can be extrapolated from the comparison of the other curves and , again , a continuous phase transition is seen @xcite .the continuous transition is also found in carefully synthesized monodomain nematic elastomers , where a second stage of the crosslinking takes place when the sample is stressed @xcite .plots of the macroscopic order parameter @xmath4 , obtained through birefringence @xcite , x - ray scattering @xcite and nmr measurements @xcite , show the same behavior .there are two possible explanations for such a deviation from the basic symmetry - based expectation of a first - order transition .inhomogeneities may cause local variations of the transition temperature inside the sample and lead to a substantial rounding of the transition , as discussed in @xcite .the other explanation considers stresses imprinted in the system during the second crosslinking stage , which add a @xmath5 term to the landau - de gennes expansion .obviously , for adequately large @xmath6 , the system would become supercritical and show a continuous transition .the analysis of the nmr spectra supports the supercritical scenario , although some degree of inhomogeneity was also found using the second moment of the spectra @xcite .another study of strain as a function of temperature over a range of applied tensile stress argues the opposite point @xcite .the blurring of the isotropic - nematic transition is attributed to the presence of heterogeneities , although the authors do not consider boundary effects that are bound to be present as discussed in @xcite .uchida @xcite has studied disordered polydomain nematic elastomers with emphasis on the role of nonlocal elastic interactions .he has shown that networks crosslinked in the isotropic phase lose their long - range orientational order due to the locally quenched random stresses , which were incorporated into the affine - deformation model of nematic rubber elasticity .simulation work was carried out to investigate the role of random bonds and random fields that might be present in elastomers @xcite , in both cases finding that the first - order isotropic - nematic transition to broaden into a smooth crossover . for random - field disorder ,the smooth crossover into an ordered state is also attributed to the long - range elastic interaction present in elastomers .a recent coarse - grained model for liquid - crystalline elastomers has also found that both homogeneous and inhomogeneous samples undergo a continuous isotropic - nematic transition @xcite . in this paperwe apply the traditional spin - glass techniques to investigate the characteristics of nematic phase transition in a system with quenched random anisotropy .the structure of the paper is as follows . in section@xmath7 we summarize a physical model of quenched disorder in nematic systems following @xcite , and introduce the replica hamiltonian .section @xmath8 applies the auxiliary fields to incorporate several constraints into this problem and obtains the effective mean - field free energy of disorder in the system . in section@xmath9 we investigate the stability of replica symmetry and discover the limits where our solutions are valid .finally , in section @xmath10 we obtain the final free energy renormalization in terms of the order parameter @xmath0 and disorder strength , and investigate the characteristics of the nematic phase transition in various situations .we conclude by discussing our results and comparing them with experiments .in the case of nematic elastomers crosslinked in the isotropic phase , the sources of quenched disorder are provided by the network crosslinks .almost independently of their specific chemical structure , the crosslinks contain anisotropic groups that locally provide easy anisotropy axes @xmath11 : it is favorable for the local director to align along @xmath11 in the vicinity of the crosslink because the anisotropic molecules in both the crosslinkers and in the nematic system interact , both sterically and via the long - range van de waals attraction .the local anisotropy axes of the crosslinks , as well as their distribution inside the sample , are quenched variables since the crosslinks can neither rotate nor move once the chemical synthesis of the elastomer is complete . although this has never been tested experimentally , there are two independent molecular models that estimate the energy of orientational confinement that a crosslink experiences from the surrounding network strands @xcite we follow earlier work @xcite in modelling the local coupling of the nematic order and the random field applied by the crosslinks . for a crosslink positioned at @xmath12 , with an anisotropy axis @xmath13 , an energy @xmath14 is raised due to the interaction with the local nematic order parameter @xmath15 , where @xmath16 is the coupling strength . employing a coarse - grained expression for the continuum density of crosslinks @xmath17 and substituting the full tensor expression for @xmath18, we get the random field energy : @xmath19 where the irrelevant constant @xmath20 has been dropped .although the random field energy is proposed specifically for a nematic elastomer , it is a general expression which is valid for all random - anisotropy systems with underlying quadrupolar symmetry . to obtain the full hamiltonian describing the nematic ordering , the gradient elasticity penalizing the fluctuations of the director fieldmust be also taken into account . in continuumelasticity the hamiltonian takes the form : @xmath21 \,=\,\int { d^3 \bm{r } \left [ \frac{k}{2}\,\left ( \nabla \bm{n}\right)^2\,-\,\gamma\,q\,\rho \left(\bm{k } \cdot \bm{n } \right)^2\right ] } , \label{hamiltonian}\end{aligned}\ ] ] where @xmath22 is the frank elastic constant in the one - constant approximation .a simple dimensional argument gives @xmath23 , where @xmath24 is the nematic coherence length , below which the meanings of the director @xmath25 and order parameter @xmath0 are ill - defined @xcite .both microscopic and phenomenological theories of nematic - isotropic transition give the elastic constant @xmath22 to scale as @xmath26 for @xmath27 .combining these two estimates , we take that for small @xmath0 the elastic constant is approximately given by @xmath28 .it is important to clarify that we are examining a homogeneous sample and as a result the magnitude of @xmath0 and the @xmath3 are uniform across the sample .there is a rich literature on the role inhomogeneities play in general first - order systems @xcite and more specifically in nematic elastomers @xcite .the important difference in our assumptions is that , although the director correlations in the polydomain nematic are short - range ( equilibrium spin - glass texture ) , the underlying nematic order parameter @xmath0 is homogeneous across the system .this assumption is based on the fact that the spin - glass like nematic textures are in fact homogeneous , in the sense that every element of the sample is equivalent to others : no ` domain walls ' ( unlike , for instance , during the stress - induced poly - monodomain transition @xcite when the domain walls localize ) .there are three established methods of dealing with quenched random fields in the replica method framework : one based on the functional renormalization group analysis , another using the gaussian variational method and the third using auxiliary fields .this paper employs the latter .we are interested in results that do not depend on the specific distributions of @xmath29 and @xmath11 because we can not control these distributions experimentally . in other wordswe are looking for the free energy averaged over the random distributions of the quenched variables @xmath29 and @xmath11 . crosslinking the sample above @xmath3makes the easy anisotropy axes point at random directions , with an isotropic probability of orientation @xmath30 .furthermore , the crosslinks are dispersed randomly in the sample with density @xmath31 .the probability that a particular distribution @xmath29 occurs is gaussian : @xmath32\,\sim\ , \exp\left[- \int{d^3r \frac{\left [ \rho(\bm{r})- \rho_0 \right]^2}{2\rho_0}}\right ] , \label{rho}\end{aligned}\ ] ] where @xmath33 is the mean density of crosslinks .it is not possible to directly average the logarithm of the partition function @xmath34 and obtain the exact free energy .so an alternative definition of a logarithm ( the limit : @xmath35 ) is used , allowing to perform the simpler average of the product @xmath36 .this way of dealing with quenched disorder is called the replica trick " , first introduced in the context of spin glasses by edwards and anderson @xcite . the expression for the free energy arising from disorderthen reads : @xmath37 \nonumber\end{aligned}\ ] ] where we now have @xmath38 identicalreplicas " of the system , labelled by the index @xmath24 .the aim of this work is to obtain @xmath39 as a function of the order parameter @xmath0 and add it to the landau - de gennes free energy to see how it influences the phase transition .a rough sketch of the averaging over disorder is given below .the density average over @xmath40 yields a random field term @xmath41,\ ] ] where ( here and throughout this paper ) @xmath42 . the reader s attention is drawn to the appearance of a second replica index @xmath43 due to the square of the @xmath44 .since the distribution of the orientations of the easy axes @xmath45 is assumed fully isotropic , symmetry arguments show that the average @xmath46 is proportional to @xmath47 .the constant of proportionality is equal to @xmath48 , where @xmath49 is the dimensionality of the unit vector @xmath11 .therefore @xmath50 since @xmath51 higher terms of the taylor expansion are smaller than the lowest order term ( @xmath52 ) and are subsequently dropped in the last line of eq .( [ kaverage ] ) .this approximate treatment retains a random field term which is overall fourth order in @xmath53 , as it was before the @xmath54-averaging , and it is most frequently met , and used , in molecular theories involving rotational diffusion . the second line of eq .( [ replicatrick ] ) is obtained after averaging over quenched disorder , sketched above , and provides the definition of the `` replica hamiltonian '' , which no longer depends on the quenched distributions of @xmath55 and @xmath56 , but instead couples different replicas of the system : @xmath57\equiv\sum_{a , b=1}^m \int d^3 \bm{r } \bigg\ { \frac{k}{2}\,\left ( \nabla \bm{n}_a\right)^2\delta_{ab}+ \nonumber\\-\,\frac{\gamma}{2 } \left[2\left(\bm{n}_a \cdot \bm{n}_b \right)^2\,+\,(\bm{n}_a \cdot \bm{n}_a)\ , ( \bm{n}_b \cdot \bm{n}_b)\right ] \bigg\ } , \label{replicahamiltonian}\end{aligned}\ ] ] where subscripts @xmath24 and @xmath43 are the replica indexes and @xmath38 is the number of replicas that will be set to zero at the end of the calculation .parameter @xmath58 , arising from completing the gaussian square between the eq .( [ rho ] ) and the random - field term in the eq .( [ hamiltonian ] ) , reflects the strength of the disorder and has a quadratic dependance on the order parameter : @xmath59 it is noted that all replicas are assumed to have equal disorder strength and equal magnitude of the local order parameter , i.e. @xmath60 and @xmath61 for all @xmath62 and @xmath43 . bearing in mind that the director is a unit vector we see that the term with @xmath63 in the random field part of the replica hamiltonian just adds an irrelevant constant to the expression .we drop this term and keep the relevant contribution @xmath64 that describes the coupling between different replicas .care must be taken to ensure that the director @xmath65 remains a unit vector .although this was assumed to be the case , it has not been implemented explicitly in the eq .( [ replicahamiltonian ] ) .one way to achieve this multiplies the partition function in eq .( [ replicatrick ] ) with the delta - function constraint @xmath66 where @xmath67 is an auxiliary field that allows the delta - function to be written in its exponential form .we proceed with a mean - field treatment of the auxiliary fields where a constant value for @xmath67 is assumed , independent of spacial coordinates .this approximation implies that the constraint @xmath68 is equally enforced across the whole sample .it is reasonable to expect this , given that the sample is spatially homogeneous .the same reasoning explains why @xmath67 has no dependance on the replica indexes : since its value is independent of the position of the crosslinks in the sample , it can not have different values for different replicas .the corresponding quadratic term @xmath69 has to be added to the replica hamiltonian in eq .( [ replicahamiltonian ] ) . to obtainthe disorder energy one must evaluate the statistical sum over all possible trajectories @xmath70 . the standard way to evaluate hamiltonians with quartic interactionsis to introduce an auxiliary field , here a tensor @xmath71 , which reduces the hamiltonian to bilinear order in @xmath70 . to employ the method all the quantities in @xmath72 $ ]must be dimensionless .since the integral over @xmath73 has the dimensions of volume , we move to a discrete summation over all points in space : @xmath74 the limits of the @xmath73-integration are @xmath75 , the size of the system , and the short - distance cutoff @xmath24 the nematic coherence length , below which the continuum representation is no longer applicable .the @xmath76 replica coupling term then becomes @xmath77 \label{fd_lambda } \\ & = & \int d\lambda_{ab}\exp\left\{\sum_r\left [ -\frac{\lambda_{ab}^2}{4\tilde{\gamma}}+\lambda_{ab}(\bm{n}_a \cdot \bm{n}_b ) \right]\right\ } , \nonumber\end{aligned}\ ] ] which involves the dimensionless constant @xmath78 .it is important to clarify the meaning of the @xmath79 term in eq .( [ fd_lambda ] ) : it is the square of the value @xmath71 rather than an element of the product of two matrices .furthermore , from now on we shall use a mean - field approximation , where it is assumed that @xmath71 has no @xmath73-dependance .this is an important limitation , but we believe it is reasonable as we are looking for homogeneous ordering in the system .summation over @xmath73 of the @xmath79 term yields @xmath80 , where @xmath81 is the number of discrete spacial points " and @xmath82 is the system s volume . moving to the corresponding discrete fourier space , the effective replica hamiltonian includes both auxiliary fields : @xmath83\,=\,\sum_{a , b}\bigg\ { -i \phi\ , \delta_{ab}\,+\ , n\frac { \lambda_{ab}^2}{4\tilde\gamma}+ \nonumber \\ & + & \ , \sum_q \left [ \left(\frac{\tilde{k}q^2}{2}+i\phi\right ) \delta_{ab } \,-\,\lambda_{ab}\right ] \left(\bm{n}_a \cdot \bm{n}_b \right)\bigg\ } , \label{heff}\end{aligned}\ ] ] with the dimensionless elasticity constant @xmath84 . the discrete sum over @xmath85is related to the integral via @xmath86 . as mentioned above ,the conversion from the integral to the discrete sum is essential so that all the quantities in @xmath87 ( such as @xmath88 , @xmath89 , @xmath90 and @xmath91 ) remain dimensionless and the logarithm of their sum can be evaluated correctly . to be able to deduce the disorder energy as a function of the order parameter, we make explicit the dependence of @xmath88 and @xmath92 on the magnitude of the order parameter @xmath0 : @xmath93 where @xmath94 is the frank elastic constant deep inside the nematic phase .it is worth noting that @xmath88 is always significantly smaller than one .the distance between crosslinks , @xmath95 , which can be deduced from the crosslink density @xmath96 , is found around @xmath97 nm in nematic elastomers @xcite .a typical coherence length for nematics is @xmath98 nm , hence @xmath99 or less .the coupling of disorder to the nematic director is deduced from the size of the domains in ref @xcite and is found to be @xmath100. therefore this crude estimate gives @xmath101 for the nematic elastomers studied in the literature . to be able to advance in the calculation , a matrix form of the auxiliary field @xmath71 has to be postulated .a reasonable starting point is to assume that it has the simplest possible form , where its elements have a constant value independent of @xmath24 and @xmath43 : @xmath102 where @xmath103 is the matrix with all its elements equal to one and @xmath104 is the identity matrix .this scheme is frequently encountered in the literature @xcite and is called the replica symmetry " limit .it is important to clarify that we are free to choose any form we like for @xmath71 , and that we will later come back to this choice and check whether it is appropriate or not , in section [ section_stability ] .there is an important reason why the diagonal elements of the auxiliary field @xmath71 are chosen to be zero in eq .( [ rslambda ] ) .had @xmath105 been our choice , then the summation over @xmath24 and @xmath43 of @xmath79 would have given @xmath106 .the quadratic dependence on @xmath38 would have meant that , after differentiating with respect to @xmath38 and setting @xmath107 , this term would be equal to zero .clearly this is not acceptable since the introduction of the auxiliary field in eq .( [ fd_lambda ] ) requires a non - zero @xmath79 quadratic term . substituting the replica - symmetric ansatz into the effective hamiltonian of eq .( [ heff ] ) , we find : @xmath108 where the propagator of the @xmath109 replica coupling is given by : @xmath110 a consequence of replica symmetry is that @xmath111 only involves matrices @xmath104 and @xmath103 , and as a result its logarithm is easily obtained : @xmath112 the path integral over configurations @xmath70 with the statistical weight determined by the effective hamiltonian ( [ h2 ] ) is gaussian and gives @xmath113 for each of the three vector components of @xmath1 . as a result the disorder free energy is given by : @xmath114 from the three ( identical ) path integrals for the components of @xmath1 .the aim of this section is to determine the particular values of the auxiliary fields ( @xmath115 and @xmath116 ) that make the disorder energy of eq .( [ fd1 ] ) a minimum . to treat the problem properly, one would have to evaluate the integrals over @xmath71 and @xmath67 , which is analytically challenging . the standard way to bypass this difficulty ,is to employ the saddle - point approximation based on the simple observation that the exponentially most significant contribution in eq .( [ fd1 ] ) will occur when the exponent is a maximum .therefore @xmath117\bigg\}\bigg|_{m \rightarrow 0 } , \label{fd2}\end{aligned}\ ] ] where @xmath118 represents the minimum of a function with respect to variations in @xmath119 and @xmath120 . after substituting the trace of the logarithm from the eq .( [ lng ] ) , we differentiate with respect to @xmath38 and then set @xmath38 equal to zero .the disorder energy then takes the form : @xmath121 \bigg\ } , \nonumber\end{aligned}\ ] ] and we are left with the task of finding the stationary point @xmath122 . from now on we move to the continuum limit of space , where the discrete sum is replaced by @xmath123 and the coherence length @xmath24 is taken to zero limit .the auxiliary field @xmath67 ensures that the nematic director is a unit vector .this constraint should of course be enforced whether the disorder is present or not .in fact , there is no physical reason why the inclusion of disorder should significantly alter this constraint .hence , as a first approximation , we look for the optimum @xmath124 when there is no disorder in the system . setting both @xmath58 and @xmath119 equal zero in the eq .( [ fd3 ] ) and differentiating with respect to @xmath120 we obtain : @xmath125 . \nonumber\end{aligned}\ ] ] in the continuous limit of @xmath126 and the arctangent is equal to @xmath127 . this equation can be re - written as @xmath128 clearly the @xmath129 term is negligible compared to @xmath130 , and can be therefore neglected .another factor that supports this omission is that we want to examine what happens close to the phase transition , where @xmath131 . in this limit , the elastic constant is known to vanish ( @xmath132 ) and the term with @xmath133 dominates in the bracket . assuming that the inclusion of disorder has only a minor effect on @xmath67 , the value we will use from now on is : @xmath134 which tends to infinity in the continuum limit of space .it is interesting to note that very close to the transition the constraint relaxes since @xmath135 .this is not surprising since the meaning of the director itself becomes ill - defined as we approach the transition point .the full calculation to obtain @xmath136 in a system with disorder is possible .however , using a disorder - dependent @xmath136 in eq .( [ fd3 ] ) makes it analytically impossible to solve @xmath137 and determine the optimum @xmath138 .even a perturbative approach : @xmath139 does not help matters . to overcomethis difficulty the @xmath140 saddle point should be found numerically , searching for a global energy minimum in the @xmath141 space .the important drawback of the latter is that we will then be unable to determine the analytical form of @xmath142 and therefore the final @xmath143 correction to the landau free energy of the phase transition .we proceed to determine the value of the auxiliary field @xmath138 that minimizes the energy in the replica symmetric approximation . differentiating the right - hand side of eq .( [ fd3 ] ) and demanding it to be zero we find : @xmath144 integration over momentum space gives the stationary condition on the auxiliary field : @xmath145 \nonumber\end{aligned}\ ] ] in the continuous limit of @xmath146 the first term vanishes and the arctangent is equal to @xmath127 .equation ( [ lambda1 ] ) has only one real solution .unfortunately , its full expression is too long and cumbersome to appear here explicitly ; instead we demonstrate its behavior in two limits .expanded in powers of @xmath147 ( weak disorder ) it takes the form @xmath148 it is a reassuring property that @xmath138 vanishes as @xmath149 .critical to our work is the behavior of @xmath138 as the order parameter @xmath0 tends to zero . both @xmath58 ( from its definition ) and @xmath22 ( for small @xmath0 )are quadratic functions of @xmath0 .therefore the leading term in the series expansion of small @xmath0 is @xmath150 the scaling @xmath151 is thus obtained , showing that @xmath138 diverges as the transition is approached , that is , even a weak disorder becomes relevant near the @xmath152 point . to find the final disorder energy the values of fields @xmath116 and@xmath138 are put back in the eq .( [ fd3 ] ) . performing the @xmath85-integrations in the continuum limitwe obtain : @xmath153 \label{fd}\end{aligned}\ ] ] the energy we are interested in arises from disorder and we can safely ignore terms that are still present when @xmath154 . at diminishing order parameter , the leading term of eq .( [ fd ] ) takes the form : @xmath155 which clearly diverges as @xmath131 .this divergence of the disorder free energy implies that the isotropic phase can never be reached and , as we shall see in greater detail in section [ section_analysis ] .this in turn leads to the rounding of the nematic - isotropic phase transition . before all thisis discussed , we examine if the replica - symmetric form of @xmath71 in eq .( [ rslambda ] ) was an appropriate choice , which is far from obvious .a necessary condition for the replica - symmetric solution to be applicable is that the disorder energy is stable for infinitesimal variations of that solution .the auxiliary field @xmath138 gives an energy extremum ; whether this extremum is a maximum or a minimum is determined by a stability analysis following the work of de almeida and thouless in spin glasses @xcite .we start by allowing the matrix of the auxiliary field @xmath71 to deviate from its replica - symmetric form : @xmath156 with @xmath157 the arbitrary infinitesimal deviation from replica symmetry .the disorder free energy will be expanded to second order in @xmath158 : @xmath159 where @xmath160 represents the energy of the replica - symmetric case ( @xmath161 ) , given by the eq .( [ fd ] ) .the second - order term in @xmath157 involves the fourth rank tensor of coefficients @xmath162 , which is called the hessian .it plays an important role in this analysis , because the replica - symmetric solution is only stable as long as @xmath162 is positive definite , or equivalently only as long as its eigenvalues are non - negative .of course there is no term linear in @xmath157 because its coefficient is given by the right hand side of eq .( [ lambda1 ] ) and is therefore equal to zero . we proceed to find the hessian of this model .expanding the first term of the disorder energy in powers of @xmath157 gives : @xmath163 where @xmath164 .hence the contribution to the hessian from this term is @xmath165 .the other contribution comes from the trace of the logarithm .the propagator of eq .( [ g^-1 ] ) can be written as @xmath166 , therefore allowing @xmath71 to vary gives the following form to the propagator @xmath167 where the replica - symmetric @xmath168 is given in eq .( [ g^-1 ] ) .therefore , @xmath169 \ \label{stabremark1}\end{aligned}\ ] ] where summation over the dummy indexes @xmath170 is implicit and @xmath171 is the inverse of eq .( [ g^-1 ] ) . in the case of two commuting matrices : @xmath172 .it turns out that the eigenvectors of the hessian ( @xmath157 ) indeed commute with matrix @xmath173 .this is largely because the latter is a combination of two very simple matrices @xmath103 and @xmath104 .breaking up the product under the logarithm : @xmath174 and expanding the second term in a taylor series , we obtain : @xmath175 there are three identical traces to be considered , one for each component of the nematic director .taking into account the @xmath176 factor in front of the hessian , this contribution is @xmath177 and the overall hessian takes the form : @xmath178 this hessian has three distinct values of its many matrix elements , depending on how its many indexes are common to the pairs @xmath179 and @xmath180 .these values are @xmath181 , @xmath182 and @xmath183 .the eigenvalues of such a 4-rank tensor have been computed by de almeida and thouless in their classical work on replica symmetry breaking in spin glasses @xcite . for @xmath184only two distinct eigenvalues exist : @xmath185 and it is straightforward to obtain them by inserting the appropriate forms of @xmath186 into the expression of the hessian of eq .( [ hessian ] ) .the first eigenvalue @xmath187dq \nonumber \\ & = & \frac{n}{2\tilde{\gamma}}-\frac{3(2i\phi+\lambda)v } { 8\sqrt{2}\pi\tilde{k}^{3/2}(i\phi+\lambda)^{3/2 } } \label{eigenvalue1}\end{aligned}\ ] ] is degenerate and corresponds to two eigenvectors @xmath188 and @xmath189 .the first , @xmath188 , is symmetric under interchange of indexes ( @xmath190 for all @xmath191 ) and determines whether the replica - symmetric fixed point of eq .( [ lambda1 ] ) is stable or not . in other words ,if @xmath192 the replica - symmetric solution corresponds to an energy maximum and has no physical relevance .the remaining two eigenvectors check the general stability of the replica - symmetric scheme .contrary to @xmath188 , the second eigenvector @xmath189 , corresponding to the degenerate @xmath193 , is symmetric under interchange of all but one index ( @xmath194 for @xmath24 or @xmath43 = @xmath195 and @xmath196 otherwise ) .this eigenvector is not symmetric and as a result a negative @xmath193 also means that replica symmetry must be broken to determine the correct @xmath71 .the second eigenvalue is non - degenerate and is given by : @xmath197 and its corresponding eigenvector , called @xmath198 , also does not have a symmetric form ( @xmath199 for @xmath200 and @xmath201 , @xmath202 for @xmath200 or @xmath203 and @xmath204 , @xmath205 otherwise ) . similarly to the previous case ,if @xmath206 then the replica symmetric solution breaks down and different forms of @xmath71 must be sought .accordingly , a sufficient condition for the replica - symmetric solution to be stable is that all the eigenvalues of the hessian remain positive .both @xmath193 and @xmath207 show a similar behavior as functions of disorder strength @xmath16 , which was defined in the text above eq .( [ rfenergy ] ) , and the order parameter @xmath0 . substituting the optimal @xmath116 and @xmath138, we find that when @xmath0 does not tend to zero , both eigenvalues remain positive , especially for weak disorder , @xmath208 . as the continuous phase transitionis approached and the order parameter diminishes , the three parameters appearing in eq .( [ eigenvalue2 ] ) become simple functions of @xmath0 : @xmath209 , @xmath210 and @xmath211 . hence ,the eigenvalues scale as @xmath212 and , therefore , become negative at some point @xmath27 .this means that for @xmath213 the replica - symmetric solution of the auxiliary field eq .( [ lambdaq ] ) and the resulting expression for the disorder energy , eq .( [ fdq ] ) , correspond to an energy maximum and should not be used . to find the exact point at which replica symmetry becomes unstable we note that it is the second ( non - degenerate ) eigenvalue , @xmath207 that becomes negative first as @xmath131 . substituting the parameters @xmath58 , @xmath22 and @xmath138 we find this crossover value for stability : @xmath214 where the relation @xmath215 has been used , with @xmath24 the short - length cutoff .bearing in mind that @xmath216 and that @xmath217 we find @xmath218 . to be consistent with previous calculations the continuum limit of space is employed and @xmath219. therefore the threshold value below which replica symmetry breaks tends to zero and the eq .( [ lambdaq ] ) is always valid in the continuum limit .another way to see this is by examining the explicit dependance of @xmath207 on the cut - off length - scale @xmath24 .the first term @xmath220 scales as @xmath221 whereas the second term scales as @xmath222 .therefore when @xmath24 is taken to zero the first term dominates and the eigenvalue remains positive . to complete the stability analysis , we note that the other eigenvalue also becomes negative for smaller values of @xmath0 when @xmath223 , but this is irrelevant since the instability of @xmath207 occurs first .the aim of this paper is to discuss how quenched orientational disorder affects the phase transition of nematic systems .the total free energy density is a combination of the landau - de gennes expansion of the order parameter plus the disorder part ( @xmath224 ) obtained in eq .( [ fd ] ) . to get the total energy as an expansion of the order parameter only the leading order contribution of @xmath39 in eq .( [ fdq ] ) is considered : @xmath225 where @xmath226 and @xmath227 , @xmath228 , @xmath229 and @xmath230 are the usual landau - de gennes parameters .one may note the explicit dependence of eq .( [ d ] ) on the short - length cutoff @xmath24 and be concerned about the sustainability of the continuum limit @xmath219 .in fact , because @xmath231 and @xmath232 , the powers of @xmath24 cancel and the free energy does not depend on this cutoff parameter , which is a reassuring test of consistency of our theory . the easiest way to illustrate the consequence of the free energy renormalization is to plot the equilibrium values of the order parameter @xmath0 .the values chosen for the landau phenomenological constants are : @xmath233 . a detailed description on how they are obtained by analyzing experimental data of ref .@xcite is given in the appendix .we consider these values to be indicative only since different nematic materials will certainly have large variations in these values .plotting the equilibrium order parameter against reduced temperature , fig .[ plotqoft ] , we can compare @xmath4 for the disorder strength increasing from @xmath234 in ( a ) up to @xmath235 in ( d ) .as mentioned in section [ section_3 ] a typical polydomain elastomer should have @xmath236 and therefore the model predicts a supercritical behavior in agreement with many experiments .note that in contrast to this result , the quenched orientational disorder was shown not to alter the continuous nematic phase transition in thin films , whose director is confined in the @xmath237-plane @xcite .the inclusion of disorder has a profound effect on the phase behavior of 3d systems whose pure versions undergo a first order phase transition .the discontinuous jump of the order parameter at the nematic transition becomes smaller as the strength of the disorder @xmath238 increases and eventually disappears altogether above a critical value making the phase transition continuous .the change in behavior is explained by the simple fact that the energy terms arising from disorder scale as negative powers of @xmath0 . as a direct result, the energy of the system increases as the transition approaches and the zero order parameter phase is never reached . to see exactly how this happens consider plots of @xmath239 of a subcritical and a supercritical system . for a subcritical system , where disorder is weak, the jump in @xmath4 is still present , albeit smaller than in the original transition of the pure system .because parameter @xmath238 is small , the disorder part of the energy becomes significant only for small @xmath0 , where it diverges .the appearance of the jump has the same origins as in the classical landau - de gennes theory .[ fig jump ] shows plots of the energy density against order parameter for four different temperatures around @xmath3 . at the lowest temperature ( @xmath240 ) the singleminimum determines the equilibrium value of @xmath0 .as the temperature increases the high-@xmath0 minimum moves to a slightly smaller value of @xmath0 and -more importantly- another low - q " minimum appears as a result of disorder . at @xmath241the high-@xmath0 is still the global minimum , but at the critical temperature ( @xmath242 ) the two minima have the same energy value .this means that two distinct phases , one with @xmath243 and the other @xmath244 , coexist .once this temperature is passed ( @xmath245 ) the low-@xmath0 minimum determines the system s order parameter .the crucial difference between these plots and the classical landau - de gennes theory is that the low-@xmath0 minimum in the latter is always placed at @xmath246 .since the disorder energy diverges at zero @xmath0 , this minimum is pushed at positive values of @xmath0 in systems with quenched disorder .the dotted and dashed lines show the landau and the disorder energy for the same temperature @xmath242 , respectively . around the high-@xmath0 minimumthe dotted line has the same shape as the actual energy ; apart from a constant shift to lower energy they are exactly equal .therefore this minimum is a result of the competition of the @xmath247 and @xmath248 energy terms of the landau expansion , which dominate at large @xmath0 .the position of the low-@xmath0 minimum is influenced by disorder .this minimum is a balance of the divergent @xmath39 term and @xmath249 . for temperatures well above @xmath3 the minimumis located at very small order parameter and the only relevant terms of eq .( [ energy ] ) are @xmath250 and @xmath249 . in a supercritical system disorderis stronger ( large @xmath238 ) and therefore its effect on the energy is more prominent . as a resultthe effect of @xmath39 is relevant for all the values of @xmath0 , not just in the small order parameter region as in the previous case .[ fig nojump ] shows the relevant energy plots .crucially there is only one minimum at any given temperature .its position shifts to smaller order parameter as temperature increases , but the phase transition is continuous . in comparison with fig .[ fig jump ] , we can say that the low-@xmath0 minimum has broadened " and absorbed " the high-@xmath0 minimum . another difference with the small @xmath238 case is that , because @xmath39 has larger magnitude , the low order @xmath249 term of the landau expansion does not influence the position of the minimum .the dotted lines in fig .[ fig nojump ] are drawn for the same temperature as in the previous case of fig .[ fig jump ] , but now this temperature is labeled as @xmath240 . in this supercritical systemthe energy minimum occurs at approximately the same value of order parameter as the high-@xmath0 minimum of the landau degennes expansion ( thin - doted line ) . in a pure systemthis would be a metastable state because there would exist a global minimum at @xmath246 .the natural question to ask next is what is the critical point at which the jump in @xmath4 disappears completely . at this pointthe free energy must have the merging of all its minima and maxima .solving for the first , second and third derivatives being zero provides three equations for the unknown critical parameters @xmath251 , @xmath252 and @xmath253 and the critical point is then given by : @xmath254 it is worth comparing this with the classical case of an applied field , which adds a @xmath255 term in the landau expansion , where the critical @xmath0 is slightly smaller and equal to @xmath256 . in the analysis of the replica symmetry stability in section [ section_stability ] we found that the above description of nematic systems breaks down for small order parameter when @xmath257 , see eq .( [ qstab ] ) .the threshold value of @xmath258 is proportional to the cube of the cut - off length @xmath24 . in order to be consistent with the line adopted in previous calculations , the continuum limit of spacewas employed and @xmath24 was taken to zero . hence @xmath258 was also taken to be zero . however , in reality the length scale @xmath24 is in fact the nematic coherence length , because below this size we can not write frank elasticity and there is no meaning to order parameter @xmath0 , or director @xmath1 .keeping @xmath24 non - zero , and assigning it the value of @xmath259 nm which is usually associated with liquid crystals , makes @xmath258 also finite .nevertheless substituting the value of @xmath260 and @xmath261 gives @xmath262 .this is an extremely small value for the order parameter and , as we see from fig [ plotqoft ] , it would only be acquired at temperatures well above the experimental range .hence , even if @xmath263 , the window of replica symmetry stability is wide enough to describe realistic nematic systems .this paper examines how the inclusion of randomly quenched orientational disorder leads to the rounding of the nematic - isotropic phase transition in three dimensions .the coupling between impurities and the local order parameter pins some mesogenic molecules and does not allow the sample to have a uniform director field @xmath264 .after quenched disorder has been averaged over using the replica method , a replica - symmetric auxiliary field is used to obtain the free energy arising from disorder .the disorder energy adds a @xmath265 term to the landau - de gennes expansion which diverges for diminishing order parameter @xmath0 . as a resultthe isotropic phase is never reached and , for sufficiently strong disorder , the phase transition becomes continuous .this is in accordance with many experiments on nematic elastomers that also show a smooth transition rather than a discontinuous as predicted by the classical landau - de gennes theory .an earlier study on xy nematics , whose director is confined to a plane , showed that the quenched disorder does not affect their continuous phase behavior @xcite .a stability analysis shows that the replica - symmetric solution we have employed does fail at small values of the order parameter , at @xmath266 .however , for realistic values of physical parameters we estimate the order of @xmath267 in nematic elastomers .a supercritical system acquires such low values of @xmath0 at temperatures well above the elastomers melting point and therefore the window of replica symmetry stability adequately describes the 3d nematic system .we would like to thank isaac prez castillo , david sherrington , paolo biscari and raphael blumenfeld for their feedback .this work has been supported by the leventis foundation , the cambridge european trust and the epsrc tcm / c3 portfolio grants . in order to determine the three phenomenological parameters @xmath268 , @xmath228 and @xmath229 of the landau - de gennes theory , defined in eq .( [ energy ] ) , we need three independent measurements .these are provided by a nmr experiment measuring the order parameter as a function of temperature of a polymer melt and its corresponding crosslinked network ( a polydomain nematic elastomer ) @xcite . as expected from the landau - de gennes theory, the melt shows a first order transition with the discontinuous jump in order parameter being approximately @xmath269 , which is smaller than @xmath270 , the value usually associated with ordinary liquid crystals @xcite .the theoretical prediction gives this jump to be equal to @xmath271 @xcite and substituting the experimental measurement we find @xmath272 .the second measurement is the width of the temperature hysteresis which is @xmath273 in the polymer melt .theory predicts @xmath274 .combining this with @xmath275 we find @xmath276 and @xmath277 . to get a value of @xmath268 a third measurementis required .a striking difference between the two @xmath4 plots , for a nematic polymer melt and its crosslinked elastomer version , is that crosslinking has reduced the overall order at temperatures below @xmath3 . for elastomers crosslinked in the isotropic phase, the energy addition arising from nematic rubber elasticity adds a fourth order term in the landau expansion @xcite : @xmath278 where @xmath279 is the rubber modulus and @xmath280 accounts for the microscopic details of an elastomer . for a freely joined polymer @xmath281 @xcite , but a side - chained polymers have @xmath280 ranging between @xmath282 and @xmath283 @xcite .let us take an intermediate case where @xmath284 .when this term is added to the landau - de gennes expansion , the fourth - order coefficient ( which is @xmath285 for the polymer melt ) now becomes larger @xmath286 .hence the transition temperature , given by @xmath287 , decreases since the renormalized @xmath229 increases .the shift in this transition temperature between the melt and the corresponding elastomer is : @xmath288 and it provides the third relation that allows to determine @xmath268 .an estimate of @xmath289 is possible in ref @xcite : @xmath3 is easily identified in the polymer melt , but it is not clear what it means in the disordered nematic elastomer with a continuous transition .it can be loosely defined as the temperature where @xmath290 .this then makes @xmath291 . substituting this back to eq .( [ delta tni ] ) with @xmath284 , we obtain @xmath292 .a typical nematic elastomer has elastic modulus of the order of @xmath293pa . putting everything together ,the phenomenological constants of a nematic elastomer are : @xmath294 these values are crude estimations , only given here to illustrate the effect of disorder in our model . however , it is comforting that , although obtained from a different set of experimental measurements , these values are quite close to the ones reported in literature @xcite .

<S> nematic elastomers do not show the discontinuous , first - order , phase transition that the landau - de gennes mean field theory predicts for a quadrupolar ordering in 3d . </S><S> we attribute this behavior to the presence of network crosslinks , which act as sources of quenched orientational disorder . </S><S> we show that the addition of weak random anisotropy results in a singular renormalization of the landau - de gennes expression , adding an energy term proportional to the inverse quartic power of order parameter @xmath0 . </S><S> this reduces the first - order discontinuity in @xmath0 . for sufficiently high disorder strength </S><S> the jump disappears altogether and the phase transition becomes continuous , in some ways resembling the supercritical transitions in external field . </S><S> + + to be published on _ phys . </S><S> rev . </S><S> e _ </S>

hep-ph9401238

the superstring offers the exciting possibility of predicting all the parameters of the standard model in terms of a single parameter , the string tension . however in order to realise the full predictive power of the superstring it is necessary to determine the origin and effects of supersymmetry breaking .only after susy is broken are the vacuum expectation values ( vevs ) of moduli determined and these determine the couplings of the effective low energy theory . also susy breaking must be responsible for the splitting of supermultiplets allowing for the superpartners to be heavier than their standard model partners .in supergravity theories soft susy breaking terms may be added to the lagrangian without introducing new divergences and thus need not spoil the supersymmetric solution to the hierarchy problem . if susy breaking is dominated by the f term of a specific field the soft terms may be related in a model independent way if supersymmetry breaking occurs in a hidden sector with coupling to the visible sector via gravitational effects only@xcite . in this letterwe will investigate the predictions for the moduli and soft terms in superstring theories in which the origin of susy breaking is through a gaugino condensate in the hidden sector .gaugino condensation offers a very plausible origin for susy breaking for it is very reasonable to expect such a condensate to form at a scale between the planck scale and the electroweak breaking scale if the hidden sector gauge group has a ( running ) coupling which becomes large somewhere in this domain .non - perturbative studies in effective supergravity theories resulting from orbifold compactification schemes suggest the dynamics of the strongly coupled gauge sector is such that the gaugino condensate will form and trigger supersymmetry breaking . the advantage of gaugino condensation as the underlying supersymmetry breaking trigger is twofold . in the first place it is generic in 4d - superstring theories to have an hidden sector with a gauge group given by @xmath0 or smaller . for such asymptotically free gauge theoriesthe coupling usually becomes strong at low scales so gaugino condensation may be expected - thus there is no need to postulate an additional source of supersymmetry breaking .secondly the scale of gaugino condensation and the associated supersymmetry breaking scale are dynamically determined and thus the mechanism offers an explanation for the magnitude of the hierarchy of masses scales .thus , when combined with the radiative breaking of the electroweak symmetry , the resultant theory not only protects the electroweak and supersymmetry breaking scales from receiving large radiative corrections proportional to the unification or planck scale ( the usual hierarchy problem ) but also predicts the supersymmetry breaking and electroweak breaking scales in terms of the planck scale and the multiplet content of the theory .in addition , after supersymmetry breaking , the moduli of the theory are fixed giving a prediction for the value of the gauge couplings and the yukawa couplings . in order to realise the full predictive power of the 4d- superstring theory ,it is necessary to compute the non- perturbative effects leading to gaugino condensation .the complete solution is clearly beyond our present - day technology so we are forced to employ approximation methods . in this paperwe will consider the implications for the soft susy breaking parameters in the case that gaugino condensation is the source of supersymmetry breaking .we parameterise the strong binding effects in the hidden sector by a simple effective ( locally supersymmetric ) 4-gaugino coupling .we then are able to use nambu - jona - lasinio ( n - j - l ) techniques @xcite to obtain non- perturbative information about the gaugino binding .this shows that a reasonable pattern of supersymmetry breaking is possible with a single gaugino condensate .we generalize the discussion of reference @xcite to consider an orbifold model which includes different moduli fields and we determine the tree level potential and the one - loop radiative corrections due to strong gaugino binding .we show that minimization of the potential leads to vacuum expectation values ( v.e.v.s ) for the moduli which either take dual invariant values or have a common value related to the v.e.v . of the dilaton .this naturally allows for squeezed orbifolds which were found to be better candidates for minimal string unification. we also include chiral matter fields in the visible sector and determine the soft susy breaking terms . unlike results obtainedpreviously we find that the radiative corrections stabilise the potential for vanishing v.e.v.s of the matter fields .we also find values for the soft parameters which lie outside the favoured range of `` model '' independent parameterisations of the soft terms due to our inclusion of large non - perturbative effects beyond string tree level in the effective potential .of course , due to the simplicity of our model for strong binding , our results should be considered as only indicative of those to be expected in a complete gauge theory .however , they do show the importance of including the full nonperturbative effects of gaugino binding before extracting predictions for susy breaking phenomena .we start with the effective theory following from a 4-d superstring orbifold model .it is specified by the kahler potential @xmath1 and the gauge kinetic function @xmath2 given by k =- ln(s+|s+2_ik_a_gs^ilnt_ri)- _iln(t+|t)_i + _ i k^i_i(t,|t)|_i|^2 , w = w_0+w_m , and f_0=s+2_i(b^i_a - k_a^i_gs)ln ( ( t_i)^2 ) where @xmath3 where @xmath4 is the quadratic casimir operator of the gauge group @xmath5 , @xmath6 is the level of the corresponding kac - moody lalgebra for the gauge group @xmath5 and @xmath7 is the green - schwarz term needed to cancel the gauge independent part of the duality anomaly .@xmath8 in eq.([eq : w ] ) is the scalar potential due to the gaugino condensate ( see below ) and @xmath9 the matter superpotential .the subscript , 0 , on the gauge kinetic function @xmath2 indicates that it is given at the string scale and the form of eq.([a8 ] ) includes the contribution from the massive kaluza - klein modes .the normalization of the kinetic term for the chiral matter superfields is given by @xmath10 which is a polynomial function of the moduli fields , k^i_i=_ja_it_rj^n_ij where @xmath11 a constant ( in most cases @xmath12 ) and @xmath13 is the modular weight of the superfield @xmath14 with respect to the real part , @xmath15 , of the moduli @xmath16 .eq.([a7 ] ) was derived for @xmath17 .note that the kahler potential includes the green - schwarz term and the dilaton transforms under duality in a non - trivial way ( @xmath18 ) so that @xmath19 is duality invariant .the form of this effective theory is fixed by the requirement that it should correspond to the strongly coupled theory in the hidden sector .we have suggested a simple approximation to this structure which , we hope , captures the essentials of the strong binding effects leading to gaugino condensate .the basis of this approximation is to start with a ( n=1 locally supersymmetric ) four fermion description of the strong hidden sector interaction and to use nambu - jona - lasinio techniques to compute the non- perturbative effects responsible for gaugino condensation .the model is defined by w_0=_i^-2(t_i ) f = f_0+ln(/ ) where @xmath20 is an auxiliary chiral superfield and @xmath21 .duality invariance fixes the mass parameter @xmath22 in eq.([a12 ] ) to be @xmath23 with @xmath24 duality invariant .we choose @xmath25 so that the parameter @xmath22 becomes , at the field theoretical limit , @xmath26 , with @xmath27 the compactification scale . since @xmath22 is non - holomorphic , consistency with supersymmetry requires to interpret @xmath22 as wave function renormalization of the dilaton field .in such a case @xmath22 would appear in the kahler potential and not in the kinetic function . from the classical equation of motion the scalar component of the auxiliary field @xmath20is given in terms of the gaugino bilinear by = |_r _ l while the fermion component is given in terms of the gaugino and chiral fermions .for the choice of @xmath32 and @xmath2 given in eq.([a11 ] ) and eq.([a12 ] ) the f - terms for the @xmath30 and @xmath31 fields are h_s =- e^g/2g_s+ f_s |_l_r = e^k/2w_0 and h_t_i = e^kw_0where @xmath33 .the tree level scalar potential is then v_0=m^2_3/2 b_0 with b_0= and @xmath34 the eisenstein modular form with weight 1/2 .the gravitino mass is given by m^2_3/2=e^k_i|(t_i)|^-4||^2 . clearly susy will be broken and a non - zero gravitino mass will result only if there is a non - vanishing v.e.v for @xmath35 generating a non - zero @xmath36 ( cf .eq.([a18 ] ) ) . assuming this to be the case and solving eq.([a12 ] ) gives a parameterisation of gaugino condensation equivalent to that derived by other methods @xcite- . to this extent the form of eq.([a20 ] ) is just a re - parameterisation of the usual model of gaugino condensation .where we depart from this analysis is our inclusion of radiative corrections in the effective potential .summing these effects to all orders shows that the gaugino condensate dynamically forms .using eq.([a13 ] ) the scalar potential may be written in terms of the gaugino field v_0=||_r_l|^2 where the factor of @xmath37 in the denominator in eq.([e29 ] ) appears because we have rescaled the gaugino fields appearing in this equation to have canonical kinetic terms .thus we see that the choice of @xmath32 and @xmath2 in eqs.([a11 ] ) and ( [ a12 ] ) leads to a four - fermion interaction with strength proportional to @xmath38 . in a pure gauge theorythe running gauge coupling is related to @xmath2 by @xmath39 with = + b_aln()+_a where @xmath40 is the level of the corresponding kac - moody algebra for the gauge group @xmath41 , @xmath42 is the n=1 @xmath43-function coefficient and @xmath44 the number of chiral fields in a representation @xmath45 .@xmath46 is the renormalization scale below which the coupling constants begins to run m^2_a=_i(t_ri)^_aim_s^2 where @xmath47 and the constants @xmath48 is model and gauge dependent ._ ai= in the case of a single overall ( 1,1 ) moduli @xmath49 , @xmath50 , with @xmath51 one obtains the `` naive '' field theoretical expression for @xmath52 .the existence of an infinite number of massive states ( kaluza and winding states ) , above the string scale , give rise to string threshold contribution @xmath53 which are relevant to the determination of the coupling constant at the string scale ( cf .eq.([a8 ] ) ) . in the effective four fermion theory defined by eqs.([a11 ] ) to ( [ e29 ] ) @xmath38 sets the scale for the strength of the interaction ( cf .eq.([e29 ] ) ) .if the theory is to model the strong binding effects due to a gauge interaction it clearly should be related to the gauge coupling @xmath54 .thus @xmath38 does indeed run like @xmath55 .however this running gives rise to an unphysical singularity in @xmath38 related to the landau singularity which we will choose to regulate when applying eq.([a13 ] ) .below the scale at which the running coupling becomes large the hidden sector states are confined and have a mass of @xmath56 ) .thus we expect the 4 fermion effective interaction to be pointlike at scales below @xmath57 and so we cutoff the growth implied by eq.([e29 ] ) when the strength of the interaction reaches @xmath58 .the tree level potential for @xmath35 , given by eqs.([a20]),([a21 ] ) and ( [ a22 ] ) , has no stable solution .this is consistent with the observation @xcite that gaugino condensation as usually parameterised does not occur in models with a single hidden sector gauge group factor .however we have argued it is essential to go beyond tree level to include non - perturbative effects in the effective potential which may allow for a non - trivial minimum even in the simple case of a single hidden sector gauge group . this non - perturbative sum ( equivalent to the njl sum )is readily obtained simply by computing the one loop correction to v. if these destabilise the potential the resultant minimum will correspond to a cancelation of tree level and one - loop terms which , as noted above , is necessarily non - perturbative in character @xcite .the one - loop radiative corrections may be calculated using the coleman - weinberg one - loop effective potential , v_1= str d^2 p p^2 ln(p^2+m^2 ) where @xmath59 represents the square mass matrices and @xmath60 the supertrace . as discussed above the 4 fermi interactionhas a form factor which falls rapidly above @xmath57 effectively giving a momentum space cutoff at the condensation scale defined as the scale where the gauge coupling constant ( cf .eq.([a1 ] ) ) becomes strong .the leading contribution to the one - loop potential @xmath61 comes from the supersymmetry breaking mass of the gaugino fields and keeping only these terms one has v_1=-_c^4j(x_g ) with j ( x)=x+x^2 ln()+ln(1+x ) @xmath62 ( @xmath63 the dimension of the hidden gauge group ) and @xmath64 . the tree level gaugino mass is given by @xmath65 with @xmath66 given in eq.([a21 ] ) and the factor @xmath67 in eq.([a30 ] ) is due to non - canonical kinetic terms for the gauginos .in addition there should be a supersymmetric contribution to the mass of the hidden sector states , gauginos and gauge bosons , generated by the strong hidden sector forces which ( in analogy with qcd ) may be expected to be confining . since the inclusion of these supersymmetric masses only results in a shift of the v.e.v . of the dilaton fieldwe will not consider them in this paper .we now consider the minimisation of the full effective potential , @xmath68 . as discussed in @xcite , a supersymmetry breaking solution with non zero @xmath35is found for a large four fermion coupling . to regulate the unphysical divergence coming from the vanishing of @xmath69 we cut off its evolution at @xmath70 , the point at which the gauge coupling becomes large .this is done to give the dimensionally correct form for the effective four fermion interaction at low energies , @xmath71 . with thisthere is a well defined minimum with the condition @xmath72 being satisfied through a cancelation of tree level with one loop terms . as such it is necessarily non- perturbative and shows that , within our four fermion approximation , gaugino condensation is dynamically generated .the condition @xmath73 can be satisfied either for @xmath31 at the dual invariant points ( @xmath74 ) or for large values of @xmath75 corresponding to squeezed orbifolds . in the latter caseonly if @xmath76 it is possible that different moduli @xmath31 and @xmath16 have a large v.e.v and their v.e.v.s will necessarily be the same ( up to modular invariant transformations ) , i.e. @xmath77 .we see that it is thus only possible to have a large overall moduli @xmath78 if all three @xmath79 have the same value , as in the case for a @xmath80 or @xmath81 orbifold where all @xmath79 vanish . for an orbifold with two completely rotated planes then only one @xmath79 is ( possibly ) different than zero in which case the v.e.v . forthe moduli attached to this plane will be different than those of the other two .a squeezed orbifold may thus naturally be obtained .the v.e.v . of @xmath82 and @xmath83are given in terms of the dimension of the hidden gauge group , its one - loop n=1 @xmath43-function coefficient and @xmath84 by y=8 and x_gln(x_g)=(3 - 1 ) from eq.([eq : e ] ) we obtain the value of the moduli _ i(1-_i)t_ri=- ln(q ) where the sum in eq.([a37 ] ) is over all moduli that acquire a v.e.v different from the dual invariant points and @xmath85 .if the gauge group is broken down from @xmath86 to a lower rank group such as @xmath87 with @xmath88 ( as can be easily done by compactifying on an orbifold with wilson lines ) a phenomenologically interesting solution can be obtained with only one gaugino condensate . the value of the dilaton is mainly given by the dimension of the hidden gauge group and for a phenomenologically interesting solution one needs @xmath89 .in particular if the hidden sector gauge group is @xmath90 the gauge coupling constant at the unification scale would be @xmath91 . using eq.([a36 ] ) , the fine structure constant at the unification scale is given by ^-1_gut=()^-1 .for an @xmath90 gauge group @xmath92 .this value is consistent with minimal supersymmetric standard model ( mssm ) unification . on the other hand , the v.e.v . of the moduli fieldsdoes depend on @xmath48 , the v.e.v . of the dilaton and @xmath93 .the value of @xmath84 is approximately @xmath94 if either the modular weight of the hidden matter fields is @xmath95 ( i.e. untwisted fields ) or if the dominant term in the n=1 @xmath43-function coefficient for the hidden gauge group @xmath93 is given by the quadratic casimir operator @xmath4 , in which case from eq.([alpha ] ) ( taking @xmath96 ) @xmath97 if we only consider the three diagonal ( 1,1 ) moduli ( which are always present in orbifold compactification ) then , for a gravitino mass of order @xmath98 and @xmath99 , @xmath100 depending on whether three , two or one moduli get a `` large '' v.e.v . , respectively . in the case discussed above of a large overall modulus @xmath101 .note that this result is quite different from the values previously obtained @xmath102 .clearly the discrepancy is due to the inclusion of the strong gaugino binding effects in @xmath61 ( cf .eq.([a29 ] ) ) . up to nowwe have only considered the scalar potential for gauginos in the hidden sector , dilaton and moduli fields .we would now like to include the contribution from matter superfields .in general the matter superpotential is given as a power series expansion in the chiral superfields @xmath105 suppressed by the planck mass . in the absence of linear and quadratic terms , the leading term in the superpotential is cubic in the matter superfields since one assumes that the matter fields at most get a small v.e.v .compared to the planck mass to be consistent with the kahler expansion in eq.([a7 ] ) , i.e. @xmath106 . the gaugino condensate term @xmath8 will then give the dominant contribution to the scalar potential and the v.e.v .of the dilaton and moduli fields obtained in eqs.([a36 ] ) and ( [ a37 ] ) will remain valid . for a generic matter superpotential @xmath9 , using the kahler potential of eq.([a7 ] ) , the tree level scalar potential is v_0=e^k((k^-1)^s_s|f_sw_0 + k_s w_m|^2 + _ i(k^-1)^i_i|f_iw_0+_i|^2 - 3|w_0+w_m|^2 ) where @xmath107 @xmath108 @xmath109 @xmath93 is the one - loop beta function coefficient for the hidden gauge group , @xmath110 is the derivative of the gauge kinetic function with respect to the i - field and the index @xmath111 runs over all matter and moduli fields . the first term in eq.([a45 ] ) is just the square of the auxiliary field of the dilaton @xmath36 while the second term is a sum of the squares of the auxiliary fields of the moduli @xmath112 and chiral matter fields @xmath113 . to determine the vacuum structure one has to include quantum corrections and the main contribution to the one - loop potentialis again given by the gaugino - loops .we would like to emphasise again that , although we are just calculating the one - loop potential , the final result is necessarily non - perturbative , since by solving the mass gap equation one is stabilizing the tree level potential by the one- loop potential , and one is effectively summing an infinite number of gaugino - bubbles .one can easily determine the contribution of these loops by calculating the gaugino mass at tree level and then using eq.([a29 ] ) .since all dependence on the gaugino bilinear in eq.([a13 ] ) is given in @xmath8 one has , m_g = e^k/2h with hf^s(k^-1)^s_s(f_s w_0+k_sw_m)+f^i(k^-1)^i_i(f_i w_0+_i)-3(w_0+w_m ) .the one - loop potential ( considering the contribution from the gauginos - loops only ) is ( cf .eq.([a29 ] ) ) @xmath114 v_1=-_c^4(2x_g+x^2_gln(x_g ) ) where we have expanded the function @xmath115 since , as we showed in eq.([eq : e ] ) , @xmath116 and it is now given by x_g = = |h|^2 .if @xmath117 one has @xmath118 and eq.([a46 ] ) reduces to eq.([a30 ] ) , as should be the case .the scalar potential @xmath119 is then @xmath120 \la{a50}\end{aligned}\ ] ] so far we have not discussed the cosmological constant at the minimum of the potential .in fact , in the absence of matter fields , it is negative , of order @xmath121 .it may be cancelled through the addition to the matter superpotential of a term linear in a chiral superfield d @xmath122 to obtain a more transparent expression for @xmath123 one can expand the scalar potential in powers of the gravitino mass ( @xmath124 ) . making use of the solutions to the extremum equations eq.([a36 ] ) and eq.([a37 ] ) the scalar potential is v = m^2_3/2b + ( m_3/2a+h.c . ) + c with @xmath125 a = e^k/2 ( d(k_d^d|d|^2c+c)+f^s(k^-1)_s^sk_s cd-3cd)(- k_d^d|d|^2 ) and @xmath126 where the @xmath111-index in eqs.([a59 ] ) runs over all chiral superfields but for @xmath127 .the leading contribution from the superpotential of eq.([eq : cc ] ) is through the auxiliary field of @xmath127 ( @xmath128 ) .this contribution is semipositive definite and for vanishing v.e.v . of @xmath127 the only term that survivesis the one proportional to @xmath129 and the one - loop potential is independent of it .it is in fact this term that gives the main contribution to the scalar potential even for non - vanishing v.e.v . of @xmath127 and since it is positive it allows for the cancellation of the cosmological constant by fine tuning @xmath130 .while such a tuning is unnatural , it does demonstrate that it is possible to have a zero cosmological constant with supersymmetry broken at a phenomenologically realistic value .it is now straightforward to compute the soft supersymmetry breaking masses in the visible sector .the main contribution comes from the auxiliary field of the dilaton @xmath36 as can be seen from eqs.([a51 ] ) and ( [ a59 ] ) since the v.e.v . of @xmath131 .the leading contribution to the scalar mass @xmath132 comes from the @xmath133 dependence implicit in b and gives ( assuming zero cosmological constant ) m_0 ^ 2=-b_0()^2 m^2_3/2 since @xmath134 and @xmath135 eq.([a57 ] ) is positive showing , contrary to analyses neglecting the effect of @xmath136 , that the scalar potential has a minimum for @xmath137 .note that the scalar mass is the same for all matter superfields regardless of their modular weights .since the auxiliary field of the dilaton is much bigger then the auxiliary field of the @xmath127 field the scalar mass is only sensitive to the fact that one has no cosmological constant and not to the details of the cancelation of the vacuum energy . in the absence of v.e.v.s for the matter fields ,the common supersymmetric mass is independent of the trilinear superpotential . using eqs.([a15 ] ) , ( [ a18 ] ) and ( [ a24 ] ) one obtains the leading contribution to the common gaugino mass , @xmath138 , of the observable sector m_1/2=-6 ( ) m_3/2 this comes from the @xmath139 auxiliary field contribution and is the same for all gaugino masses at the unification scale where the gauge coupling constants meet .the remaining soft terms are proportional to the a component of the matter superpotental , @xmath140 .they are given by @xmath141 where n is the degree of the normalized superpotential @xmath142 and @xmath143 .\label{eq : r}\ ] ] for the leading trilinear terms , @xmath144 , whose coupling , @xmath145 , is independent of the moduli @xmath78 this reduces to @xmath146 for a bilinear term , such as the mssm term @xmath147 , with @xmath22 independent of @xmath78 we have @xmath148 one problem associated with such a term is that there is no obvious reason why @xmath149 as must be the case if the hierarchy problem is not to be re - introduced . to answer this it has been suggested that the @xmath22 term originates from a term in the kahler potential proportional to @xmath150 .after supersymmetry breaking this generates the @xmath151 term in the superpotential with @xmath152 . in this case the soft term coming from the b term in eq.([a51 ] ) is again of the form of eq.([eq : asoft ] ) with b&=&()^2 [ eq : r2p ] + & = & ( ) ^2 another suggestion is that the @xmath22 term comes from a term in the superpotential proportional to @xmath153 . in this casewe find from the b term the contribution of the form of eq.([eq : asoft ] ) with b&=&()^2 + & = & ( ) ^2 [ eq : r2pp ] this completes the derivation of the soft terms arising from gaugino condensation .it is also interesting to calculate the fine tuning parameter @xmath154 , defined as || < where @xmath11 are the soft susy parameters and @xmath155 the mass of @xmath156 boson .this quantity measures the allowed degree of the fine tunning needed to get the correct mass @xmath157 ; for @xmath158 the value of @xmath11 must be fine tuned to @xmath159 of the @xmath156 mass . at a crude approximation @xmath160 with @xmath161evaluated at the electroweak scale and it is given by @xmath162 with @xmath163 .it is illuminating to compute them for realistic examples with an @xmath90 gauge group in the hidden sector with different matter content .they are shown in * tables 1 - 3*. notice that the fine structure constant at the unification scale is of the right order of magnitude as required by mssm .it is the dimension of the gauge group in the hidden sector that sets the value of @xmath164(cf .eq.([eq : sc ] ) ) .these results suggest that @xmath165 is large contrary to the analyses based on assuming supersymmetry breaking originates at tree level from the dilaton f term even though in this case too the dilaton f term dominates .there are two reasons for this . in the first place the mechanism explored here , being nonperturbative , necessarily requires that loop corrections to the potential be as important as the tree level contribution .indeed there is a strong suppression of the @xmath166 terms due to a cancelation of the tree and loop contributions .secondly the cancelation of the cosmological constant in gaugino condensation requires a contribution from the matter superfield sector .both these effects are absent in @xcite and explains the difference between the resulting soft terms . for a wide range of values for @xmath84 , a large hierarchy can be obtained .however , the soft supersymmetric terms are more sensitive to @xmath84 . in * fig.1 * we show @xmath167 , @xmath168 and the trilinear term @xmath169 as a function of @xmath84. for increasing @xmath84 , @xmath132 and @xmath169 become smaller .the common supersymmetric mass , @xmath132 , is always larger than @xmath170 . on the other hand ,the gaugino mass is always smaller than @xmath170 and it approximates the gravitino mass for @xmath171 . the ratio of @xmath168 is independent of the value of @xmath170 , but @xmath167 is not .the reason is that @xmath167 depends on @xmath172 , which sets the hierarchy .running the masses down to the electroweak scale , neglecting corrections due to possible large yukawa couplings , the common supersymmetric mass remains the same while the gluino mass increases by a factor of three .if we demand all supersymmetric masses to be in the range @xmath173 , then @xmath174 . from the range of values of the parameter @xmath84 , the preferred value is for @xmath84 positive and close to @xmath175 . for this value of @xmath84the difference between the supersymmetric masses of the scalar and gaugino fields is smallest .this allows for having a small fine tuning parameter @xmath154 . at the same time, the trilinear term @xmath176 decreases .this is welcome since one - loop diagrams , associated to the complex phases of the trilinear terms , contribute to the electric dipole moment of the neutron . for @xmath177 ,this contribution is about @xmath178 bigger than the experimental values .furthermore an @xmath179 minimizes the unification scale and allows for unifying the standard model gauge groups with the hidden sector one and the fine tuning parameter @xmath154 is smallest .thus , from all the range of possible values for the parameters @xmath180 and @xmath84 , consistency with the mssm and electroweak breaking , selects a very narrow band .it is remarkable that the values of all soft supersymmetric terms improve within this band taking quite acceptable values . to summarise ,we have considered the phenomenological implications of susy breaking triggered by a gaugino condensate in the hidden sector . in a supergravity theory the strong binding effects play a role in all sectors of the theory and must be included .here we have modelled these effects via a simple ( n=1 locally supersymmetric ) four gaugino interaction which allowed us to use n - j - l techniques to estimate the nonperturbative physics .we thus constructed the modular invariant effective potential which determines the magnitude of the gaugino condensate and also the moduli of the theory .we have generalized the work in by including several moduli fields . as a result of minimizing the scalar potential we found that the v.e.v . of the dilatonis mainly given by the dimension of the hidden sector gauge group .for reasonable hidden sector gauge groups a large mass hierarchy can arise together with an acceptable value for the fine structure constant at the string scale with only one gaugino condensate .we found that the extremum eqs .for the moduli are solved for a unique value of the parameter @xmath84 ( cf .eq.([alpha ] ) ) and thus the v.e.v . of the modulitake a common value ( @xmath77 ) related to @xmath183 and the v.e.v of the dilaton or they take the dual invariant values ( @xmath184 ) .this result naturally accommodates the squeezed orbifold solutions which are better candidates for minimal string unification .we also showed that it is possible to fine tune the cosmological constant to zero while having susy broken at a phenomenologically interesting value . incorporating chiral matter fields we computed the soft susy breaking terms .the ratio between the scalar and gaugino mass is given and it depends on @xmath183 and the v.e.v . of the dilaton .the numerical values for phenomenologically interesting solutions are consistent with results obtained from requiring acceptable values for the gauge couplings , electroweak breaking , dark matter abundance and @xmath185 assuming a minimal unification of the mssm@xcite .v. kaplunovsky and j. louis , cern - th.6809/93 ; r. barbieri , j. louis and m. moretti cern - th.6856/93 .a.brignole , l.e.ibanez and c.munoz , universidad autonoma de madrid preprint ftuam-26/93 a. font , l.e .ibez , d. lst and f. quevedo , phys .b245 ( 1990 ) 401 .m.cvetic , a.font , l.e.ibanez , d.lust and f.quevedo , nucl.phys .b361(1991)194 . g.veneziano and s. yankielowicz , phys .lett . 113b ( 1984 ) ; t.r .taylor , phys .lett 164b ( 1985 ) 43 s. ferrara , n. magnoli , t.r .taylor and g. veneziano , phys .lett . b245 ( 1990 ) 409 ; p. binetruy and m. k. gaillard , phys .232b ( 1989 ) ; nucl .b358 ( 1991 ) 121 .nilles and m. olechowski , phys .b248 ( 1990 ) 268 ; p. binetruy and m.k .gaillard , phys .b253 ( 1991 ) 119 ; j. louis , _ `` status of supersymmetry breaking in string theory '' _ , slac - pub-5645 ( 1991 ) ; d. lst and t.r .taylor , phys .b253 ( 1991 ) 335 ; b. carlos , j. casas and c. muoz , phys .b263 ( 1991 ) 248 ; s. kalara , j. lopez and d. nanopoulos , _`` gauge and matter condensates in realistic string models '' _ , preprint ctp - tamu-69/91 .l. dixon , v. kaplunovsky and j. louis , nucl.phys .b355 ( 1991 ) 649 ; g. lopes cardoso and b. ovrut , nucl.phys .b369 ( 1992 ) 351 ; _ `` sigma model anomalies , non - harmonic gauge and gravitational couplings and string theory , '' _preprint upr-0481 t ( 1991 ) .n.v.krasnikov , phys .193b(1987)37 ; l.dixon , in proceedings of the aps dpf .meeting , houston , 1990 ; j.a.casas , z.lalak , c.munoz , and g.g.ross , nucl.phys . b347(1990)243 ; t.taylor , phys.lett b252(1990)59 ; b. de carlos , j. a. casas and c. munoz , cern - th.6436/92 and ref . therein .r. barbieri and s. cecotti , z. phys .c 17 , ( 1983 ) 183 ; m. srednicki and s. theisen , phys .54 , ( 1985 ) 278 ; p. binetruy , s. dawson , m.k .gaillard and i. hinchliffe , phys .d 37 ( 1988 ) 2633 .g. costa , j. ellis , g.l .fogli , d.v .nanopolous and f. zwirner , nucl.phys .b297 ( 1988 ) 244 ; j. ellis , s. kelley and d.v .nanopoulos , phys .b249 ( 1990 ) 441 ; phys .b260 ( 1991 ) 131 ; p. langacker , _ precision tests of the standard model " , _ pennsylvania preprint upr-0435 t , ( 1990 ) ; u. amaldi , w. de boer and h. frstenau , phys .b260 ( 1991 ) 447 ; p. langacker and m. luo , phys.rev.*d44 * ( 1991 ) 817 ; r. roberts and g.g .ross , preprint ral-92 - 005 ( 1992 ) .g. f. giudice and a. masiero , phys .206b ( 1988 ) 480 ; j.e .kim and h .-nilles , phys .b 138b ( 1984 ) 150 ; j.l . lopez and d.v .nanopoulos , phys .b b251 ( 1990 ) 73 ; e.j . chun , j.e .kim and h .-nilles , nucl .b370 ( 1992 ) 105 .k. inoue et al . , prog.theor.phys .* 68 * ( 1982 ) 927 ; l.e .ibez , nucl.phys .b218 ( 1983 ) 514 ; l.e .ibez and c. lpez , phys .b126 ( 1983 ) 54 ; nucl.phys .b233 ( 1984 ) 511 ; l. alvarez - gaume , j. polchinsky and m. wise , nucl.phys .b221 ( 1983 ) 495 ; l.e ibanez , c. lopez and c. munoz , nucl .b256 ( 1985 ) 218 .

<S> we consider the predictions for the hierarchy of mass scales , the fine structure constant , the radii of compactification and the soft susy breaking terms which follow if susy breaking is triggered by a gaugino condensate . </S><S> -1 cm outp-93 - 32p + * unification of couplings and soft supersymmetry breaking terms in 4d superstring models * + and g. g. ross + [ 5 mm ] _ department of physics , university of oxford , + 1 keble rd , oxford ox1 3np _ + [ 8 mm ] </S>

nucl-ex0608019

the hot and dense medium created in heavy - ion collisions at the relativistic heavy ion collider ( rhic ) is extremely short - lived which requires probes from the early stage of the collision to extract information about the medium and its constituents .the @xmath1-meson , which is the lightest bound state of hidden strangeness ( @xmath13 ) , can be used as a clean probe from the early stage : the @xmath1 is assumed to have a small cross - section for interactions with non - strange particles @xcite and it has a relatively long lifetime ( @xmath1441 fm/@xmath5 ) which means it most likely decays outside the fireball , ensuring that its decay daughters do not lose information through rescattering with other hadrons .previous star measurements @xcite have ruled out @xmath15 coalescence as a @xmath1-meson production mechanism and therefore the information carried by the @xmath1 should remain unmodified by the hadronic phase .the elliptic flow parameter , @xmath0 , is an observable which provides information from the early stage of the system s evolution .information concerning the partonic collectivity of the medium can be extracted from @xmath0 measurements of the @xmath1 by comparing its @xmath0 values with those of particles comprised of the lighter @xmath16 and @xmath17 quarks and with other multistrange hadrons ( @xmath18 and @xmath19 ) .the @xmath1-meson can also provide information on particle production mechanisms since it is a meson , but at the same time is almost as heavy as a @xmath9 baryon . comparing the @xmath1-meson nuclear modification factor , @xmath8 , with other identified hadrons can help to differentiate between mass - type and meson / baryon - type dependencies of particle production .medium effects on particle production may be studied through comparison of the @xmath1-meson transverse momentum , @xmath3 , distributions with model expectations .analysed data are from the au+au runiv 200 gev dataset. for the @xmath0 and spectral analysis @xmath1413.5 million minimum bias events ( 0 - 80@xmath20 ) were analysed , while @xmath144 million central triggered events were used to obtain @xmath3 spectra in ( 0 - 5@xmath20 ) central collisions .the @xmath1-meson was reconstructed through its decay to two charged kaons ( branching ratio = @xmath21 ) which were identified through their ionisation energy loss in the star time projection chamber ( tpc ) . to extract the raw yield of @xmath1-mesons , an invariant mass ,@xmath22 distribution was constructed using all combinations of positively and negatively charged kaons from each event .the combinatorial background from uncorrelated @xmath23 pairs was estimated using event - mixing , scaled to the same - event distribution and then subtracted .the resulting @xmath22 distribution was fitted with a breit - wigner distribution ( to describe the shape of the @xmath1 mass peak ) plus a straight line ( for the residual background ) .the first measurement of the @xmath1-meson elliptic flow , @xmath0 , at @xmath2 gev is shown in fig . [fig : v2 ] by the solid circles .the curves represent quark number ( nq ) scaling @xmath0 parameterizations ( dashed for nq = 2 , dot - dashed for nq = 3 ) from @xcite which indicate the meson and baryon @xmath0 trends as a function of @xmath3 .elliptic flow , @xmath0 , as a function of @xmath3 for the @xmath1-meson ( solid circles ) compared to the @xmath7 ( open squares ) and @xmath9 ( open circles ) @xcite . statistical errors are represented by the error bars and the systematic errors are shown by the shading around the data points .the dashed(dot - dashed ) curve represents a parameterization for @xmath0 based on nq = 2(3 ) scaling from @xcite . ]the @xmath1-meson @xmath0 is the average of results obtained with two different , yet consistent , methods .the first method has been previously established @xcite for measurements of other identified particle @xmath0 while the second method , the @xmath0 vs. @xmath22 method , is based on @xcite . the first important point to notice in fig .[ fig : v2 ] is that the @xmath1-meson has a significant @xmath0 , which is comparable to that of other identified particles such as the @xmath7 ( open squares ) and @xmath9 ( open circles ) @xcite as well as the multistrange baryons @xmath18 and @xmath19 @xcite ( not shown here ) . for @xmath24 gev/@xmath5 ,the @xmath1-meson @xmath0 seems consistent with a hydrodynamic - type mass - ordering while at intermediate @xmath3 , ( @xmath25 gev/@xmath5 ) , it is more consistent with the @xmath26 of the @xmath7 and the scaling for nq = 2 , than the @xmath9 ( nq = 3 ). therefore the @xmath1 @xmath0 seems to be determined by its meson properties rather than its mass .this is further evidence of the particle species - type dependence of @xmath0 observed at intermediate @xmath3 at rhic , and can be described in terms of particle production through quark recombination or coalescence @xcite .the nuclear modification factor , @xmath8 , can provide insight on the system created in central compared to peripheral collisions since it describes the yield of @xmath1-mesons produced in central compared to peripheral collisions , scaled by the number of binary collisions ( calculated using a glauber model ) and is given by @xmath27 . nuclear modification factor , @xmath8 , for two choices of peripheral bin : 40 - 60@xmath20 ( top panel ) and 60 - 80@xmath20 ( bottom panel ) , for the @xmath1-meson ( solid circles ) . also shown are star @xmath8 measurements for @xmath7 ( open triangles ) and @xmath9 ( open squares ) from @xcite .error bars represent statistical and systematic errors added in quadrature .uncertainties due to the glauber @xmath28 calculations are indicated by the shaded bands . ]figure [ fig : rcp ] shows @xmath8 of the @xmath1-meson ( filled circles ) for two choices of peripheral bin compared to @xmath8 measurements for @xmath7 ( open triangles ) and @xmath9 ( open squares ) @xcite . for both choices of peripheral bin , the @xmath1 @xmath29 is less than unity , implying that the yield in central collisions is suppressed compared to peripheral collisions .similarly to the trend noticed in the @xmath0 measurements , the @xmath1 @xmath8 follows very closely the trend of the @xmath7 rather than the @xmath9 .this is additional confirmation of the meson / baryon - type dependence rather than a mass - type dependence of @xmath8 at intermediate @xmath3 at rhic @xcite .this grouping of mesons and baryons into separate bands in @xmath29 has been predicted by recombination and coalescence models of particle production @xcite .left panel shows the @xmath1 and @xmath19 @xmath3 spectra for the most central collisions compared to recombination model expectations from @xcite .the sum of all(thermal - thermal ( tt ) recombination ) contributions to the model is shown as the solid(dashed ) lines .the right panel shows the @xmath11 ratio as a function of @xmath3 compared to the recombination model expectations . ]comparison of the @xmath1 @xmath3-spectra with model expectations may also provide further insight into the constituents of the medium produced in au+au collisions .the left panel of fig .[ fig : specratio ] shows the central @xmath1-meson ( solid circles ) and @xmath19 ( @xmath30 ) ( open circles ) spectra compared to recombination model expectations from @xcite .this model predicts that the production of @xmath1-mesons and @xmath19-baryons at intermediate @xmath3 ( @xmath318 gev/@xmath5 ) is predominantly due to the recombination of thermal @xmath10-quarks in the medium .this is shown by the sum of all contributing recombination terms ( solid lines ) being almost identical to the term due to thermal recombination alone ( dashed lines ) i.e. the contribution to @xmath1 and @xmath19 production from fragmentation processes in this model is very small .these expectations seem to match the trend of the data .the ratio of the @xmath11 yields is shown in the right panel of fig .[ fig : specratio ] . in this case, the model predicts a monotonic increase in the @xmath11 ratio until @xmath32 gev/@xmath5 .although the data deviate from the model predictions at higher @xmath3 , their trend can be described up to @xmath12 gev/@xmath5 .in summary , the large @xmath0 of the @xmath1 at low @xmath3 , despite its low cross - section for interactions in the hadronic phase as well as the fact that the @xmath26 scales with nq = 2 , are strong indications of partonic collectivity and deconfinement of the medium created in au+au collisions at rhic .the good agreement of the @xmath1-meson @xmath29 with @xmath8 values of other mesons rather than the heavier baryons , provides additional confirmation of the meson / baryon - type dependence of this observable at intermediate @xmath3 .the scaling of the @xmath0 and @xmath8 values of the @xmath1-meson with nq=2 is expected from recombination and coalescence models of particle production .in addition , the trend of the central @xmath11 ratio can be described up to @xmath12 gev/@xmath5 by a model based on recombination of thermal strange quarks in the medium .

<S> we present the first elliptic flow , @xmath0 , measurement for the @xmath1-meson in au+au collisions at @xmath2 gev . at low @xmath3 ( @xmath4 gev/@xmath5 ) , </S><S> the @xmath0 is consistent with mass ordering expected from hydrodynamics , while at higher @xmath3 ( @xmath6 gev/@xmath5 ) , it follows the same trend as the @xmath7 @xmath0 and a parameterization for number of quarks = 2 . </S><S> the nuclear modification factor , @xmath8 , has also been measured and it follows the same trend as the @xmath7 rather than the @xmath9 which confirms the baryon / meson - type dependence of @xmath8 at rhic . </S><S> a model based on the recombination of thermal @xmath10-quarks describes the central @xmath1-meson @xmath3 spectrum as well as the baryon / meson ratio of @xmath11 up to @xmath12 gev/@xmath5 . </S>

1311.5787

dynamic robot locomotion is a particularly challenging study .one of its main difficulties is due to the fact that the robot is sub - actuated , that is , on a dynamic walking gait it is impossible to freely control each orientation of the robot s links .it is straightforward to see this limitation as there is no actuator between the ground and the support polygon , the gravity alone will impose an additional torque that will change the robot orientation . aside from this , a dynamic walking bipedal robot can achieve greater speeds than a static walker , so there is a practical interest in this domain of research . in @xcite the problem of sub - actuated dynamic walkingis treated in a systematic way that will lead to controller design with assured stability and also , with a step evolution clocked , not by the time , but by its own geometry .this self clocked characteristic will be instrumental in this work .we can see each step as a trajectory for the state vector of the dynamic model .the classical controller design will ensure the state of the model to track some reference tied to the time and some control effort will take place even if the robot state is valid for the reference trajectory but is not synchronized with the reference . to successfully perform a step ,the robot links must be driven in some specific way , even for coordinates that can not be directly controlled . to accomplish this it may be necessary , for example , to swing the robot s torso and this can be undesirable in some cases . to have complete control over one coordinate , it is necessary to give up the control of another one .so , if it is possible to ignore the position or the orientation of a link , it will be possible to control every other important coordinate . by using a inertial disc ,its orientation will be a cyclic variable , and thus , could be ignored . that is the main idea introduced by the design presented in @xcite and studied in many works as @xcite , @xcite and @xcite .theoretically it is possible to ignore the disc position and speed , but in practice , there will be some limitation on the maximum speed achieved by the disc , so that the actuators do not saturate .one solution to this problem is proposed in @xcite by means of a supervisory control , that will perform a different trajectory if the disc speed is beyond some limit value .section [ sec : dynamic - model ] presents the hybrid dynamic model as proposed in @xcite .there is also a brief presentation of the hypotheses and terminology therein and inherited by this paper .next , it will be introduced the trajectory planning in section [ sec : design - of - thetrajectory ] .this is the main practical benefit of our approach , valid only for this specific class of bipedal robots . in @xcite ,the proposed strategy is to find the evolution of the robot geometry as function of the absolute orientation and then check if it will lead to a monotonic evolution of that absolute orientation . in our approach, it is possible to fix the robot geometry as a desired function of the absolute orientation . then find a monotonic evolution of the absolute orientation that lead to a repeatable evolution of the disc speed .another benefit is that the proposed control does not deal at all with the disc position , but can continuously control its speed .the control law and the convenient change of coordinates is also presented , introducing the core contribution of this paper .section [ sec : controlstability ] present a demonstration of exponential stability around some neighborhood of the planed trajectory .finally in section [ sec : numerical - simulations ] there are some simulation results .the robot will be modeled by a chain of @xmath0 rigid links , each one with known parameters such as mass , center of mass and inertia .the links are connected by @xmath1 frictionless joints independently actuated .there will be at least one and at most two point of contact with the ground , called foots .the foots are punctual and unactuated . for illustration purposes ,a very simple bipedal robot is depicted at figure [ fig : chap2_examplebiped ] .the coordinate @xmath2 is the only absolute coordinate , @xmath3 is a relative coordinate to represent the disc position and @xmath4 is a relative coordinate representing the angular displacement between the two legs . ] on a higher order robots @xmath4 will be a vector representing the angular displacements between each link .indeed , the robot s shape will be determined by @xmath4 .the walking takes place on a surface , is restricted to the sagittal plane , and is composed of alternating phases of single and double support . during the single support phase , the stance foot acts as an ideal pivot .the double support phase is instantaneous and associated impacts are modeled as a rigid contact . at the impact , the swing leg touches the ground with no slip nor rebound and , the former stance leg releases without interaction .the robot model is derived under the hypotheses briefly cited above and detailed in @xcite , hr1~hr5 , hgw1~hgw7 and hi1~hi7 ( ( * ? ? ?* p48 p50 ) ) .a small change is introduced in hr6 . * the modelis expressed in one relative coordinate @xmath3 for the disc , n-2 relative coordinates @xmath4 for the rest of the body and only one absolute coordinate @xmath2 .using lagrangian formulation , the dynamic equations for the continuous phase , between two impacts , will be @xmath5 where@xmath6 @xmath7 , @xmath8 , @xmath9 and the vector of actuator torques being defined by @xmath10 .@xmath11 is called the inertia matrix , @xmath12 is called the coriolis matrix , @xmath13 is called the gravity vector and @xmath14 will map the actuator torques as generalized forces. it will be more convenient to define@xmath15 where @xmath16 is the angular moment of the system , calculated as @xmath17 @xmath18 being the last line of the inertia matrix .its derivative will be @xmath19 @xmath20 being the last line of the gravity vector. there will be a diffeomorphism between @xmath21 and @xmath22 and so it is possible to define the state vector@xmath23 and denote by @xmath24 the set of valid values for the vector state @xmath25 .at some moment , the swing foot will touch the ground with state @xmath26 and the system will be mapped to a new state @xmath27 . it is assumed that the walking gait is transversal to the impact surface .so , the impact forces will change the speeds on @xmath22 .after the impact the roles of the legs are reversed with a change in @xmath28 . as result , the system state , @xmath27 , after the impact is @xmath29 with @xmath30 being a constant involutive matrix and @xmath31 a function of the robot geometry @xmath32 and absolute orientation @xmath33 .note that any value with a superscript @xmath34 means the value just before the impact at the end of the step , and a superscript @xmath35 means the value immediately after the impact at the beginning of a new step .this is the same notation utilized in @xcite. the impact , modeled by the mapping @xmath36 , will take place at a given configuration that can be represented by a manifold @xmath37 , for example@xmath38 where @xmath39 is a convenient can express for example , the vertical position of the swing foot and will reach zero only when the swing foot touch the ground . ]function for which @xmath40 and @xmath41 .the complete hybrid model will be @xmath42 for this particular class of robots , if the coordinate vector is choose as ( [ eq : vector_state ] ) then it is possible to demonstrate that the first column of @xmath31 is always @xmath43 .this is due to @xmath3 being a cyclic coordinate .in this section we present some aspects of the trajectory generation . the idea of a geometry tied to the absolute orientation is presented in @xcite . in this kind of design, all the actuators will be used to drive the robot geometry as a function of the absolute orientation .this geometry must be chosen in a way that , apart from not being directly controlled , the absolute orientation will have a monotonic evolution .also , the interaction of the robot with the ground must produce a stable limit cycle . in our robot , the disc position will be ignored .this allows us to get a new input available .every relative link orientation , except for the disc coordinate @xmath3 , should track a predefined function of @xmath2 .this way , at any time , the robot configuration will be a function of its absolute orientation .this kind of control is equivalent to drive @xmath44 to zero on the output defined as @xmath45 with @xmath46 being the desired evolution of @xmath4 as a function of @xmath47 .the output described by ( [ eq : chap2_ouroutput ] ) has dimension @xmath1 which is the same dimension of the input @xmath48 .the dynamic system can be written in terms of the global diffeomorphism defined by@xmath49 if the decoupling matrix @xmath50 is square and @xmath48 have both the same dimension . ] and invertible , there exists a input @xmath51 so that by the use of @xmath52 and the coordinate change ( [ eq : coordinate_p ] ) , the system can be viewed as @xmath1 independent series of 2 integrators driven by the virtual input @xmath53 .thus , if it is possible to find @xmath52 , it will be possible to control every entry of @xmath54 by the input @xmath55 .more information on the exact linearization control can be found in @xcite .applying the change of coordinates defined by ( [ eq : coordinate_p ] ) it is possible to write @xmath56 using the input ( [ eq : chap2_linearinput ] ) , the change of coordinates ( [ eq : coordinate_p ] ) and discarding @xmath3 , it is possible to rewrite the continuous phase as@xmath57 now define an embedded manifold @xmath58 from equation ( [ eq : linearizedsystem ] ) it is possible to see that the input @xmath59 can be used to drive @xmath60 .the input @xmath61 remains free to control the absolute speed @xmath62 and also , when restricted to @xmath63 , the angular momentum @xmath16 .assume that it is possible to find functions @xmath64 and @xmath65 , such that if @xmath66 , then @xmath67 , where @xmath68 is some constant bias .it is possible to use the input @xmath61 to exponentially attenuate this bias .the manifold @xmath63 is said to be forward invariant if solutions starting at @xmath63 will remain in @xmath63 .it will be said to be impact invariant if @xmath69 and @xmath70 .observe that @xmath71 .if @xmath63 is both forward and impact invariant , it is said to be hybrid invariant ( * ? ? ?* p 96 ) . while the feedback control can lead to forward invariance, the impact invariance is a design property . for the nominal trajectory to be repeated, each variable evolution must be hybrid invariant .so , not only @xmath63 , but also the angular momentum reference @xmath65 , and the absolute speed reference @xmath72 , has to be hybrid invariant themselves . one procedure to find such nominal trajectory would be : 1 .fix the absolute orientations at the beginning , @xmath73 , and at the end , @xmath74 , of the step , and the nominal configurations @xmath75 and @xmath76 .2 . determine a desirable evolution @xmath46 for the robot configuration , such that 1 .@xmath77 , as @xmath3 is a cyclic coordinate , it can be ignored ; this will ensure impact invariance on the robot shape @xmath46 2 . depending on the value of @xmath78 , determine @xmath79 and @xmath80 for impact invariance of @xmath81 .there will be also a fixed increment or decrement of the angular momentum @xmath82 .this will lead to impact invariance on @xmath63 .3 . choose a impact invariant and convenient function candidate for @xmath64 , then numerically calculate @xmath83 such that @xmath84 will satisfy the impact invariance restriction for @xmath65 .the control must ensure forward invariance and also a stability around this planned trajectory . by construction of @xmath72 and @xmath65 , when restricted to @xmath63 , where @xmath85 , the following relation holds@xmath86 and so it is possible to write@xmath87 where @xmath88 is a unknown but bounded function .suppose now that exist hybrid impact invariant functions @xmath46 , @xmath72 and @xmath65 .then if we choose @xmath89 with @xmath90 and @xmath91 positive definite , this will ensure convergence of @xmath92 .the convergence @xmath93 will be independent of @xmath2 , @xmath62 , @xmath3 or @xmath94 and also @xmath63 will be forward invariant .the input @xmath61 can be freely used to drive the absolute orientation of the robot .it is theoretically possible to track any reference trajectory by the use of the output ( [ eq : chap2_ouroutput ] ) as we found a linearizable part of dimension @xmath95 , the side effect being that we do not have direct control over the speed accumulated by the disc .but there is an interesting property of this robot , as @xmath3 and @xmath94 can be ignored for the dynamic model .so the remaining coordinates will be @xmath96 , @xmath2 , @xmath16 , @xmath97 , @xmath62 . as @xmath2 is the only absolutecoordinate , @xmath98 is also independent of @xmath2 .now define [ eq : newcoordinates]@xmath99 with @xmath100 and @xmath101 positive constants .it is possible to define the manifold @xmath102 and it will be impact invariant . when the robot is performing the nominal step , @xmath103 and @xmath104 .as @xmath2 is monotonic at the continuous phase , it is possible to integrate ( [ eq : linearizedsystem ] ) at @xmath2 and rewrite it using the coordinates @xmath105 so , defining @xmath106 using the input @xmath59 as defined in ( [ eq : chap2_controllaw_vr ] ) and @xmath61 as @xmath107 then the continuous part of the dynamic system , subject to the given hypotheses , can be written as@xmath108 note also that by using the input ( [ eq : controllaw_vn ] ) , the impact invariant manifold @xmath109 will be forward invariant and thus , hybrid invariant . in thisnew coordinates , the impact map will be [ eq : newsystemwithinput_impactmap]@xmath110 where @xmath111 from ( [ eq : newsystemwithinput_impactmap_a ] ) and ( [ eq : impactmanifold ] ) it is possible to find the impact effect over @xmath112 and the actual integration limits @xmath113 and @xmath33 , @xmath114 .from ( [ eq : newsystemwithinput_impactmap_b ] ) it is possible to find how the remaining part of ( [ eq : newvectorstate ] ) are affected by the impact .the impact map with respect to the vector state ( [ eq : newvectorstate ] ) will be@xmath115 and can be obtained by reshaping ( [ eq : newsystemwithinput_impactmap ] ) adequately . from ( [ eq : newvectorflowfunction ] ) and ( [ eq : newvectorimpactmap ] ) , the complete hybrid system will be@xmath116 when walking , the dynamic system ( [ eq : newhybridsystem ] ) will take a periodic evolution composed of a continuous phase , followed by the impact map and another continuous phase . to evaluate the stability of this periodic orbit , it will be important to define poincar return map for the manifold @xmath37 as@xmath117 where @xmath118 is a neighborhood of @xmath119 , such that@xmath120 with @xmath121 being the integral curve of ( [ eq : newhybridsystem ] ) with initial condition @xmath122 , starting at @xmath123 and ending at @xmath124 .so @xmath125 . by the hybrid invariance of the manifold @xmath109, @xmath126 will be a fixed point of @xmath127@xmath128 as expected , there exist a periodic orbit of @xmath129 and , the interception of this periodic orbit with the impact surface @xmath37 , that is @xmath130 , occours at @xmath119 .under the given robot hypotheses and the existence of @xmath46 , @xmath72 and @xmath65 such that the manifolds @xmath63 and @xmath109 are hybrid invariants , then : - the poincar return map @xmath127 is regular in some neighborhood @xmath131 around @xmath132 . - the poincar return map @xmath127 can be made locally exponentially contracting in some neighborhood @xmath133 of @xmath132 , by adjusting the parameters @xmath90 , @xmath134 , @xmath101 and @xmath100 .if it is possible to find ( [ eq : chap2_linearinput ] ) for every possible value of @xmath25 , then the the system ( [ eq : hybridsystem ] ) , except for the disc position , can be made exponentially stable around some neighborhood @xmath135 of @xmath109 by using the inputs defined by ( [ eq : chap2_controllaw_vr ] ) , ( [ eq : controllaw_vn ] ) and adjusting the parameters @xmath90 , @xmath134 , @xmath101 and @xmath100 .the poincar return map is well defined for @xmath119 , @xmath136 , and this is its fixed point .the manifold @xmath109 is transversal to @xmath37 and solutions of the hybrid system ( [ eq : newhybridsystem ] ) are continuous .so for some sufficiently small perturbation @xmath137 around @xmath138 before the impact , the continuous part of the system will reach again the impact surface .if @xmath139 is defined to be the set in which , in the event of a impact , at least one another impact will follow , then it is possible to conclude that @xmath140 .the impact map ( [ eq : newimpactmapwithinput ] ) can be linearized at @xmath136 .this will lead to@xmath141 where @xmath142 will be a matrix whose non constant elements will increase at most with @xmath143 . also , the integration limits @xmath113 and @xmath33 , being determined by the impact condition , will change if @xmath144 . they can be written as functions and when linearizated around @xmath119 will lead to@xmath145 where @xmath146 and @xmath147 are constants .note that @xmath148 , so the differential equation ( [ eq : newvectorflowfunction ] ) , linearizated around @xmath109 , will be @xmath149 with@xmath150 if the linearized system is exponentially stable , then the non linear system will be stable around some neighborhood @xmath135 of @xmath109 .we can find an explicit solution of ( [ eq : analyticallinearized ] ) for the k - nth step@xmath151 where@xmath152 @xmath153 @xmath154 @xmath155 @xmath156 @xmath157 @xmath158 @xmath159 @xmath160 @xmath161 @xmath162 @xmath163 @xmath164 @xmath165 it is possible to see that the parameters @xmath166 , @xmath134 , @xmath101 and @xmath100 can be chosen large enough to have any initial condition @xmath167 at @xmath168 attenuated as much as we want at the point @xmath169 , end of the step .this will be true even if there are some small pertubation on the integration limits due to @xmath144 .choosing @xmath37 as the poincar surface , it will be possible to write its linearization as@xmath170 and this product will be basically composed of terms like @xmath171 that , after some peak point , will exponentially became smaller as @xmath172 increases . by adjusting the tunning parameters, the linearized poincar map can be made as small as we want and , as result , its eigenvalues can be allocated inside the unit circle . even starting in a neighborhood of @xmath63 , where the integration limits@xmath113 and @xmath33 will be different from the nominal values , the stability is assured .in fact , appropriate values of @xmath90 and @xmath91 will lead , along each step , to the convergence of @xmath173 and @xmath174 to the nominal values @xmath175 and @xmath176 . under the conditions of the corollary there exist a diffeomorphism between ( [ eq : newhybridsystem ] ) and ( [ eq : hybridsystem ] ) , except for the disc position. if the jacobian of the poincar return map evaluated at @xmath136 has its eigenvalues inside the unit circle , then the system ( [ eq : newhybridsystem ] ) will be stable around some neighborhood @xmath135 of @xmath109 .this can be achieved by choosing the parameters of the control law . as conclusion , the system ( [ eq : hybridsystem ] )can be made stable in the neighborhood @xmath135 , except for the disc position .by now some simulation results are presented for a model like the one illustrated on picture [ fig : chap2_examplebiped ] .the corresponding author can submit the numerical values and other details by email .the proposed control law was simulated with parameters @xmath177 and @xmath178 , with the robot starting from rest at positions @xmath179 and @xmath180 .this means that the robot state is outside @xmath63 .the simulated behavior can be viewed on figure [ fig : chap4_speedcontrol-2 ] .manifold.[fig : chap4_speedcontrol-2 ] ] figure [ fig : chap4_speedcontrol-3 ] show the results when some parametric errors are included on the model .anyway , the control remains stable and can drive the robot close to the reference .manifold.[fig : chap4_speedcontrol-3 ] ]we have derived a control law for the presented class of bipedal robots , so that the robot configuration will be tied to the absolute orientation .also the speed of this absolute orientation and the angular momentum will follow predetermined references . as result, the robot will asymptotically converge to a walking gait at the same time as the average disc speed can be driven to zero ( or any other desired value ) by an appropriate offset in the reference for the angular momentum .the stability of the proposed control was proved around some neighborhood of the nominal step and could be verified in numerical simulations .the simulations shows that the domain of attraction is somewhat big as the nominal step is reached even if the robot starts with no speed and the results could also be validated for some parametric errors

<S> in this paper we exploit some interesting properties of a class of bipedal robots which have an inertial disc . </S><S> one of this properties is the ability to control every position and speed except for the disc position . </S><S> the proposed control is designed in two hierarchic levels . </S><S> the first will drive the robot geometry , while the second will control the speed and also the angular momentum . </S><S> the exponential stability of this approach is proved around some neighborhood of the nominal trajectory defining the geometry of the step . </S><S> this control will not spend energy to adjust the disc position and neither to synchronize the trajectory with the time . </S><S> the proposed control only takes action to correct the essential aspects of the walking gait . </S><S> computational simulations are presented for different conditions , serving as a empirical test for the neighborhood of attraction . </S><S> nonlinear control , walking , robots , hybrid model , autonomous mobile robots </S>

1409.0110

form factors ( ffs ) of specimens of various types , from nuclear compounds to biological tissues , have been widely investigated in the past and also recently in different branches of science , since they provide detailed information concerning the electric charge distribution of the specimens @xcite . by systematically measuring ffs , information about the electric charge distribution of targetshas been extensively extracted from scattering experiments @xcite .atomic ffs , especially , play an important role in different topics of applied science such as : radiation absorption in shielding and medical diagnostics @xcite , structural factors of crystals in crystallography @xcite , and image contrast in transmission electron microscopy @xcite . during the last decades ,extensive amount of numerical calculations on nonrelativistic ffs ( nffs ) and relativistic ffs ( rffs ) for atoms and ions have been carried out , mostly by using hartree - fock ( hf ) wave - functions .such numerical calculations can be found in several tables @xcite .simultaneously , there has been many attempts toward analytical evaluations of atomic ffs .for instance , belkic reported several works on analytical atomic ffs including both the hydrogen - like and clementi - roetti wave functions for multi - electron atomic systems @xcite .in addition , partial analytical evaluations of atomic ff has been reported , for instance , in ref .@xcite and also eq .( 4.4.5.5 ) in ref .@xcite .in contrast , here , we show that it is possible to obtain an analytical expression for the atomic ff , alternatively , by using recent mathematical results : we derive an analytical expression for the nff based on the independent particle model ( ipm ) constructed from screened hydrogenic wave functions .to derive such an expression , we make use of the jacobi - anger expansion that yields a multipole decomposition of the nff , with a full separation of the angular and radial parts .we then derive an analytical form for the radial part by means of recently discovered solutions of integrals involving the product of bessel functions and associated laguerre polynomials @xcite .such analytical solutions of atomic ffs based on multipole decomposition , to the best of our knowledge , have not been presented in the literature yet .even though approximative , ipm is widely used in many fields of science in order to grasp basic properties of the system under investigation ( e.g. , in atomic and nuclear physics ) and is always presented in textbooks as a first approach to describe many - body systems . due to this and to the wide applicability of ffs , we state that our analytical expression has a fundamental value . besides , such an analytical expression might have also computational value , as remarked in @xcite .first , numerical calculations of ff require specific programs ( e.g. cowan @xcite ) and programming capabilities , whereas an analytical approach can be implemented by nearly anyone in any symbolic software ( e.g. mathematica ) .second , numerical calculations might involve problems due to rapid oscillation of the integrand at high energies , due to high multipoles , and are approach dependent .therefore , they need to be checked for each element and energy value . on the other hand ,an analytical formula is general and valid for any element .third , numerical values of the ff in the literature for specific momentum - transfer - grids are limited and dependent on the particular technique of interpolation , while the analytical ff function can generate ff values for any selected momentum transfer . in thisregard , empirical formulae for the ff have been obtained and largely used in the literature ( see , for instance , refs .@xcite , and also international table of crystallography ( itc ) @xcite , where the atomic ff is analytically expressed by means of four gaussians whose parameters are obtained from least square fitting of numerical values ) . an analytical formula that can provide easy access to reliable ff values for a wide linear momentum rangecan thus be very profitable for users , especially for experimentalists .the range of validity of our analytical expression is assessed by comparing the analytically obtained ffs with the ones obtained within the hf method . as a use case example, we apply our analytical expression for the atomic ff to rayleigh scattering by neutral atoms @xcite . the differential cross section ( dcs ) for rayleigh scattering off some selected neutral atoms is here calculated in two ways : i ) by using our analytical expression for the ff based on ipm , and ii ) by using the ff obtained from single - configuration hf numerical calculations . both resultsare compared with experiments and other numerical calculations presented in the literature .finally , it is shown how our analytical formula for the evaluation of the atomic ff can be implemented in configuration interaction ( ci ) numerical algorithms , extended to a relativistic framework , and applied to structural crystallography .si units are used throughout this article .within the independent particle approximation ( ipa ) , the total wave function of the atom , @xmath0 , is considered as product of @xmath1 one - electron spin orbitals in the form of one or more slater determinants .the one - electron spin orbitals can be hydrogenic or non - hydrogenic for a self - consistent potential generated by the electrons and nucleus .the hf method , which is also called the self - consistent field ( scf ) method , is based on ipm . in hydrogenic ipa , which we refer to as ipm , the total wave function is expressed by means of a slater determinant constructed from hydrogenic solutions of the schrdinger equation ( i.e. , coulombian wavefunctions ) @xcite . in ipm , only the electron kinetic energies and the electron - nucleus coulomb interaction terms are retained in the hamiltonian .however , electron - electron interactions are partially taken into account by using an effective nuclear charge , for each electronic orbital .the effective nuclear charge thus corresponds to the nuclear charge as seen by a given electron due to the screening of the other electrons . in what follows, the effective nuclear charge of the @xmath2th orbital will be denoted by @xmath3 . in our analytical method , values of @xmath3are directly taken from tabulation of clementi @xcite , which are currently available in the website of webelements @xcite .clementi computed the self - consistent - field ( scf ) function for atoms with a minimal basis set of slater - type orbitals .the orbital exponent of the atomic orbitals , @xmath4 , are optimized as to ensure the energy minimum . with such analysis, they obtained simple and accurate rules for the electronic screening constant , @xmath5 .their rules accounts also for the screening due to the outside electrons in comparison to the slater s rule @xcite .finally they obtained effective nuclear charge with formula : @xmath6 , where @xmath7 is the atomic number . here, we shall use such @xmath3 , since coulombian orbitals are well approximated by slater - type orbitals @xcite .the electronic charge distribution of atoms can be written as @xmath8 where @xmath9 is the differential solid angle related to the variable @xmath10 , @xmath1 is the number of electrons and @xmath0 is the @xmath1-body wave function of the atom . as evident from eq ., we take the distribution @xmath11 to be normalized to @xmath1 when integrated over @xmath12 . within ipm , the electronic charge distribution can be expressed as @xmath13 where @xmath14 are hydrogenic radial wave functions that have the form @xcite @xmath15 here , @xmath16 and @xmath17 are the principle and orbital angular momentum quantum numbers respectively , while @xmath18= is the normalization coefficient , @xmath19 is the bohr radius of hydrogen atom and @xmath20 is the associated laguerre polynomial of degree @xmath21 . in writing eq ., we have neglected cross term of the type @xmath22 which arise due to the fact that radial coulombic orbitals with different effective nuclear charge are not orthogonal to each other .for spherically symmetric charge distributions , the ff is given by @xcite @xmath23 where @xmath11 is the target charge distribution , @xmath24 is the modulus of the momentum transferred in an elastic scattering , @xmath25 is the reduced planck constant , and @xmath26 is the scattering angle ( the angle between the incoming and outgoing projectile directions ) .moreover , @xmath27 is the modulus of the linear momentum of the scattering projectile . by rewriting @xmath28 appearing in eq . in exponential form and by making use of the jacobi - anger expansion , we can re - express the ff in the simple form @xmath29 where the summation over @xmath30 is restricted to odd numbers , as explicitly denoted .the radial part of the ff ( @xmath31 ) , which will be hereinafter referred to as `` radial form factor '' , is defined as @xmath32 the simple form for the ff in eq .allows for a few physical considerations .firstly , being @xmath33 the bessel function of lowest order included in the summation , for @xmath34 one may use the equivalence @xmath35 and therefore the radial ff would be @xmath36 . consequently , the ff would simply equal @xmath1 , as it is expected to be for a point - like target whose charge is @xmath1 times the electronic charge .this is physically plausible , since for @xmath34 the resolution of the probe ( projectile ) is not enough to resolve the target structure or , in other words , the target is `` seen '' as point - like by the probe .secondly , we can interpret the summation in eq . as a multipole expansion of the ff , where each multipole is specified by the integer @xmath30 .such a multiple expansion can be used with any model and allows to calculate the ff with the necessary accuracy by setting the maximum allowed multipole .the angular and radial parts of the ff in this expansion are fully separated . whereas the angular part is expressed analytically ,the radial part is in integral form .below we shall see that , within ipm constructed from screened hydrogenic wave - functions , the radial part can also be expressed analytically . in this section , we shall proceed to evaluate the radial ff within the ipm .such quantity shall be denoted by @xmath37 and is given by eq . with the replacement @xmath38 .using howell s formula ( see ref .@xcite or eq .( 1 ) of ref .@xcite ) , the square of the associated laguerre polynomial in eq . is decomposed into a sum of associated laguerre polynomials .after such a decomposition , the function inside the integral reduces to a product of a bessel function and an associated laguerre polynomial .alassar recently derived the result for integrals involving such functions in terms of hypergeometric function ( see eq .( 6 ) of ref .@xcite ) . by thus using alassar s result and eq ., we find the following analytical expression for the ff : @xmath39 where @xmath40 is the gaussian hypergeometric function . here@xmath41 and @xmath42 are coefficients that can be directly traced back to howell s and alassar s formulas @xcite , and are given by @xmath43 and @xmath44 the gaussian hypergeometric function in eq .can be evaluated analytically ( see eqs .( 15.3.19 ) and ( 15.4.10 ) in ref .equation is the main result of this article : by setting the orbital effective nuclear charges , the electronic configuration , the projectile wave - vector and the scattering angle , the atomic ff is analytically obtained .for example , in fig .[ fffig ] , we show the ffs of some selected elements as obtained from .the ffs are plotted against @xmath45 , as usual in crystallography , where @xmath46 .ipm s ffs are compared with hf s . for light atoms, ffs obtained within ipm are hardly distinguishable from the ones obtained within hf .as the atomic number increases , the difference between the two calculations is , as expected , more significant .this is due to the fact that the cross terms of the type @xmath22 , which have been here neglected in the calculation of the radial charge density , have non - negligible contribution in medium- and high-@xmath7 regime when compared to the direct terms . as an example , below we shall apply eq . to evaluate the dcs for rayleigh scattering by some selected neutral atoms .[ tab : i ] lrrrrrrrrrrr + & & & + + & & & & & & & & & + + @xmath47&@xmath48&@xmath49&@xmath50&@xmath51&@xmath52 & & @xmath48&@xmath49&@xmath50&@xmath51&@xmath52 + + @xmath53 & - & - & 3.432[-7]&3.339[-7]&3.344[-7 ] & & - & - & 1.460[-10]&1.482[-7]&1.457[-10 ] + @xmath54 & - & - & 0.133&0.120 & 0.115 & & @xmath55 ( 0.00035)&0.0165 & 0.0169&0.0166 & 0.01621 + @xmath56 & 1.53 ( 0.13 ) & 1.560 & 1.365&1.387 & 1.519 & & - & - & 0.106&0.097 & 0.095 + @xmath57 & 1.69 ( 0.15 ) & 1.745 & 1.517&1.543 & 1.648 & & 0.127 ( 0.006 ) & 0.132 & 0.124&0.121 & 0.112 + @xmath58 & 4.20 ( 0.37 ) & 3.845 & 5.206&5.364 & 4.188 & & @xmath59 ( 0.028 ) & 0.665 & 0.562&0.591 & 0.480 + @xmath60 & 4.90 ( 0.44 ) & 5.019 & 6.045&6.211 & 5.005 & & 0.709 ( 0.035 ) & 0.752 & 0.622&0.665 & 0.553 + @xmath61 & 5.82 ( 0.50 ) & 6.045 & 6.879&7.087 & 6.101 & & 0.759 ( 0.038 ) & 0.824 & 0.697&0.734 & 0.632 + @xmath62&18.40 ( 1.66)&20.468 & 19.054&20.280 & 16.922 & & 1.92 ( 0.10 ) & 1.93 & 2.27&2.560 & 1.90 + @xmath63&20.79 ( 1.83)&22.853 & 21.683&23.055 & 18.509 & & 2.38 ( 0.12 ) & 2.27 & 2.64&3.009 & 2.13 + + +during the last decades , calculations of rayleigh scattering amplitudes have been widely performed , both within relativistic and nonrelativistic frameworks @xcite . up to now , there have been two main methods to perform such calculations , namely relativistic second - order s - matrix ( sm ) and ff approaches @xcite .sm approach is more accurate than ff , and can provide rayleigh amplitudes as accurate as 1% , for a wide range of photons energies ( approximately from 50 kev to 1.4 mev @xcite ) .nevertheless , since precise sm calculations are computationally very demanding , ff approaches are very much used . a comparison between ff and sm approachescan be found elsewhere and will not be made here in detail ( for details , see refs .we briefly mention that , although both sm and ff approaches have their own limitations depending on the energy of the incident photon , atomic number , atomic shell , and scattering angle , they provide dcs in good agreement with experimental data over limited intervals of these parameters @xcite . more specifically , ff approach can be used provided that the photon energy be much larger than the atomic binding energies but much smaller than the electron rest - mass energy @xcite .more accurate rayleigh amplitudes for a wider photon energy range can be achieved within the ff approach by considering a modified form factor ( mff ) and by considering the anomalous scattering factors usually denoted by @xmath64 and @xmath65 @xcite .it must be underlined that both sm and ff approaches treat the target as an isolated atom ( isolated atom approximation ) , with the electronic wave - functions being obtained within ipa .this restricts their use in situations where electron correlations are important , such as at very low photon energy and when the atom is influenced by the environment , as it might be the case in solids or in a plasma . for coherent scattering by several atoms in a unit cell ,the more general bragg - laue ( bl ) or thermal diffuse ( td ) diffraction laws must be used , which correspond to the cases where the bragg planes are aligned or misaligned , respectively .both bl and td diffraction laws depend on the atomic form factor through the crystal structure factor @xcite . within the ff approach ( or ff approximation ) , the dcs for elastic scattering of photons by an isolated atom is given by @xcite @xmath66 where @xmath67 is the thomson dcs , and @xmath68 is the classical electron radius .below , we shall obtain the dcs for rayleigh scattering by substituting into . in order to assess the validity of our analytical approach, we shall also numerically calculate the dcs by evaluating the ff directly from eqs . , and , using wave functions obtained from single - configuration hf calculations , as implemented in cowan s code @xcite .we checked that hf and relativistic hf ( so called _ dirac - hartree - fock _ ( dhf ) ) calculations give the same results for the ff , within few percent .we also checked that the influence of the breit - interaction in the calculation of the ffs is negligible .we calculated the dcs for rayleigh scattering off some selected elements in the ground state , for scattering angle @xmath69 , for energies 22.1 kev and 59.54 kev .our analytical nonrelativistic dcss ( hereinafter referred to as anff ) and hf results are shown in table [ tab : i ] and compared with sm calculations of kissel @xcite , numerical nff calculations of hubbell @xcite ( hereinafter referred to as nnff ) , and recent measurements of mandal @xcite and kumar @xcite . as can be seen from table [ tab : i ] , for the energy value 22.1 kev , anff cross sections are inside the experimental error bars for all elements except for @xmath63 . in comparison ,all sm cross sections are inside the error bars except for @xmath62 and @xmath63 ( for @xmath58 , the cross section is in the border ) . on the other hand , nnff and hfcalculations are , in general , outside the experimental error bars and are in excellent agreement with each other .another point which is worth to notice is that in the case of @xmath58 , our anff cross section describes the measurement remarkably better than all other calculations .comparison between anff and hf values shows that anff results agree better with experiments than hf results for photon energy @xmath70 kev .the reason for this might be : i ) in hf calculations , all parameters are calculated self - consistently . on the other hand , in the ipmthe effective nuclear charge is given as external parameter , which is calculated independently .this difference in procedure might be the reason for which our analytical formula is favorable with respect to hf ; ii ) we must remember that the application of the ff to the dcs for light scattering is itself an approximation .thus , the accuracy of ff , via either analytical or numerical methods , is not directly responsible for the discrepancies between experimental data and theoretical value . for the energy value 59.54 kev ,sm calculations have in general better agreement with experiments than other calculations .this is not unexpected , since sm calculations numerically include the full contribution of the rayleigh scattering second order amplitude , while nnff , hf and anff cross sections include only the leading part of it @xcite .we would like here to outline few further theoretical analysis and applications which could be performed using the presented results : \i ) our nonrelativistic analytical treatment can be extended to the calculation of rff by using dirac hydrogenic wave functions . in fact , radial components of dirac hydrogenic wavefunctions can be expressed in terms of associated laguerre polynomials . by repeating the same analysis showed in this article for each component , one can derive the relativistic equivalent of eq . .relativistic effects in rayleigh scattering off atoms are nonetheless expected to be small , as briefly discussed in sec .[ sec : ray ] ( see also ref .@xcite , where relativistic effects in ff calculations are discussed to be not more than few percent , when momentum transfer is not very large ) .\ii ) the presented formalism may be directly generalized to ci numerical calculations .the ci atomic state function ( asf ) is a linear combination csfs which , for instance , might be taken as screened hydrogenic wave functions .accordingly , the ff related to the asf is simply given by ( neglecting cross terms ) @xmath71 where @xmath72 are the ffs presented in eq . for the @xmath2th csf and @xmath73are determined numerically by diagonalizing the hamiltonian matrix .iii ) in structural crystallography , the structure factor @xmath74 , which may be written as a linear combination of ffs of the atoms in the unit cell @xcite , is used to solve the crystallographic phase problem , and , therefore , to determine the crystal structure . todaythe number of solved structures ( inorganic , organic , metallorganic , biological tissues ) is approaching one million and they are deposited in various databases @xcite .the accuracy of the structural model is estimated by the crystallographic index , @xmath75 , and then verified by means of chemical validation .presently the most popular atomic ffs used in crystallography are listed in the international tables for crystallography @xcite .the analytical expression may be used as a possible alternative .\iv ) our analytical expression for the atomic ff can be applied for studying other electrodynamical processes , such as electron scattering off atoms @xcite . in forthcoming articles, we will refine the presented results by including the cross terms which have been neglected in the calculation of the radial charge distribution .this work was supported by the research council for natural sciences and engineering of the academy of finland . all authors gladly thank prof .paul indelicato , prof .nivaldo lucio speziali , and prof .carmelo giacovazzo for discussions about possible applications of the presented results in structural crystallography . f.f . acknowledges support by fundao de amparo pesquisa do estado de minas gerais ( fapemig ) and by conselho nacional de desenvolvimento cientfico e tecnolgico ( cnpq ) .f.f . acknowledges hospitality of the nel institute in grenoble ( france ) , where part of the work has been carried out .p.a . acknowledges support by the german research foundation ( dfg ) .j.p.s . and p.a .acknowledge support by fundao para a cincia e a tecnologia ( fct ) in portugal , through the projects no .pestoe / fis / ui0303/2011 and no . ptdc / fis/117606/2010 .

<S> atomic form factors are widely used for the characterization of targets and specimens , from crystallography to biology . by using recent mathematical results , here </S><S> we derive an analytical expression for the atomic form factor within the independent particle model constructed from nonrelativistic screened hydrogenic wavefunctions . </S><S> the range of validity of this analytical expression is checked by comparing the analytically obtained form factors with the ones obtained within the hartee - fock method . as an example </S><S> , we apply our analytical expression for the atomic form factor to evaluate the differential cross section for rayleigh scattering off neutral atoms . </S>

1310.3587

this work was supported by nsfc , china ( grant no . 11205048 ) .

<S> similar to the superradiant effect in reissner - nordstrm black hole , a charged scalar field can be amplified when impinging on the charged black hole in string theory . according to the black - hole bomb mechanism , the mass term of the incident field can works as the reflecting mirror , which may trigger the instability of black hole . </S><S> we study the possible instability triggered by superradiant effect and demonstrate that the charged black hole in string theory is stable against the massive charged scalar perturbation . </S><S> the reason is that there is no trapping potential well in the black hole exterior which is separated from the horizon by a potential barrier and there is no bound states in the superradiant regime . </S><S> the stability tests of black holes become an important topics since the initial work of regge and wheeler @xcite . </S><S> the schwarzschild black hole is well known to be stable @xcite . </S><S> the studying of stability for the charged or the rotating black holes becomes involved because of the significant effect of classical black holes , i.e. superradiance . </S><S> if an incident bosonic wave whose frequency lies in the superradiant regime , the wave scattered by the event horizon get amplified @xcite . </S><S> this classical effect allows extracting rotational energy efficiently from the rotating black holes or extracting the coulomb energy from charged black holes @xcite . </S><S> the stability analysis of these types of black holes should be adequately addressed . </S><S> press and teukolsky @xcite proposed to use the superradiance to built the black - hole bomb by adding an reflecting mirror outside the black hole ( see also @xcite for a recent studies on this topic ) . in this mechanism </S><S> , the amplitudes of superradiant modes trapped in the potential well between the mirror and the event horizon will grow exponential . later </S><S> , it is found that the mass term of the scalar field or the boundary of anti de - sitter ( ads ) spacetime can play the roll of the reflecting mirror @xcite . in these cases , </S><S> the superradiance of the scalar field perturbation will lead to the instability of the black hole . </S><S> recently , it is proved by hod in @xcite that the existence of a trapping potential well outside the black hole and superradiant amplification of the trapped modes can not be satisfied simultaneously . </S><S> this means that the reissner - nordstrm ( rn ) black holes are stable under the perturbations of massive charged scalar fields . however , whether all of the charged black holes are stable is still an open question worth studying . in this work , we will study the stability of the charged black hole in string theory . </S><S> similar to the superradiant effect in reissner - nordstrm black hole , a charged scalar field can be amplified when impinging on the charged black hole in string theory . according to the black - hole bomb mechanism </S><S> , the mass term of the incident field can effectively works as the reflecting mirror , which may trigger the instability of black hole . </S><S> however , we will demonstrate that the charged black hole in string theory is stable against the massive charged scalar perturbation . by carefully studying the behavior of the effective potential outside the horizon </S><S> , we will show that there is no trapping potential well in the black hole exterior which is separated from the horizon by a potential barrier . </S><S> so there is no bound states in the superradiant regime which can leat to the instability of charged stringy black hole . </S><S> we consider the static charged black holes in low energy effective theory of heterotic string theory in four dimensions . </S><S> besides the einstein - hilbert term , the action also includes the contributions from the dilaton field and the maxwell s field . </S><S> the charged black hole in string theory which is firstly found by gibbons and maeda in @xcite and independently found by garfinkle , horowitz , and strminger in @xcite a few years later is described by the metric @xmath0 and the electric field and the dialton field @xmath1 the parameters @xmath2 and @xmath3 are the mass and electric charge of the black hole respectively . </S><S> the event horizon of black hole is located at @xmath4 . </S><S> the area of the sphere of the charged stringy black hole approaches to zero when @xmath5 . </S><S> therefore , the sphere surface of the radius @xmath5 is singular . when @xmath6 , this singular surface is surrounded by the event horizon . </S><S> we will consider the black hole with the parameters satisfying the condition @xmath6 in this paper . </S><S> when @xmath7 , the singular surface coincides with the event horizon . </S><S> this is the case of extremal black hole . </S><S> we start with analysing the scalar field perturbation in the background of the charged stringy black hole . </S><S> the dynamics of the charged massive scalar field perturbation is governed by the klein - gordon equation @xmath8\psi=0\ ; , \end{aligned}\ ] ] where @xmath9 and @xmath10 denote the charge and the mass of the scalar field . by taking the ansatz of the scalar field @xmath11 , where @xmath12 is the conserved energy of the mode , </S><S> @xmath13 is the spherical harmonic index , and @xmath14 is the azimuthal harmonic index with @xmath15 , one can deduce the radial wave equation in the form of @xmath16 where we have introduced a new function @xmath17 and the potential function is given by @xmath18\;. \end{aligned}\ ] ] firstly , we want to demonstrate that the classical superradiance is present for scalar field perturbation in charged stringy black hole </S><S> . in particular , we shall study the asymptotic solutions of the radial wave equation near the horizon and at spatial infinity with the appropriate boundary conditions and obtain the superradiant condition of the charged scalar field . to this end </S><S> , it is convenient to define the tortoise coordinate @xmath19 by the equation @xmath20 , and introduce the new radial function as @xmath21 . </S><S> the domain of the tortoise coordinate @xmath19 is @xmath22 corresponding to the domain @xmath23 of coordinate @xmath24 . in terms of the tortoise coordinate and </S><S> the new radial function , the radial wave equation can be rewritten as @xmath25 with @xmath26 it can be easily checked that when @xmath27 the above equation reduced to the regge - wheeler equation @xcite as expected . it is easy to obtain the asymptotic behavior of the potential @xmath28 as @xmath29 so , the radial wave equation has the following asymptotic behavior when @xmath30 @xmath31 with the parameter @xmath32 . by calculating the radial flux of the scalar field modes respectively , one can see that the modes of the form @xmath33 and @xmath34 are outcoming from the event horizon and ingoing to the event horizon respectively , while the modes of the form @xmath35 correspond to the outgoing flux at the spatial infinity . </S><S> this boundary conditions then correspond to the flux coming from the event horizon of the black hole which is partially reflected back to the black hole and partially sent to the spatial infinity . </S><S> given any two linearly independent solutions @xmath36 and @xmath37 of the radial equation ( 10 ) , their wroskian is constant , i.e. @xmath38 because the potential @xmath28 is real , the asymptotic solution ( 14 ) and its complex conjugate are linearly independent to each other . evaluating the wronskian for the asymptotic solution ( 14 ) and its complex conjugate at the horizon and spatial infinity respectively , we can get @xmath39 now , we can see that if @xmath40 , we have @xmath41 . </S><S> this means that an incident wave from the horizon is reflected back to the horizon with an increased amplitude . </S><S> this phenomenon is known as superradiance . from @xmath40 </S><S> , we can get the condition to occur the superradiance @xmath42 with @xmath43 being the electric potential at the horizon . </S><S> in fact , this condition has already been obtained in @xcite , where the superradiant effects for the dilaton black holes are studied . </S><S> now we shall analysis whether this superradiant effect will lead to the instability of the charged stringy black hole . </S><S> the purpose can be achieved by studying whether there exists a trapping potential well outside the horizon that can trap the superradiant modes . if there exists a trapping potential well , superradiant amplification of the bound state in the potential well will trigger the instability of black hole . </S><S> otherwise , the black hole is stable although the superradiance is present . as a matter of convenience </S><S> , we want to introduce a new radial function as @xmath44 . </S><S> the radial wave equation ( 5 ) can be rewritten as @xmath45 with the effective potential @xmath46\;. \end{aligned}\ ] ] now , we analyze the behavior of the effective potential outside of the horizon . from eq.(20 ) , after some algebra , we can get the derivative of the effective potential as @xmath47\;. \end{aligned}\ ] ] it is convenient to define a new variable @xmath48 , in terms of which the derivative of the effective potential can be rewritten as @xmath49 with @xmath50 the behavior of the effective potential can be roughly described by the properties of the roots of @xmath51 . </S><S> we now analysis the case of the nonextremal black hole . obviously , @xmath52 is a root of @xmath53 . </S><S> the other three roots are denoted as @xmath54 . </S><S> it is well - known that the relations between the roots and the coefficients of the cubic equation @xmath55 are given by @xmath56 for the scalar field satisfying the superradiant condition and the bound state condition , the frequency @xmath12 of the mode satisfies @xmath57 then , it is obvious that @xmath58 this implies that @xmath59 note also that @xmath60 and @xmath61 this implies that there exists at least one maximum point in the region @xmath62 and at least one maximum point in the region @xmath63 . </S><S> we denote this two maximum points as @xmath64 and @xmath65 respectively </S><S> . then we have @xmath66 for the nonextremal black hole , @xmath67 . </S><S> then , we have @xmath68 this implies @xmath69 this means that in the superradiant regime the effective potential @xmath70 has only one maximum outside the event horizon . </S><S> this implies that there is no trapping potential well outside of the event horizon which is separated from the horizon by a potential barrier . for different @xmath12 . </S><S> the parameters are given by @xmath71 and @xmath72 . from top to down , the three curves correspond to @xmath73 and @xmath74 respectively.,scaledwidth=40.0% ] in fig . 1 , we have plotted the shape of the effective potential @xmath75 given in eq.(15 ) for different @xmath12 . the analytical conclusion for the nonextremal black hole case </S><S> is explicitly shown in this figure . </S><S> the mass of the scalar field is never able to generate a potential well outside of the horizon to trap the superradiant modes . </S><S> thus , there are no meta - stable bound states of the charged massive scalar field in the superradiant regime . in other words , the superradiance in the nonextremal </S><S> charged stringy black hole can not trigger the instability . for the case of the extremal black hole , @xmath7 . </S><S> the effective potential becomes @xmath76\;. \end{aligned}\ ] ] the derivative of the effective potential is then given by @xmath77 with @xmath78 the coefficient @xmath79 is still unchanged , i.e. @xmath80 in the superradiant region . </S><S> the root of @xmath53 is simply given by @xmath81 we shall analysis the following two situations in detail . </S><S> case i : @xmath82 or @xmath83 the effective potential behaves @xmath84 but @xmath85 , i.e. the root of @xmath53 locates at the non - physical regime @xmath86 . </S><S> this implies that there is neither an maximum point nor an minimum point outside the horizon . </S><S> in the black - hole exterior , the effective potential will gradually bring down to a finite value . for different @xmath12 . </S><S> the parameters are given by @xmath87 and @xmath72 . from right to left , the three curves correspond to @xmath88 and @xmath74 respectively.,scaledwidth=40.0% ] in fig . </S><S> 2 , we have plotted the shape of the effective potential given in eq.(33 ) for different @xmath12 . </S><S> the parameters are selected to satisfying the extremal condition and @xmath83 simultaneously . </S><S> one can see that , outside the horizon , there exists neither a potential barrier nor a potential well . in this case , the superradiance can not lead to the instability . </S><S> case ii : @xmath89 or @xmath90 the effective potential behaves @xmath91 but @xmath92 , i.e. there is a root of @xmath53 in the region of @xmath62 . </S><S> this implies that the effective potential have only one maximum point outside the horizon , i.e. there is only a potential barrier outside of the horizon . for different @xmath12 . </S><S> the parameters are given by @xmath87 and @xmath93 . from top to down , the three curves correspond to @xmath88 and @xmath94 respectively.,scaledwidth=40.0% ] in fig . </S><S> 3 , we have plotted the effective potential for different @xmath12 which satisfy the condition @xmath90 . </S><S> the shape of the effective potential in this case is very similar to that of the nonextremal black hole . </S><S> so for the same reason , in the present case , the superradiant modes can not be trapped as well and the black hole is also stable . in summary </S><S> , we have firstly shown that the classical superradiance phenomenon presents in the charged stringy black holes for the charged scalar field perturbations . </S><S> the superradiant condition is also obtained by analyzing the asymptotic solutions near the horizon and at the spatial infinity , which is similar to that of rn black hole . </S><S> then we investigate the possibility of instability triggered by the superradiance . </S><S> it is shown by analyzing the behavior of the effective potential that for both the nonextremal black holes and the extremal black holes there is no potential well which is separated from the horizon by a potential barrier . </S><S> thus , the superradiant modes of charged massive scalar field can not be trapped and lead to the instabilities of the black holes . </S><S> this indicates that the extremal and the nonextremal charged black holes in string theory are stable against the massive charged scalar field perturbations . at last </S><S> , we should note that although the mass of the scalar field can not provide an effective potential well outside the black hole , one can still make the black hole unstable by placing a reflecting mirror around the black hole @xcite . </S><S> it would be interesting to study the behavior of the scalar field in this black hole - mirror system in the future work . </S>

0705.0309

the development of spintronics devices @xcite has motivated a large body of research on diluted magnetic semiconductors @xcite with the ultimate aim of creating materials with curie temperatures ( @xmath7 ) above room temperature .this ambitious goal can only be achieved by a detailed understanding of the underlying mechanisms that govern the behavior of currently available dms .most theoretical approaches to study these materials start with one of two extreme regimes : _( i ) _ the limit of high mn doping in which holes are directly doped into the valence band and , thus , are uniformly distributed in the sample ( vb scenario ) @xcite and _ ( ii ) _ the limit of very low mn doping in which holes are electrically bound to the impurity cores and an impurity band develops due to wave function overlap as the number of holes increases ( ib scenario).@xcite researchers using the vb limit claim that it is valid for all the relevant dopings , namely @xmath0@xmath8@xmath9 in ga@xmath10mn@xmath11as , and some experimental results support their view.@xcite however , a similar claim is advanced by the groups promoting the ib scenario , i.e. that the ib exists up to the largest value of @xmath0 that has been reached experimentally ( @xmath12 ) .this view also appears supported by the analysis of some experimental data.@xcite to solve this apparent puzzle , it is very important to study theoretically the dms problem using unbiased techniques that provide reliable estimations for the value of @xmath0 where the ib to vb crossover takes place .such unbiased approaches could be provided by numerical techniques : in fact , the mc and dmft methods have already been applied to a variety of phenomenological models for the dms.@xcite these previous studies have been able to determine a crossover between the vb and ib behaviors , but only as a function of increasing values of the @xmath1-@xmath2 exchange @xmath3 . however , most experimental results appear to indicate that the realistic @xmath3 for ( ga , mn)as is approximately 1 ev,@xcite which corresponds to the weak coupling regime in which no ib is generated by @xmath3 alone .in fact , recent results obtained applying mc techniques to a six - orbital microscopic model , in which both the correct lattice geometry and the spin - orbit interactions were considered , indicate that ( ga , mn)as is indeed in the vb regime for @xmath13.@xcite in addition , dmft techniques , which allow for the study of the very diluted ( @xmath14 ) regime , have shown that for values of @xmath3 in the weak coupling region , an ib never develops as a function of @xmath0.@xcite however , experiments based on electron paramagnetic resonance,@xcite infrared spectroscopy,@xcite and magnetization measurements@xcite of the electronic structure of one mn ion doped in gaas have actually shown the existence of a shallow hole state with binding energy @xmath15@xmath16@xmath17 mev centered at the @xmath18=@xmath19 mn ion .moreover , analytical studies indicated that @xmath15 has contributions from @xmath20 the spin - dependent @xmath1-@xmath2 hybridization and the coulomb attraction between the hole and the @xmath21 trapping center.@xcite when additional mn ions are added , the wave functions of the bounded holes will start to overlap and an ib will develop . further increasing @xmath0 should widen the ib , locating it closer to the vb and eventually a regime of complete hybridization with the holes doped into the vb is expected to occur .thus , it is clear that a crossover from the ib to the vb regime should take place in ( ga , mn)as as a function of @xmath0 . in this paper, it will be argued that an ib - vb crossover will be missed in theoretical studies of materials with a weak @xmath3 if the coulomb attraction is disregarded , while materials with very strong @xmath3 will be in the ib regime regardless of doping .in fact , here we explicitly show that by the simultaneous consideration of @xmath3 and @xmath6 in the formalism , the experimentally observed transition from ib to vb with increasing @xmath0 can be understood .the organization of the paper is the following : in section [ section2 ] the non - magnetic interactions in dms are described ; the model used and the dmft technique are presented in section [ section3 ] ; in section [ section4 ] the results , including mc simulations , are discussed , and section [ section5 ] is devoted to the conclusions .as remarked in the introduction , most of the numerical work on dms has been performed on models that focused on the role of the spin dependent @xmath1-@xmath2 exchange @xmath3 interaction between the spins of the localized impurities and the doped holes.@xcite this is certainly sufficient to capture qualitatively many of the properties of these compounds , including the generation of ferromagnetism .however , non - magnetic interactions between holes and impurities must be considered in order to improve the quantitative agreement with experiments .this additional potential term in the model has been generally referred to as `` chemical disorder''(@xmath6),@xcite and it summarizes all the non - magnetic interactions between the localized impurities and the holes . in this context , tworzydlo @xcite used a short range potential ( less than nearest - neighbors range ) with a square - well form of depth @xmath22 , and considered both positive ( repulsive ) and negative ( attractive ) values of @xmath22 .the potential was introduced to explain an apparent @xmath0-dependence of the @xmath1-@xmath2 exchange in cd@xmath10mn@xmath11s .dietl @xcite recently used the same approach to address apparently contradictory experimental results for ga@xmath10mn@xmath11n .he also pointed out @xcite that this kind of extra potential term leds to a chemical shift in the standard impurity language , or to a valence - band offset in the alloy nomenclature , and that @xmath3 and @xmath6 are actually related @xcite through the expression @xmath23 where @xmath24 is an effective correlation energy for the 3@xmath2 shell , and @xmath25 is its energetic position with respect to the top of the valence band .however , the value of @xmath6 is not easy to determine and , thus , it has been added as an extra free parameter by some authors ( with @xmath6 allowed to take both positive and negative values).@xcite other efforts focused just on the attractive coulomb interaction between the holes and the impurities.@xcite only some of the previously mentioned investigations have attempted to study the effects of the coulomb attraction at finite @xmath0 with unbiased techniques .the authors of ref .[ ] studied the case of a single mn impurity , considering the long - range coulomb potential supplemented by a central cell correction with a gaussian or square - well shape , that is routinely introduced in calculations of bound state energies for impurities in semiconductors.@xcite for higher dopings , it is believed that the most important coulombic term is the central - cell contribution since the long - range potential is screened . in ref . , the coherent potential approximation ( cpa ) , very similar in spirit to dmft , was applied to a single orbital model which included both the spin dependent @xmath1-@xmath2 hybridization @xmath3 and an on - site central - cell coulomb attraction @xmath6 .it was claimed that the ib - vb crossover for ( ga , mn)as using @xmath6=0.6 ev ( chosen to reproduce , in combination with @xmath3=0.8 ev , the single impurity bound state energy ) should occur for @xmath26 . in ref .[ ] , a repulsive on - site potential was added .both the repulsive and attractive cases were considered in ref .. however , these important previous efforts did not present a systematic analysis of results as a function of @xmath3 , @xmath6 , and @xmath0 , which is part of the goals of the present study .in this work we apply dmft to a model that includes @xmath3 and the coulomb attraction @xmath6 .the density of states ( dos ) and @xmath27 are studied in a wide range of couplings , hoppings , carrier fillings @xmath1 , and mn concentrations @xmath0 , and estimations of the most appropriate values for different materials are made .we obtain the ib - vb crossover for a large class of dms s and show that with a suitable strength @xmath6 included , the ib regime can always be reached by decreasing the mn concentration .the spin - fermion hamiltonian used here and in several previous studies contains a kinetic @xmath28-term that describes the hopping of holes between two neighboring @xmath29 and @xmath30 lattice sites ( @xmath28 is set to @xmath31 to define the energy unit ) , an exchange interaction ( ei ) @xmath32-term that anti - aligns the carrier s spin with the magnetic moment of the impurity ( considered classical ) at site @xmath33 , and a @xmath6-term that takes into account the on - site central - cell part of the attractive coulomb potential,@xcite @xmath34 here , @xmath35 ( @xmath36 ) is the creation ( destruction ) operator for a hole with spin @xmath37 at site @xmath29 , @xmath38=@xmath39 is the hole s spin , @xmath40=@xmath41 is the classical spin of the local moment , and @xmath42 is the number of holes at @xmath33 .several details on the dmft calculations were already presented in ref .[ ] for the case v=0 , thus here only a brief summary is given and the modifications introduced by a non - zero v are remarked .dmft uses the momentum independence of the self - energy in infinite dimensions [ @xmath43@xmath44@xmath45 , @xmath46=@xmath47 @xcite and reproduces the physics of diluted correlated systems in lower dimensions.@xcite within dmft , the bare green s function @xmath48 contains all the information about the hopping of carriers onto and off magnetic ( with probability @xmath0 ) and nonmagnetic ( with probability @xmath31-@xmath0 ) sites . with ( [ ham ] ) the full green s function @xmath49 is solved by integration obtaining the result : @xmath50=@xmath51^{-1}\rangle$]+@xmath52 , where @xmath3=@xmath53 .@xcite this equation , complemented with the relation @xmath54=@xmath55 valid within the assumption of a bethe lattice,@xcite can be solved with a semicircular noninteracting @xmath56=@xmath57{(w/2)^{2}\smash{-}\omega^{2}}}/\pi w$ ] ( @xmath58=@xmath59+@xmath60 , @xmath59 is the chemical potential , and @xmath61=@xmath62 is the bandwidth ) .being spin diagonal , @xmath63 and @xmath64 are expanded in powers of @xmath65 as : @xmath66=@xmath67+@xmath68 , where @xmath69@xmath70@xmath71 , @xmath72 being the order parameter used to detect the fm transition . to linear order in @xmath72we write @xmath73=@xmath74 + @xmath75 and then @xmath76 is found from a @xmath77-th order equation , @xmath78}{[b_{\pm}+v]^{2}-j^{2 } } \!-\!(1\smash{-}x)\frac{w^{2}}{16}\frac{1}{b_{\pm}},\label{re}\ ] ] that at @xmath59=@xmath79 and with @xmath60@xmath44@xmath80 gives us the low - temperature interacting @xmath81=@xmath82/\pi$ ] for up ( @xmath83 ) and down ( @xmath84 ) spin configurations.@xcite the expression for @xmath85 : @xmath86^{2}}\right\}\nonumber\\ & { } & + ( 1-x)\frac{w^{2}}{16}\frac{q}{b^{2}},\label{qu}\end{aligned}\ ] ] leads us to an implicit equation for @xmath27 in the form : @xmath87\{[b\smash{+}v]^{2}\smash{-}j^{2}\}^{2 } -3xw^{2}b^{2}\{[b\smash{+}v]^{2}\smash{-}j^{2}\ } -2xw^{2}j^{2}b^{2}}=1,\label{tc}\ ] ] where @xmath76 is given by eq .( [ re ] ) at @xmath72=@xmath79 .the @xmath27 contained in @xmath46 can be obtained from eq.([tc ] ) numerically .let us start the discussion of results by considering the general dependence of a variety of quantities with the parameters of the model .the dos obtained from eq .( [ re ] ) at @xmath0=@xmath88 is displayed in fig .[ fig1 ] for various values of @xmath3 , @xmath72 , and @xmath6 . as observed in fig .[ fig1](a ) , the @xmath3-term alone is able to generate an ib but only if @xmath89 exceeds a critical value @xmath90@xmath70@xmath91 . at realistic couplings for ( ga , mn)as ( namely , @xmath89@xmath92@xmath93 if we assume @xmath3@xmath94@xmath28@xmath70@xmath31ev ) there is no ib generated by the @xmath3-term alone .however , with the addition of coulomb attraction , when a value @xmath95@xmath96@xmath97 is reached , then a well - defined split ib forms , as shown in fig.[fig1](b ) .no `` symmetric '' impurity band exists at high energies since the observed one is due to the carriers that are trapped in the vicinity of the core spins through the influence of @xmath6 , and are fully aligned for @xmath72=@xmath31 ( fig . [ fig1](c ) ) .the growth of @xmath89 produces asymmetric low- and high- energy impurity bands if @xmath6@xmath98@xmath79 ( fig .[ fig1](d ) ) .we have observed that the coupling strength @xmath99 for which the ib develops is a function of @xmath0 , namely the larger @xmath0 is , the larger @xmath99 becomes .thus , we used eq .( [ re ] ) to draw the phase diagram @xmath99 vs. @xmath0 at various values of @xmath6 . when @xmath6=@xmath79 the occurrence of an ib due only to the @xmath3-term requires a @xmath99@xmath94@xmath93 when @xmath0@xmath44@xmath79 , as seen in fig.[fig2](a ) . when @xmath0@xmath44@xmath79 and @xmath89@xmath100@xmath93 the addition of a potential @xmath6 leads to the relation @xmath101@xmath94@xmath93 to establish the boundary of the region where an ib develops .our calculations also show that the boundary between the ib and vb regions in the full @xmath3-@xmath0 plane just moves down by an amount @xmath102 after the introduction of the coulomb attraction .this @xmath102 is independent of @xmath0 indicating that @xmath103 as it can be seen in fig .[ fig2](a).@xcite this means that an ib will be generated by a @xmath104 if a @xmath6 such that @xmath101@xmath94@xmath105 is added .then , intuitively the effect of the addition of @xmath6 is to renormalize @xmath3 to a larger value .this result is not surprising because @xmath3 has a dual effect : _( i ) _ it induces ferromagnetism , but _( ii ) _ it also tends to localize the holes near the impurity so that they take advantage of the antiferromagnetic coupling . this last property is similar to the effect produced by the coulomb attraction @xmath6 .however , it would be expected that as @xmath0 increases and more holes are added to the system , the wave functions of the holes will start to overlap , and as the holes become delocalized the effects of @xmath6 should become less important.thus , we would expect that the crossover boundaries between the ib and vb regions indicated in fig .[ fig2](a ) should become closer to the @xmath106 curve as @xmath0 increases , instead of remaining parallel as in the figure .similar results have been observed in mc simulations .@xcite we believe that the reason for this unexpected behavior is related to the fact that here an on - site central - cell potential is being considered .this behavior can be corrected by considering a nearest - neighbor - range potential @xcite or , within the dmft framework , by considering a phenomenological on - site potential that depends on @xmath0 such as @xmath107 where @xmath108 can be roughly estimated using mott s criterion @xcite as @xmath109 with @xmath110 being the side of the cubic cell of the material and @xmath111 the bohr radius for the bound impurity . for a material such as ( ga , mn)as , which has an estimated @xmath112 , we obtain @xmath108=0.0014 . the resulting boundary between the ib and vb regions is presented in fig .[ fig2](b ) which indicates that for realistic values of @xmath3 ( 0.2w ) and @xmath22 ( 0.1w ) for ( ga , mn)as , the crossover would occur for @xmath113 . after having remarked that some paradoxes of the results can be solved by extending the size - range of the attraction or , similarly , by reducing its strength with increasing @xmath0 , here we will continue the discussion of the qualitative aspects for the case of the on - site central - cell potential .the main reason for it is to be able to compare our conclusions with previous results in the literature since an on - site potential is the only approach used in previous numerical investigations.@xcite there are still some quantitative aspects that may need the @xmath0 dependent potential of the previous paragraphs , and those will be clarified below .focusing on the on - site potential , it can be observed that even if @xmath89@xmath100@xmath99 , the ib regime can in general be reached either by increasing @xmath6 at fixed @xmath0 , or by decreasing @xmath0 at fixed @xmath6 ( see fig .[ fig2](c ) ) .while at @xmath0@xmath44@xmath79 the carriers trapped due to @xmath6 in the vicinity of each mn core spin reside in an impurity - like bound state , as @xmath0 increases the wave functions that describe the bound state at the manganeses start overlapping ( due to the combined effects of @xmath6 and @xmath3 ) producing an ib that at a critical @xmath114 merges with the vb .the renormalization condition obtained in our calculations yields an ib - vb boundary in the diagram @xmath89 vs. @xmath95 , for a fixed @xmath0 , as shown in fig .[ fig2](d ) .this boundary deviates from linear only for very small values of @xmath89 which is not a physically interesting region . according to the results in fig .[ fig2](d ) the area of the vb region is a minimum for @xmath115 and increases with increasing @xmath0 .the literature does not provide a unique value of @xmath6 for the case of ( ga , mn)as . the main reason is that the value of @xmath6 necessary to generate a bound state upon doping by one hole is a function of both @xmath3 and the bandwidth @xmath61 , as it can be observed from the results presented in table [ table ] .thus , in ref . [ ] a value of @xmath6 = 2.3 ev is determined for @xmath3=0.9 ev with @xmath116 since a luttinger - kohn energy band is used , while in ref ., @xmath6=0.6 ev is used with @xmath3=0.8 ev and @xmath61=4 ev . in both cases , @xmath6 is determined by requesting that for a single impurity doping a bound state at @xmath117= 112 mev appears as the combined result of the magnetic and coulomb interactions .our calculations indicate that the parameters of ref .provide an ib - vb crossover at @xmath118 while we recovered the value @xmath119 of ref.using the parameters that they provided .the discrepancy shows that the values assumed for @xmath61 and @xmath3 play an important role in the determination of @xmath6 and @xmath5 . the expression given by dietl,@xcite provides an estimation of the non - magnetic impurity potential that may include more than coulomb interactions .it is evaluated using experimental data . for @xmath120 @xcite with @xmath121 and @xmath3=@xmath31 ev, the ratio @xmath122=@xmath123 is obtained .the potential turns out to be repulsive @xmath6=@xmath124 ev .notice that while the estimations of @xmath6 performed for @xmath115 provides positive values , compatible with an attractive potential , the estimations at finite doping do not .as pointed out in the previous section , this indicates that it may be necessary to use an @xmath0-dependent expression for the non - magnetic interactions . the phenomenological potential proposed in eq .( [ pot ] ) will provide an ib - vb crossover at @xmath0@xmath70@xmath125 for all the attractive values of @xmath6 provided above , as seen in fig .[ fig2](b ) .we can make estimations of @xmath5 for ( ga , mn)as and for other mn doped iii - v materials as well . the value of @xmath3 is expected to be inversely proportional to the volume of the cubic cell of the material @xmath126 , according to the chemical trends , and the energy of the bound state for one mn impurity has been measured.@xcite from these data , we can estimate @xmath6 for different values of @xmath61 , with results given in table [ table ] , that also includes @xmath110 for each material and the estimated value of @xmath127 where @xmath128 with @xmath129 the electron mass and @xmath130 the luttinger parameters.@xcite then @xmath131 can be obtained from eq .( [ x0 ] ) and is also shown in the table .@xmath5 ( @xmath132 ) indicates the estimated values of the doping for which the ib - vb crossover occurs for an on - site ( @xmath0-dependent ) potential ( @xmath133 given by eq .( [ pot ] ) ) .[ table ] .[rep ] dmft calculated values of @xmath6 that produce a bound state with energy @xmath134 for the values of @xmath3 and bandwidth @xmath61 shown corresponding to the indicated dmss .the calculated doping density @xmath114 ( @xmath135 ) at which the ib / vb crossover occurs for an @xmath136independent ( dependent ) potential is listed .the ib label indicates that the material is in the ib regime at all @xmath0@xmath137@xmath138 $ ] .values of @xmath110 , @xmath111 , and @xmath131 ( see text ) for each material are also shown . [ cols="^,^,^,^,^,^,^,^,^,^",options="header " , ] it is clear that for all relevant values of @xmath0 , ( ga , mn)as is in the vb regime .the crossover , for realistic values of @xmath61 , occurs at @xmath0@xmath139@xmath9 for both on site and @xmath0-dependent potentials .thus , even including the coulomb attraction , our results indicate that the ib regime is not expected to play a relevant role in this material .a similar picture emerges for ( ga , mn)sb . in this casethe ib - vb crossover is expected to occur for such small values of impurity doping that for all practical purposes the coulomb attraction can be neglected . on the other hand, the ib regime seems to dominate the physics of ( ga , mn)n .considering @xmath3=@xmath140 ev , within our model we found that even for the largest value of w considered ( namely , @xmath61=10 ev ) @xmath89 is strong enough to generate an ib region below some finite @xmath141 , even if no coulomb attraction is considered .however , since the single hole bound energy for gan is 1.4 ev , i.e. much larger than the 0.113 ev value observed in gaas , it is clear that the coulomb - attraction term has to be incorporated . in the tablewe show the values of @xmath6 that together with @xmath3 will produce the bound state for different values of the bandwidth @xmath61 .our calculations show that with an on - site potential ( ga , mn)n will be in the ib regime for all relevant values of @xmath0 ( we studied up to @xmath142 ) .this is still true when an @xmath0-dependent @xmath6 is considered since even in the case for the largest bandwidth considered the crossover is expected to occur at @xmath143 .coulomb attraction should therefore be included to study this material . our results for ( ga ,mn)p indicate that despite the deeper position of the bound state in the gap , studies neglecting the coulomb attraction could be performed , particularly for @xmath13 . for completeness , and to compare with previous calculations,@xcite we present the @xmath27 vs. @xmath1 dependence obtained from eq .( [ tc ] ) at @xmath0=@xmath144 , for different values of @xmath3 s and no coulomb attraction in fig.[fig3](a ) . for @xmath89@xmath145@xmath99, @xmath27 is low and almost independent of @xmath1 . when @xmath89@xmath8@xmath99 , i.e. in the ib regime , @xmath27 vs. @xmath1 is semicircular with a maximum at @xmath1=@xmath146 , in agreement with previous results for one - orbital models.@xcite the behavior of @xmath27 vs. @xmath1 at different values of @xmath95 for @xmath147 is shown in fig.[fig3](b ) .comparing with the curves in part ( a ) of the figure it is clear that @xmath6 increases the effective value of @xmath3 .our results agree with ref . and confirm that an on - site square - well @xmath6 simply renormalizes @xmath3 .the dependence of @xmath27 on @xmath3 for different values of @xmath6 is shown in fig.[fig4](a ) .@xmath6 boosts @xmath27 at small and intermediate @xmath89 , while at large @xmath89 s no change is observed because within dmft the @xmath27 saturates as @xmath148 .as it will be discussed in the following section , we believe that the renormalization of @xmath3 for the physically relevant values of @xmath0 , such as the one used in our figures , is an artifact of the on - site range of the coulomb attraction and , thus , we do not expect it to play a role in enhancing the @xmath27 of real materials .hamiltonian ( [ ham ] ) was also studied here using a real - space mc technique with the mn core spins treated classically .details are not provided since the technique has been widely discussed before in the context of studies of manganites.@xcite the simulations were performed using cubic lattices with @xmath149 sites at @xmath0=@xmath93 .finite - size effects have been monitored by running some points on @xmath150 clusters .a random starting spin configuration has been selected as the starting point for each temperature @xmath151 .the spins were allowed to evolve for a total of @xmath152 mc steps , with the first @xmath153@xmath154@xmath155 steps being discarded to thermalize the starting configuration .at @xmath156=@xmath31 , and @xmath157=@xmath158=@xmath159 , a value @xmath6=@xmath31 for the on - site coulomb attraction increases @xmath27 by as much as 33% , as shown in fig .[ fig4](b ) .this agrees qualitatively with the dmft results .the figure shows clearly how @xmath6 effectively `` renormalizes '' @xmath3 .since the curve @xmath27 vs. @xmath3 for @xmath6=@xmath79 has a maximum at @xmath160 the effect of @xmath6 is to increase @xmath161 for values of @xmath162@xmath139@xmath160 , while @xmath27 decreases with @xmath6 for values of @xmath163 .although this renormalization has been previously reported,@xcite we do not believe that it will play a role in the relevant range of doping for most dms . as we pointed out in subsection [ general ] , the on - site range of the coulomb attraction induces unphysical behavior by exaggerating hole localization for values of @xmath0 for which overlap of the hole wave functions should occur .while finite range attraction can not be studied with dmft , it can be done with mc simulations but at the price of not being able to access the low doping regime at which the ib - vb crossover would be expected to occur for a material such as ( ga , mn)as . in fig .[ fig4](c ) , we present the dos obtained with mc for @xmath156=@xmath31 and @xmath6=@xmath79 for @xmath0=@xmath164 indicated by the black continuous line .the peaks are due to the finite size of the system , and each of them can be identified with the spikes that appear in the dos of a non - interacting system in the same lattice .thus , at this value of @xmath3 , there is only a vb in the dos , i.e. , the magnetic interaction is not strong enough to develop an impurity band .the position of the chemical potential @xmath59 is indicated by the black dashed line . upon adding an on - site coulomb attraction @xmath6=@xmath165, we observe that an ib develops as indicated by the red line in the figure , that has been shifted upwards along the vertical axis for clarity .this ib is due to the localization of the holes induced by the on - site potential .the chemical potential denoted by the dotted red line indicates that only states in the ib are occupied .however , when the range of the potential is increased to next - nearest neighbors , as indicated by the green dashed line in the figure , it can be seen that the ib dissapears although the intensity of the potential has not changed .this occurs because , at this large doping , the extended potential allows for a more uniform distribution of the holes .as it can be seen in the figure , the dos for @xmath6=0 and for finite extended @xmath6 have an almost perfect overlap .this shows that the use of on - site coulomb attraction potential can lead to missleading results and authors have to be cautious when using this approximation .our combined dmft - mc study shows that the coulomb attraction by acceptors needs to be considered to obtain correctly the ib - vb crossover as a function of impurity doping concentration @xmath0 in models for dms .however , for most materials we find that the crossover occurs at very low levels of doping , outside the regime in which high @xmath166 would be expected .we also find that a doping - independent on - site square - well potential acts as a renormalization of the coupling @xmath3 in an extended doping range up to @xmath142 .however , this apparent boost to the @xmath3-term at all mn dopings is unphysical , since the effect of @xmath6 should be @xmath0-dependent beyond some critical value .our mc simulations demonstrate that this @xmath0-dependence is achieved naturally by considering a longer range ( next - nearest neighbors ) square - well attraction , which is beyond the capability of the single site dmft which can deal with on - site interactions only .thus , a phenomenological @xmath0-dependent coulomb attraction was introduced . with this modification, we have shown that for ( ga , mn)as , the coulombic attraction @xmath6 influences the physics of the material only at small mn doping , i.e @xmath0@xmath139@xmath167 .this result shows that it is correct to apply theories that consider the @xmath3-term only for studying the properties , including the curie temperature , of ( ga , mn)as at the relevant values of mn concentrations @xmath0@xmath70@xmath9@xmath168 . on the other hand, we found that the coulomb attraction will play a relevant role , and should be included , in studies of mn - doped gan .summarizing , here we have shown that the addition of an attractive coulomb potential is the necessary ingredient to explain the transition from the ib to the vb regime as a function of mn - impurity doping concentration in materials for which the magnetic interaction @xmath3 is not strong enough to bind a hole .however , we find that , except for the case of ( ga , mn)n , the crossover occurs at very low doping in a regime in which high ferromagnetic critical temperatures would not be expected and , thus , the effective value of @xmath3 will not be affected . as a consequence , it is not necessary to include the coulomb attraction in the calculations .in addition , we show that an on - site attractive potential does not capture the overlap of localized hole wave - functions that should occur as a function of doping and it provides unphysical results . thus , to study materials such as ( ga , mn)n , in which the coulomb attraction is relevant , a nearest - neighbor finite range potential has to be used .we acknowledge helpful discussions with t. dietl and j. sinova . this research was supported in part by the national science foundation grants dmr-0443144 and dmr-0454504 , and also in part by the division of materials sciences and engineering , office of basic energy sciences , u.s .department of energy , under contract de - ac05 - 00or22725 with ornl , managed by ut - battelle .h. ohno , science * 281 * , 951 ( 1998 ) ; t. jungwirth , j. sinova , j. maek , j. kuera , and a.h .macdonald , rev .phys . * 78 * , 809 ( 2006 ) ; a.h .macdonald , p. schiffer , n. samarth , nature materials * 4 * , 195 ( 2005 ) .potashnik , r.f .wang , s.h .chun , p. schiffer , n. samarth , m.j .seong , a. mascarenhas , e. johnston - halperin , r.c .myers , a.c .gossard , and d.d .awschalom , appl .. lett . * 82 * , 2302 ( 2003 ) .one - band models with the exact diagonalization of the fermions were studied in g. alvarez , m. mayr , and e. dagotto , phys .lett . * 89 * , 277202 ( 2002 ) and m. mayr , g. alvarez , and e. dagotto , phys .b * 65 * , 241202(r ) ( 2002 ) . in m. kennett , m. berciu , and r. n. bhatt , phys . rev .b * 66 * , 045207 ( 2002 ) , the ib band model was numerically analyzed .combined mean - field and percolative ideas were used in s. das sarma , e.h .hwang , and a. kaminski , phys .b * 67 * , 155201 ( 2003 ) . mc results for the kinetic - exchange model are in j. schliemann , jrgen knig , and a.h .macdonald , phys .b * 64 * , 165201 ( 2001 ) ; and lsda estimations of @xmath166 are in j. xu , m. van schilfgaarde , and g.d .samolyuk , phys .* 94 * , 097201 ( 2005 ) .r.s . fishman and m. jarrell , j. appl. phys . * 93 * , 7148 ( 2003 ) ; r.s .fishman , j. moreno , t. maier , and m. jarrell , phys .b * 71 * , 180405(r ) ( 2003 ) ; r.s . fishman , f. popescu , g. alvarez , t. maier , j. moreno , and m. jarrell , new j. phys .* 8 * , 116 ( 2006 ) .the average @xmath173=@xmath174 is over the orientations @xmath175 of the local moment on site @xmath79 and is a matrix in @xmath165@xmath154@xmath165 spin space , @xmath176 is the probability for the local moment to point in the @xmath175 direction . while @xmath176=@xmath177 above @xmath161,@xcite in the fm phase near @xmath27 , @xmath178 ( see m. auslender and e. kogan , phys . rev .b * 65 * , 012408 ( 2001 ) ; _ ibid . _* 67 * , 132410 , ( 2003 ) , m. auslender and e. kogan , europhys59 * , 277 ( 2002 ) ) , where @xmath179=@xmath180 and @xmath72=@xmath181@xmath137@xmath182 $ ] is the local - moment order parameter . while @xmath72=@xmath31 deeply in the fm state , at the transition @xmath72=@xmath79@xmath183 is the unity matrix .the behavior displayed in fig .[ fig2](a ) was observed up to @xmath190 , i.e. , a percentage of impurity doping well beyond the experimental posibilities . as @xmath0 approaches@xmath191 the crossover is difficult to monitor because in this limit the system would develop antiferromagnetic long range order and the valence band is splitted in two sub - bands .mott criterion states that upon doping a semiconductor a mit occurs when the wave functions of the holes localized around the impurities start to overlap .this occurs when @xmath192@xmath70@xmath93 .@xmath193 is the @xmath194 density per unit volume and @xmath195 is the bohr radius of the hydrogenic impurity .thus , we can obtain the relative concentration @xmath131 of @xmath21 for which the wave functions would start to overlap in ( ga , mn)as as @xmath108=@xmath196 , where the number of @xmath197 ions per unit volume @xmath198=@xmath199 . hence , mott s criterion becomes @xmath200=@xmath93 .hwang , y. ishida , m. kobayashi , h. hirata , k. takubo , t. mizokawa , a. fujimori , j. okamoto , k. mamiya , y. saito , y. muramatsu , h. ott , a. tanaka , t. kondo , and h. munekata , phys .b * 72 * , 085216 ( 2005 ) .

<S> the crossover between an impurity band ( ib ) and a valence band ( vb ) regime as a function of the magnetic impurity concentration in models for diluted magnetic semiconductors ( dms ) is studied systematically by taking into consideration the coulomb attraction between the carriers and the magnetic impurities . </S><S> the density of states and the ferromagnetic transition temperature of a spin - fermion model applied to dms are evaluated using dynamical mean - field theory ( dmft ) and monte carlo ( mc ) calculations . </S><S> it is shown that the addition of a square - well - like attractive potential can generate an ib at small enough mn doping @xmath0 for values of the @xmath1-@xmath2 exchange @xmath3 that are not strong enough to generate one by themselves . </S><S> we observe that the ib merges with the vb when @xmath4 where @xmath5 is a function of @xmath3 and the coulomb attraction strength @xmath6 . </S><S> using mc calculations , we demonstrate that the range of the coulomb attraction plays an important role . </S><S> while the on - site attraction , that has been used in previous numerical simulations , effectively renormalizes @xmath3 for all values of @xmath0 , an unphysical result , a nearest - neighbor range attraction renormalizes @xmath3 only at very low dopings , i.e. , until the bound holes wave functions start to overlap . </S><S> thus , our results indicate that the coulomb attraction can be neglected to study mn doped gasb , gaas , and gap in the relevant doping regimes , but it should be included in the case of mn doped gan that is expected to be in the ib regime . </S>

1412.0829

over the coming decades , the uk energy market faces significant challenges as it strives to meet its climate change targets .low carbon and renewable generation will need to play a more dominant role in our future electricity supply market .however , renewable energy is intermittent and its availability is driven by uncontrollable elements , such as wind speed and solar intensity .other options , such as nuclear power and carbon capture and storage , are generally considered to be less flexible than traditional thermal plants .the problems are clear .the unreliability of renewable supply means that there will always be a need for a quick - reacting back - up , in order to ensure that demand is met . on the other hand , at off - peak demand times, supply may be curtailed ( often at a high cost ) if the system is not flexible enough to respond . on the demand side , despite improved energy efficiency measures , we can expect to see an overall increase in consumption as the electrification of transport and heating become more commonplace @xcite . if all environmental targets are met on time , then national grid s gone green scenario predicts a 12.6@xmath3 increase in great britain s peak demand by 2035 , from 2013 levels @xcite .in addition , future demand profiles are likely to look very different from today .electric heat pumps , for example , will create a marked seasonal peaking of demand during the colder , winter months . electric vehicle charging and new smart technologies , on the other hand , could distort our daily demand patterns. electricity storage is one potential option for improving the flexibility and reliability of our electricity system .it could offer services such as peak - shaving , frequency response , reactive power regulation and the provision of reserve . in the uk, large - scale electricity storage currently comes only in the form of pumped hydro power plants , which are implemented largely to meet early evening winter peak demands between 4 pm and 8 pm @xcite .their main sources of revenue are through providing the balancing services , short - term operating reserve ( stor ) and fast reserve , which are funded by national grid .the stores then replenish their supply during the night , when electricity prices are at their lowest @xcite .much research has been dedicated to analyzing storage viability .some papers focus on the benefits which a store could bring to the electricity system if operated in a socially optimal way ( see , for example , @xcite ) . typically , the approach of these papers is to solve the unit dispatch problem i.e. to select a suitable configuration of generators ( including storage ) to run at each time , in order to ensure that demand is met ( or that demand is met with a certain high probability ) .an advantage of this method is that it can incorporate the interactions between all assets of the energy system into a single optimization problem , allowing a comparison between the actions of storage against its competing options .a drawback , however , is that the system optimum does not necessarily coincide with each individual firm s optimal strategy . in many cases , it is not even clear that each firm would make a profit under these solutions. therefore , the whole - system approach is generally better suited to questions where the store is not privately owned , but instead owned and operated by a central controller , such as the system operator .other papers focus on the profits available to a store which faces stochastic or probabilistic prices .some of these papers restrict the behaviour of the store to a pre - defined set of operating strategies ( e.g. @xcite ) .another approach is to implement dynamic programming techniques ( e.g. @xcite ) .a drawback of this latter approach , however , is that it tends to be computationally heavy and requires information about prices over the entire time horizon over which we wish to optimize .additionally , such methods do not give much scope for mathematical insight into the dynamics of the solutions .in this paper , we present a method to determine the maximum profit available to an electricity store which is operating in the wholesale electricity market .we assume that the store can predict an electricity price function @xmath4\to\re$ ] ( or alternatively , a cost function ) where @xmath0 $ ] is the period of time over which we wish to optimize .however , the method only ever uses price information over a smaller time interval , thus reducing the required prediction horizon and the amount of computation needed .it is worth commenting here that the prediction of future prices in this way is not an unreasonable assumption , since most electricity today is traded through `` over - the - counter '' bilateral contracts , which are agreed ahead of operating time .our method is derived from standard calculus of variations techniques and is intended as an extension to @xcite .our work differs from @xcite in several respects .firstly , we present our model in a continuous time setting , rather than the discrete setting employed in @xcite .even if prices are declared at discrete time intervals , in accordance with our current market system , our approach allows for the input of piecewise constant prices but with continuous variation in the operation of the store .secondly , @xcite allows only for convex cost functions , whereas here we allow for much more general operating costs and , in particular , we remove the convexity assumption .this is realistic : in general , there is little reason to believe that the cost of running a motor , for example , is a convex function of the output power .moreover , there may be cost jumps involved in switching on a new motor if additional power is required .although these more general conditions do not guarantee the existence of solutions , we prove that optimal operating strategies of the form given in proposition [ prop : ps ] exist if and only if the associated algorithm does not terminate early .finally , we allow the inclusion of additional operating constraints . if these constraints can be expressed as bounds on the instantaneous power output ( e.g. in the case of enforced periods of closure for maintenance ) , then they can easily be incorporated into the main result ( proposition [ prop : ps ] ) and the optimal strategy can be determined using the algorithm of section [ sec : alg ] .this method reduces the optimization problem to a set of smaller problems which are localized in time . for other constraints , however , this localization in time is not guaranteed . in section [ sec : switching ] , we demonstrate this issue and propose a method for handling an additional condition if the store must wait for a fixed amount of time ( called the minimum switching time ) whenever it wishes to change its mode of operation between the charging and the discharging mode .the structure of the paper is as follows : over the remainder of this section , we introduce our storage model and its associated costs . in section 2 , we present the main result , which characterizes the optimal strategies via a reference price function . in section 3, we present an algorithm which determines both the optimal strategy and the reference price for the basic case where the store is only constrained by its power ratings and its capacity constraints .we prove that an optimal strategy exists in this case only if the algorithm does not terminate early .section [ sec : constraints ] contains a discussion of a simple storage model and in section [ sec : switching ] we investigate the inclusion of a minimum switching time constraint . in section 5 ,we present and discuss some illustrative results .finally , the conclusions of the paper are summarized in section 6 .we use a simple technology - agnostic model for storage , which incorporates a number of key features , both physical and economic . to allow for multiple applications, we define all quantities in a general setting . towards the end of the paper , in section [ sec : constraints ], we present a concrete ( and technically relevant ) example .let @xmath0 $ ] be the interval of time over which we want to optimize the actions of the store , for some @xmath5 .at each time @xmath1,$ ] we denote by @xmath6 the set of admissible power outputs associated with the store at that time .the operator of the store may then choose an _ operating strategy _@xmath7\to\re$ ] which allocates , at each time @xmath8 the amount of power @xmath9 that is to enter the store ( using the convention that if @xmath10 then the store discharges @xmath11 units of power at time @xmath12 ) . allowing the power constraint set @xmath13 to vary with time in this way means that we may incorporate constraints such as planned periods of closure or transmission congestion into the model .a planned closure over a time period @xmath14,$ ] for example , corresponds to the power constraint set @xmath15 for all @xmath16 the amount of stored energy , or the _ level of the store _ , which results from the strategy @xmath17 at time @xmath12 is denoted @xmath18(t)$ ] and evolves according to the differential equation @xmath19(t)=-\alpha\ell[q](t)+q(t).\end{aligned}\ ] ] here , @xmath20 is the leakage rate , which is introduced to reflect the self - discharge which occurs due to imperfect sealing or porous casing .given an initial condition @xmath18(0)=\ell_0,$ ] the level of the store at each time @xmath1 $ ] is thus given by @xmath21(t)=e^{-\alpha t}\ell_0+e^{-\alpha t}\int_0^te^{\alpha s}q(s)ds.\end{aligned}\ ] ] we will only consider strategies @xmath7\to\re$ ] which are piecewise continuous functions of time .this is a physically reasonable assumption , since any strategy which has an accumulation point of discontinuities or a non - jump discontinuity would be hard to achieve by any piece of equipment .throughout the remainder of this paper , we denote by @xmath22 the space of piecewise continuous functions @xmath23 for any subset @xmath24 the _ capacity constraints _ of the store are characterized by two functions @xmath25\to\re,$ ] so that any strategy @xmath7\to u$ ] is constrained by the inequalities @xmath26(t)\leq e^+(t ) \ \ \ \ \ \ \ \ \ \ \forall t\in[0,t].\end{aligned}\ ] ] we assume that the initial and terminal levels of the store are pre - specified , so that @xmath27 note that , if not specified , then the terminal condition is in fact implicit since any optimal strategy @xmath28 should satisfy that @xmath29(t)=e^-(t).$ ] we also assume , without loss of generality , that @xmath30 for all @xmath1 $ ] ( otherwise @xmath31 could be adjusted so that this is true , without changing the set of admissible strategies ) . the set of possible levels of the store , at each time , is represented by the _admissible energy domain _ @xmath32\times\re,$ ] which is the set of all pairs @xmath33\times[e^-(t),e^+(t)].$ ] often , one may choose to replace the capacity constraints @xmath34 and @xmath35 with constants , so that @xmath36 is the physical size of the store and @xmath37 is the minimum technically feasible level of the store ( which may be strictly positive , for example in the case of compressed air energy storage , where a cushion of air needs to be maintained in the store at all times in order to provide sufficient pressure to instigate discharging ) .we have chosen here to represent @xmath31 in this more general form , to allow the store to participate in multiple markets .for example , if a store decides to participate in a balancing mechanism reserve contract , it may be required by national grid to make supply available over certain agreed periods .thus , at these contracted times the store will have less capacity available for use in the wholesale market .in addition to the capacity and power constraints outlined above , the operation of a store may be restricted by further factors , such as warm - up times , ramping constraints and minimum switching times between modes of operation .we denote the collection of additional constraints associated with the store by @xmath38 the only condition on these constraints is that they must be entirely expressible as functions of the power output function @xmath39 a maximum ramp - up rate @xmath40 on the charge side , for example , may be written as inequalities @xmath41 \ \textnormal{such that } \q(t)>0.\ ] ] we do not allow for constraints of this kind in section [ sec : nocon ] .however , the impact of a non - empty set @xmath42 is discussed in section [ sec : constraints ] . the total cost incurred by the store , given an operating strategy @xmath43is denoted @xmath44\in\re.$ ] here , @xmath45 is a functional which maps @xmath46 into @xmath47 this can incorporate factors such running costs , warming - up costs and costs of storing , as well as the cost of purchasing power and the payments ( counted as negative cost ) received for providing power .the aim of this paper is to identify those operating strategies @xmath17 which minimize the total cost @xmath44,$ ] whilst adhering to all of the physical constraints outlined above .if the store is profitable to run , then the minimal cost should , of course , be negative .[ defn : storage ] we say that @xmath7\to\re$ ] is an admissible strategy if @xmath17 is piecewise continuous and satisfies the following properties : 1 . for each @xmath1,$] we have @xmath48 and @xmath18(t)\in[e^-(t),e^+(t)].$ ] 2 .if @xmath49 then the strategy @xmath17 satisfies each of the constraints in @xmath38 we denote by @xmath50 the set of all admissible strategies and say that @xmath51 is an _ optimal strategy _ if @xmath52\leq c[q ] \ \ \ \ \ \ \ \ \ \ \ \ \ \forall q\in x. \end{aligned}\ ] ]the following proposition provides a characterization of optimal strategies in terms of a `` reference price '' function @xmath53 we introduce the _ indicator functional _ @xmath54 which is defined by the property that @xmath55 $ ] takes the value 0 if and only if @xmath17 satisfies each of the constraints in @xmath38 [ prop : ps ] suppose there exists a constant @xmath56 a piecewise differentiable function @xmath57\to\re$ ] and a strategy @xmath51 with the following properties : a. the strategy @xmath28 is a minimizer of @xmath58-\int_0^te^{\alpha t}\mu^*(t ) q(t)dt+\rho^*s[q]\end{aligned}\ ] ] over all piecewise continuous functions @xmath7\to\re$ ] such that @xmath48 for all @xmath1.$ ] b. if @xmath59 is differentiable at @xmath1,$ ] then the following complementary slackness conditions are satisfied : @xmath60(t)<e^+(t),\label{eq : mucond1}\\ { \frac{d{\mu^*}}{dt}}(t)\geq 0 \ & \ \ \ \textnormal{if } \ \ell[q]^*(t)=e^+(t),\label{eq : mucond2}\\ { \frac{d{\mu^*}}{dt}}(t)\leq 0 \ & \ \ \ \textnormal{if } \ \ell[q^*](t)=e^-(t).\label{eq : mucond3}\end{aligned}\ ] ] c. if @xmath59 is not differentiable at @xmath1,$ ] then the following `` jump '' complementary slackness conditions are satisfied : @xmath61(t)<e^+(t),\label{eq : mucond1'}\\ \mu^*(t^+)-\mu^*(t^-)\geq 0 \ & \ \ \ \textnormal{if } \ \ell[q^*](t)=e^+(t),\label{eq : mucond2'}\\ \mu^*(t^+)-\mu^*(t^-)\leq 0 \ & \ \ \ \textnormal{if } \ \ell[q^*](t)=e^-(t)\label{eq : mucond3'},\end{aligned}\ ] ] where @xmath62 and @xmath63 are the left and right limits respectively of @xmath59 at @xmath64 then @xmath28 is an optimal strategy .let @xmath65 and let @xmath66 be a partition such that @xmath67=[0,t]$ ] and @xmath68 are continuous and @xmath59 is differentiable over each @xmath69 notice that ( [ elldefn ] ) can be equivalently written as @xmath70(t)=e^{\alpha t}q(t ) \ \ \ \ \ \ \ \ \ \ \ \forall t\in[0,t].\ ] ] hence , ( [ eq : propmax ] ) implies @xmath71&-c[q]\leq\int_0^te^{\alpha t}\mu^*(t)\big(q^*(t)-q(t)\big)dt+\rho^*\big(\underbrace{s[q^*]-s[q]}_{=0}\big)\\ & \stackrel{(\ref{elldefn})}{=}\sum_{i=1}^n\int_{a_i}^{a_{i+1}}\mu^*(t)\left\{\frac{d}{dt}\big(e^{\alpha t}\ell[q^*](t)-e^{\alpha t}\ell[q](t)\big)\right\}dt\\ & = \sum_{i=1}^n\big[e^{\alpha a_{i+1}}\mu^*(a_{i+1}^-)\left(\ell[q^*](a_{i+1})-\ell[q](a_{i+1})\right)-e^{\alpha a_i}\mu^*(a_i^+)\left(\ell[q^*](a_i)-\ell[q](a_i)\right)\big]\\ & \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ -\sum_{i=1}^n\int_{a_i}^{a_{i+1}}e^{\alpha t}{\frac{d{\mu^*}}{dt}}(t)\left(\ell[q^*](t)-\ell[q](t)\right)dt\\ & = \sum_{i=2}^ne^{\alpha a_i}\left(\mu^*(a_i^-)-\mu^*(a_i^+)\right)\left(\ell[q^*](a_i)-\ell[q](a_i)\right)\\ & \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ -\sum_{i=1}^n\int_{a_i}^{b_i}e^{\alpha ( t)}{\frac{d{\mu^*}}{dt}}(t)\left(\ell[q^*](t)-\ell[q](t)\right)dt\end{aligned}\ ] ] where we have used the fact that , since @xmath17 and @xmath28 are admissible , then @xmath72=s[q]=0.$ ] the second equality follows from integration by parts and final equality results from a rearrangement of the first sum on the previous line , together with the condition that @xmath18(0)=\ell[q^*](0)=e^-(0)$ ] and @xmath18(t)=\ell[q^*](t)=e^-(t)$ ] if @xmath73 applying the complementary slackness conditions ( [ eq : mucond1])-([eq : mucond3 ] ) to the final equality above , we obtain @xmath74-c[q]\leq 0.$ ] the reference price @xmath59 provides a reference cost per unit of power .roughly , if the cost per unit of power for operating the store is lower than the reference price , then this indicates the store should charge at that time ( and similarly for discharging ) .it is worth mentioning here that exactly the same result follows if we apply the karush - kuhn - tucker ( kkt ) conditions to the capacity constraints of the original minimization problem ( [ minprob ] ) . by following such an approach ,one instead looks for minimizers of the relaxed functional @xmath75-\int_0^t\big(\lambda_1(t)\left(\ell[q](t)-e^-(t)\right)+\lambda_2(t)\left(e^+(t)-\ell[q](t)\right)\big)dt+\rho^*s[q]\end{aligned}\ ] ] over all piecewise - continuous @xmath7\to\re$ ] such that @xmath48 for all @xmath1,$ ] where @xmath76\to\re$ ] are piecewise - continuous functions which satisfy the complementary slackness conditions @xmath77(t)-e^-(t)\right)=\lambda_2(t)\left(e^+(t)-\ell[q](t)\right)=0 \ \ \ \ \ \ \ \forall t\in[0,t].\ ] ] the relation between the two approaches is that the reference price function @xmath59 in proposition [ prop : ps ] can be expressed as an integral of the kkt functions : @xmath78 and , in particular , @xmath79 the following lemma shows that , in certain cases , we should expect the determination of minimizers @xmath28 to reduce to a sequence of smaller minimization problems which are localized in time .[ lemma : constraints][localization of the minimization problem ] let @xmath80\to\re$ ] be the admissible domain , @xmath81 be the set of admissible power outputs at each time @xmath1 $ ] and let @xmath45 be a cost functional of the form @xmath82=\int_0^tl(t , q(t ) ) \ dt \end{aligned}\ ] ] for some running cost function @xmath83 where @xmath84 is the set of time power pairs @xmath85\times\re$ ] such that @xmath86 for each @xmath1,$ ] let @xmath87 and @xmath88 assume that there exists an optimal strategy which satisfies the conditions of proposition [ prop : ps ] and suppose that the optimal level of the store at some time @xmath89 $ ] is known to be @xmath90 with @xmath91 for any @xmath92,$ ] let @xmath93 solve the minimization problem @xmath94 over all piecewise continuous @xmath95\to\re$ ] such that @xmath48 for all @xmath1 $ ] and such that the end conditions @xmath18(\sigma)=\ell_i(\sigma)$ ] and @xmath18(\sigma_0)=\ell$ ] hold .if there exists a time @xmath96 such that @xmath97(t_0)=\ell_j(t_0),$ ] with @xmath98 then there exists an optimal strategy @xmath51 which coincides exactly with @xmath93 over the interval @xmath99 assume first that @xmath100 so that @xmath97(t_0)=\ell_2(t_0)$ ] and @xmath97(\sigma)=\ell_1(\sigma).$ ] let @xmath51 be an optimal strategy , let @xmath93 satisfy the hypotheses above and suppose that the claim of the lemma is not true. then @xmath29(\sigma)>\ell_1(\sigma)$ ] ( since otherwise we could replace @xmath28 with @xmath93 over @xmath101 $ ] to either reduce the total cost or keep the cost the same ) .thus there must exist at least one interval @xmath102 with @xmath103 such that @xmath29(a)-\ell[q_{\sigma}](a)=\ell[q^*](b)-\ell[q_{\sigma}](b)=0 $ ] and @xmath29(t)\neq\ell[q_{\sigma}](t)$ ] for all @xmath104 however , this implies that both @xmath28 and @xmath93 must be minimizers of @xmath105 over all admissible @xmath106\to u$ ] such that @xmath18(a)=\ell[q^*](a)$ ] and @xmath18(b)=\ell[q^*](b).$ ] in particular , we may adapt @xmath28 if necessary so that @xmath107 for all @xmath108 repeating this argument for all such intervals @xmath109 completes the proof that the optimal strategy can be chosen to coincide with @xmath93 over @xmath99 a similar argument holds for the case where @xmath110 the significance of the above lemma is that @xmath93 depends only on knowledge of a previous time @xmath111 when the store is either full or empty , and on the action of the cost functional over the time interval @xmath112.$ ] in particular , starting at time 0 , one sets @xmath113 and finds a time @xmath114 $ ] such that the minimizer @xmath93 of ( [ eq : lemmamin ] ) hits the upper capacity constraint at some time in @xmath115 the optimal strategy @xmath51 can thus be chosen to coincide with @xmath93 over the interval @xmath116.$ ] we then set @xmath117 and @xmath118 and repeat the process . continuing in this way , we eventually construct a global optimizer @xmath119we assume in this section that the only constraints faced by the store are its capacity and power constraints ( i.e. @xmath120 ) . for simplicity of notation , we assume from now on that there is a fixed set of power constraints @xmath121 such that @xmath122 for all @xmath1.$ ] all of the results extend naturally to the more general case where @xmath123 is allowed to vary with time .we present an algorithm to determine @xmath124 and consequently the optimal strategy @xmath125 for the class of functionals @xmath45 which take the form ( [ c ] ) for some @xmath126\times u\to\re$ ] with @xmath127 for all @xmath1.$ ] we assume that , for each @xmath1,$ ] the map @xmath128 is piecewise differentiable and that the associated partial derivative @xmath129 has a continuous inverse at almost every @xmath130 this class of functionals contains costs which depend only on the time @xmath12 and the power output of the store at that time .the piecewise differentiability assumption allows for jumps in the cost .these jumps may relate to the cost of switching on an additional motor in order to provide a higher power output , for example .proposition [ prop : ps ] states that , if we can find the appropriate reference price function @xmath124 then the optimal strategy @xmath51 solves @xmath131\leq\l[q,\mu^ * ] \ \ \ \ \ \ \forall \ q:[0,t]\to u,\end{aligned}\ ] ] where for any functions @xmath132\to\re,$ ] we define the relaxed functional @xmath133=\int_0^t\big(l(t , q(t))-{e^{\alpha t}}\mu(t)q(t)\big ) \dt.\end{aligned}\ ] ] the regularity assumptions on @xmath134 @xmath17 and @xmath135 imply that ( [ argmin ] ) reduces to a minimization problem which is pointwise in time . specifically , for almost all @xmath1,$ ] we search for @xmath136 which solves @xmath137 our algorithm is a generalization of that provided in @xcite .an important result is proposition [ prop : algpot2 ] , which states that if the algorithm does not terminate early , then it does indeed provide an optimal strategy @xmath119 in particular , in the special case that the cost functional @xmath45 is strictly convex , then there exists a unique optimal strategy , and this coincides exactly with @xmath119 by proposition [ prop : ps ] , we aim to find a piecewise differentiable @xmath57\to\re$ ] such that the optimal strategy @xmath51 solves ( [ argmin ] ) .moreover , @xmath59 should be constant over intervals of time where the level of the store is away from the capacity constraints .this motivates the following construction : given @xmath1 $ ] and @xmath138 we want to define a quantity @xmath139 as a solution of @xmath140 if we know that @xmath141 over some connected interval @xmath14,$ ] then we hope that @xmath142 for all @xmath16 care needs to be taken , however , because there may be multiple minimizers associated with @xmath143 with this in mind , we denote by @xmath144 the set of @xmath145 which admit multiple minimizers in ( [ algmin1 ] ) .note that , if @xmath146 is assumed to be strictly convex in its second argument , then minimizers of ( [ eq : propmax ] ) are of course unique , implying that each @xmath147 [ lemma : xinc ] for each @xmath148\times\re,$ ] let @xmath149 be any solution of ( [ algmin1 ] ). then , at each @xmath1,$ ] the mapping @xmath150 is monotone increasing and piecewise continuous .moreover , the set @xmath151 is finite and discontinuities in @xmath152 occur only at points in @xmath153 let @xmath1 $ ] and suppose that the map @xmath150 is not monotone increasing .then , there exists @xmath154 such that @xmath155 but then , @xmath156 the first inequality follows from the definition of @xmath157 and @xmath158 as solutions to ( [ algmin1 ] ) , and the second inequality follows from the supposition .however , the above contradicts the definition of @xmath159 and we conclude that the map @xmath160 is monotone increasing at all @xmath161.$ ] the piecewise continuity of the map @xmath150 follows immediately from the regularity assumptions on @xmath162 precisely , if for almost all @xmath138 the minimizers @xmath149 lie away from discontinuities of @xmath134 then they must satisfy that @xmath163 the continuous invertibility assumption on the partial derivative of @xmath146 therefore implies the piecewise continuity of the map @xmath164 if , on the other hand , there is a non - degenerate subset @xmath165 such that if @xmath166 then @xmath149 lies at a discontinuity of @xmath134 then the above monotonicity property and the piecewise continuity of @xmath146 imply that the map @xmath150 must be piecewise constant over @xmath167 in either case , the map @xmath150 is piecewise continuous . finally , we prove the finiteness of the set @xmath168 by supposing that the opposite is true . to this end , let @xmath169 be an infinite set such that , for each @xmath170 there are two distinct local solutions @xmath171 to ( [ algmin1 ] ) , where @xmath172 denotes the boundary of the set @xmath173 let @xmath174 be such that there exists a sequence @xmath175 in @xmath176 with @xmath177 and set @xmath178 and @xmath179 assume without loss of generality that @xmath180 is satisfied at @xmath181 where @xmath182 denotes the partial derivative of @xmath146 with respect to the second argument , and that @xmath183 and @xmath184 lie away from any discontinuity of @xmath162 setting @xmath185 and @xmath186 we obtain @xmath187 and similarly , @xmath188 hence , if @xmath189 and @xmath190 are both global minimizers solving ( [ algmin1 ] ) but @xmath191 then there exists @xmath192 such that @xmath193 for all @xmath194 in particular , @xmath195 and @xmath196 can not both be minimizers for @xmath194 this contradicts the assumption that there exists such a set @xmath197 the definitions which follow will be employed in the algorithm , and are written under the assumption that there exists a pair @xmath198 which satisfies the complementary slackness conditions of proposition [ prop : ps ] .condition ( [ eq : mucond1 ] ) motivates the following definition . for any choice of @xmath199 and @xmath200,$ ]we say a piecewise continuous function @xmath201\rightarrow u$ ] is @xmath202admissible if @xmath203 over @xmath204 $ ] and if , for each @xmath1,$ ] @xmath205 is a solution of @xmath206 in particular , the mapping @xmath207 coincides with @xmath208 over @xmath209,$ ] for some choice of map @xmath210 which solves ( [ algmin1 ] ) at each @xmath1.$ ] note that , in general , a @xmath202admissible strategy @xmath211 is not admissible : for a general choice of @xmath212 one of the capacity constraints is likely to be broken at some time in @xmath213.$ ] the idea of the algorithm is to piece together @xmath202admissible strategies in order to construct the optimal strategy @xmath214 the reference price @xmath59 attains the appropriate value of @xmath145 over intervals of time when @xmath59 is constant in the special case that one can find a @xmath215admissible strategy which is also admissible ,then of course we are done , since we may set @xmath216 such a strategy satisfies condition ( [ eq : mucond1 ] ) at every @xmath1.$ ] the key feature of the algorithm is thus to determine : ( i ) the intervals of time over which @xmath59 is constant and ( ii ) the value that @xmath59 attains over these intervals .intuitively , we expect the store to completely fill whenever costs are sufficiently low .the store will then wait for costs to rise before fully discharging .correspondingly , the reference price should increase over this waiting period , from @xmath217 say to @xmath218 .if we did not allow for this increase in reference price , then the associated strategy would cause the store to begin discharging too soon and to empty too much , thus breaking the lower capacity constraint in order to take advantage of the low costs .this behaviour is an indication that @xmath217 is the correct choice of reference price over the charging period .this motivates the following characterization of strategies .[ fnset ] let @xmath200,$ ] @xmath199 and @xmath219.$ ] we characterize each @xmath202admissible @xmath201\to u$ ] according to sets @xmath220 and @xmath221 as follows : a. we write @xmath222 if there exists @xmath223 and @xmath224 such that @xmath225(t)&=e^-(t ) \qquad \textnormal{and } \qquad m+\ell[x](t+\varepsilon)<e^-(t+\varepsilon)\end{aligned}\ ] ] for all @xmath226 with @xmath227(t')\leq e^+(t ' ) \qquad \forall t'\in(\tau , t ) .\end{aligned}\ ] ] + in other words , if the level of the store at time @xmath228 is @xmath229 and if the store adopts the strategy @xmath211 over @xmath213,$ ] then the first violation of a capacity constraint occurs at the lower bound of @xmath230 we write @xmath231.$ ] b. similarly , we write @xmath232 if there exists @xmath223 and @xmath224 such that condition ( [ defncond ] ) holds and @xmath225(t)&=e^+(t ) \qquad \textnormal{and } \qquad m+\ell[x](t+\varepsilon)>e^+(t+\varepsilon)\end{aligned}\ ] ] for all @xmath233 in other words , if the level of the store at time @xmath228 is @xmath229 and if the store adopts the strategy @xmath211 over @xmath213,$ ] then the first violation of a capacity constraint occurs at the upper bound of @xmath230 we write @xmath231.$ ] c. we say @xmath234 if there exists a @xmath202admissible @xmath235 similarly , @xmath236 if there exists a @xmath202admissible @xmath237 lemma [ lemma : xinc ] implies that @xmath238 and @xmath239 are each connected subintervals of @xmath176 which satisfy @xmath240\times[0,m ] .\end{aligned}\ ] ] in particular , the interiors of the two sets are disjoint : @xmath241\times[0,m ] .\end{aligned}\ ] ] we are now in a position to outline the algorithm . to thisend , fix @xmath242 and suppose that @xmath59 and @xmath28 are known over @xmath204.$ ] we assume without loss of generality that @xmath243(\tau)\in\{e^-(\tau),e^+(\tau)\}$ ] ( otherwise , move @xmath228 backwards until this is satisfied ) .the algorithm will identify an @xmath192 and two increasing sequences of times @xmath244,@xmath245 such that @xmath246 for each @xmath247 and such that @xmath248 whenever @xmath249 the reference price @xmath59 is only allowed to jump in value at times of the form @xmath250 or @xmath251 we may assume without loss of generality that @xmath252 for some @xmath253 ( otherwise , again , move @xmath228 backwards until this is true ) . recall from definition [ fnset ] that @xmath254 $ ] is the time at which a strategy @xmath211 first breaks a capacity constraint , given that @xmath255 + + * step 1 : * if there exists @xmath199 and a @xmath256admissible @xmath211 such that @xmath257 then set @xmath258 set @xmath259 for all @xmath260 $ ] and define the restriction of @xmath28 to @xmath261 $ ] to be @xmath262 * the algorithm is complete . *if there is no @xmath145 satisfying these conditions , proceed to step 2 . +* step 2 : * set @xmath263 there are three cases . for each case, we define the pair @xmath198 restricted to an interval @xmath264\subset(\tau_k , t].$ ] * if @xmath265 choose a @xmath256admissible @xmath211 such that @xmath266(\sigma_k)=e^+(\sigma_k)$ ] for some @xmath267),$ ] and select the latest such @xmath268 set @xmath259 for all @xmath269 $ ] and define the restriction of @xmath28 to @xmath264 $ ] to coincide with @xmath262 if @xmath270 proceed to step 3 ; if @xmath271 proceed to step 4 . *if @xmath272 choose a @xmath256admissible @xmath211 such that @xmath266(\sigma_k)=e^-(\sigma_k)$ ] for some @xmath267),$ ] and select the latest such @xmath268 set @xmath259 for all @xmath269 $ ] and define the restriction of @xmath28 to @xmath264 $ ] to coincide with @xmath262 if @xmath270 proceed to step 3 ; if @xmath271 proceed to step 4 .* if @xmath273 choose a @xmath256admissible @xmath211 such that either @xmath266(\sigma_k)=e^+(\sigma_k)$ ] or @xmath266(\sigma_k)=e^-(\sigma_k)$ ] for some @xmath267),$ ] and select the latest such @xmath268 set @xmath259 for all @xmath269 $ ] and define the restriction of @xmath28 to @xmath264 $ ] to coincide with @xmath262 if @xmath270 proceed to step 3 ; if @xmath271 proceed to step 4 . * step 3 : * let @xmath274 be the first time in @xmath275 $ ] such that , on relabeling @xmath274 as @xmath276 and @xmath266(\tau_{k+1})$ ] as @xmath229 and on determining the new associated @xmath277 from steps 1 and 2 , we have @xmath278 relabel @xmath274 as @xmath276 and return to step 1 .if there is no such @xmath279 then we have found all intervals @xmath280 over which @xmath281 proceed to step 4 .+ * step 4 : * over each interval @xmath282,$ ] corresponding to the times found in the steps above , and for each @xmath283,$ ] define @xmath284(\sigma_i)=e^+(\sigma_i)\\ \frac{d}{dt } e^-(t)+\alpha e^-(t ) & & \textnormal{if } \\ell[q](\sigma_i)=e^-(\sigma_i ) .\end{array } \right .\end{aligned}\ ] ] * the algorithm is complete * if , for each each @xmath285 there exists a piecewise differentiable @xmath286 such that the following conditions hold : * @xmath287 is a minimizer of @xmath288 over all @xmath289 * if @xmath29(\sigma_i)=e^+(\sigma_i),$ ] then @xmath290 is non - decreasing over @xmath291 and satisfies that @xmath292 and @xmath293 * if @xmath29(\sigma_i)=e^-(\sigma_i),$ ] then @xmath59 is non - increasing over @xmath282 $ ] and satisfies that @xmath294 and @xmath295 if these conditions do not hold , the algorithm ends here and yields no solution @xmath119 + * end of algorithm * + the above algorithm implicitly reduces the original optimization problem to a series of new optimization problems which are localized in time . at each time@xmath296 one only needs to look ahead to a time @xmath297 to know how to operate the store over @xmath298 where @xmath297 is defined as @xmath254 $ ] for the appropriate choice of @xmath211 and @xmath299 from step 2 of the algorithm ( or as the the minimum value of @xmath254 $ ] whenever there there is a choice of @xmath211 ) .hence , whilst we originally assumed knowledge of the entire price function @xmath4\to[0,\infty),$ ] in practice we may not need so much information .the following result states that , provided the algorithm does not terminate early , then it does indeed give an optimal strategy @xmath119 proposition [ prop : algpot2 ] then states the converse : if there is an optimal strategy , which satisfies the conditions of proposition [ prop : ps ] , then the algorithm will find it ( in other words , the algorithm will not terminate early ) .[ prop : algopt ] if the above algorithm yields a strategy @xmath300\to u,$ ] then @xmath28 is an optimal strategy .let @xmath198 be the pair which is determined through the algorithm and let @xmath244 and @xmath301 be the associated sequences of time , so that @xmath59 is constant over each interval @xmath302 assume that the conditions of proposition [ prop : ps ] are satisfied up until time @xmath303 at each @xmath285 we have @xmath304(\tau_i))\leq\mu^*(\tau_i^+)\leq\inf\lambda'(\tau_i,\ell[q^*](\tau_i)).\end{aligned}\ ] ] if the conditions of step 1 are satisfied at @xmath296 then @xmath28 is clearly admissible and satisfies the properties of the proposition .hence , @xmath28 is optimal and the algorithm is complete .assume therefore that the conditions of step 1 do not hold at @xmath276 so that @xmath305 it is clear that the strategy defined by the algorithm is admissible , and it remains to check that conditions ( [ eq : mucond1])-([eq : mucond3 ] ) are satisfied by @xmath53 to this end , note that , the algorithm ensures that the conditions are satisfied over the intervals @xmath306 and @xmath307,$ ] and it remains to check that the conditions hold at @xmath268 suppose first that case a ) of step 2 is satisfied at time @xmath296 so that @xmath29(\sigma_k)=e^+(\sigma_k).$ ] then , by construction , we have @xmath308 thus , if @xmath309 then , together with ( [ lambdamu ] ) , this implies @xmath310 and condition ( [ eq : mucond2 ] ) is satisfied at @xmath268 if , on the other hand , @xmath311 then the construction of @xmath59 through the algorithm immediately implies that @xmath312 similarly , if case b ) of step 2 holds at @xmath296 then @xmath29(\sigma_k)=e^-(\sigma_k)$ ] and @xmath313 thus , if @xmath309 then , together with ( [ lambdamu ] ) , this implies @xmath314 and condition ( [ eq : mucond3 ] ) is satisfied at @xmath268 if , on the other hand , @xmath311 then the construction of @xmath59 through the algorithm immediately implies that @xmath315 finally , if case c ) holds at @xmath316 then the above two cases together imply that the conditions of proposition [ prop : ps ] are satisfied .[ prop : algpot2 ] if the algorithm terminates early , then there is no pair @xmath198 which satisfies the conditions of proposition [ prop : ps ] .let @xmath198 be a pair generated by the algorithm .( if the algorithm terminates early , at a time @xmath317 say , then @xmath59 and @xmath28 are both considered as functions over the domain @xmath318.$ ] ) let @xmath319 and @xmath320 be the sequences of time generated by the algorithm , so that @xmath59 is constant over each @xmath321 we write @xmath322(\tau_k)$ ] and assume that each interval @xmath306 is maximal in the sense that if @xmath323 is a strictly larger interval , then @xmath59 is not constant over this interval .suppose also that there is a pair @xmath324 which satisfies the conditions of proposition [ prop : ps ] .analogously , let @xmath325 and @xmath326 be increasing sequences which define the end - times of all maximal intervals @xmath327 over which @xmath135 is constant .we write @xmath328(t_k)$ ] and @xmath329 we make the following additional assumptions on the pair @xmath324 : suppose there is a pair of times @xmath330 $ ] such that @xmath331 and @xmath18(t)\in\{e^-(t),e^+(t)\}$ ] for each @xmath332 if there exists a pair @xmath333 such that ( i ) the conditions of proposition [ prop : ps ] are satisfied , ( ii ) @xmath334(\tau)=\ell[q](\tau)$ ] and @xmath334(\tau')=\ell[q](\tau'),$ ] and ( iii ) @xmath335 for all @xmath336 then we assume that @xmath324 agrees with @xmath333 over the larger interval @xmath337 in other words , we always take @xmath135 to be constant whenever there is the choice . throughout this proof , we will use the notation @xmath338 if the set of @xmath339admissible functions coincides exactly with the set of @xmath340admissible functions over the interval @xmath341 this assumption is only required to handle the degenerate cases where @xmath17 is identically zero over @xmath341 in these cases , there may be a choice of values to assign to @xmath342 each of which would result in the same @xmath343admissible @xmath17 over @xmath341 we introduce the ordering @xmath344 if @xmath345 and @xmath346 the proof consists of proving that the following statements are true at each @xmath253 : 1 .@xmath347 and @xmath17 is @xmath343admissible over the interval @xmath341 2 .there exists @xmath348 such that @xmath349 3 . if @xmath348 and @xmath350 then @xmath351 and @xmath18(t_k)=\ell[q](\sigma_k).$ ] the first statement is required to prove statement 2 , and implies that the algorithm determines the correct value of @xmath135 over each intervaltogether , statements 2 and 3 then complete the proof , since they imply that the algorithm uncovers all intervals @xmath327 over which @xmath135 is constant .moreover , statement 3 implies that the strategy uncovered by the algorithm agrees with @xmath28 at the end point of each of these intervals .hence , in between each @xmath277 and @xmath279 the strategies @xmath17 and @xmath28 necessarily lie on the capacity constraints and so must coincide .in particular , the existence of the pair @xmath324 implies that the algorithm does not terminate early but instead generates a pair @xmath198 which are defined over @xmath0.$ ] _ proof of statement 1 : _ let @xmath352 it is clear from the definition of @xmath343admissibility and from the requirement that @xmath17 must be a minimizer of ( [ minptwise ] ) , that @xmath17 must be @xmath343admissible over the interval @xmath327 .we need to show that @xmath353 recall that the maps @xmath150 of lemma [ lemma : xinc ] are piecewise continuous and monotone increasing at each time @xmath1.$ ] these maps are constructed in such a way that one can always find a suitable @xmath354 to coincide with @xmath17 over @xmath341 thus , the monotonicity of @xmath210 with respect to its second argument implies the relations : @xmath355 suppose now that @xmath356 if @xmath357(s_k)=e^+(s_k),$ ] then this immediately implies that @xmath358 since , otherwise , the continuity and monotonicity of @xmath210 would imply the existence of @xmath359 such that @xmath360 and a @xmath361admissible @xmath362 such that @xmath363(t)\geq e^+(t)$ ] for all @xmath364,$ ] with strict inequality at @xmath365 this would imply that @xmath366 which contradicts ( [ beta1 ] ) .if , on the other hand , @xmath357(s_k)=e^-(s_k),$ ] then proposition [ prop : ps ] requires that @xmath367 if @xmath135 is constant over an interval @xmath368,$ ] then the monotonicity property of lemma [ lemma : xinc ] implies that the lower capacity constraint will be broken by @xmath43 and no admissible solution will be found .if @xmath135 is not constant over any such interval , then proposition [ prop : ps ] requires that @xmath369 for all @xmath370 by the monotonicity of @xmath371 this is only possible if @xmath372 at some @xmath373 or if @xmath374 since otherwise the fact that @xmath375 implies that @xmath17 will break the lower capacity constraint .however , this contradicts condition ( [ eq : mucond3 ] ) of the proposition .hence , we must have @xmath358 or @xmath376 if @xmath377 then similar arguments as above arrive at analogous contradictions to proposition [ prop : ps ] andwe conclude that we must therefore have @xmath378 this completes the proof of statement 1 ._ proof of statement 2 : _ suppose first that @xmath379 but that there is no @xmath348 such that @xmath380 then , statement 1 implies that @xmath381 and that @xmath17 must coincide with some @xmath382admissible @xmath211 over @xmath383 however , since @xmath228 does not coincide with any @xmath384 the conditions of steps 1 and 2 of the algorithm do not hold at time @xmath385 if @xmath386 then proposition [ prop : ps ] implies that @xmath387 and @xmath18(s_k)=e^-(s_k)$ ] ( since the conditions of step 2 do not hold ) . thus , proposition [ prop : ps ] implies that either ( i ) @xmath388 and @xmath389 or ( ii ) @xmath390 and @xmath391 for all @xmath392.$ ] in case ( i ) , the monotonicity property of lemma [ lemma : xinc ] implies that @xmath17 breaks the lower capacity constraint . in case ( ii ) , since @xmath211 breaks the lower capacity constraint , we must have that @xmath393 for all @xmath394 in some interval @xmath395.$ ] however , since we require that @xmath396 for all @xmath397 this means that we must have @xmath398 for all @xmath394 in some subset @xmath399 this contradicts the conditions of proposition [ prop : ps ] , and we conclude that there exists @xmath348 such that @xmath380 the same result follows from similar arguments if @xmath400 this completes the proof of statement 2 ._ proof of statement 3 : _ let @xmath401 first notice that the proof of statement 2 yielded a contradiction when we assumed that @xmath17 coincided with a @xmath343admissible @xmath402 such that @xmath387 and @xmath18(s_k)=e^-(s_k).$ ] since statement 1 implies that @xmath17 must coincide with some @xmath343admissible @xmath211 over @xmath403 it follows that , if @xmath386 then either @xmath404 or @xmath18(s_k)=e^+(s_k).$ ] similarly , if @xmath405 then either @xmath404 or @xmath18(s_k)=e^-(s_k).$ ] the assumption that the intervals @xmath327 are maximal thus implies that @xmath406 coincides with @xmath407 and that @xmath18(s_k)=\ell[q^*](\sigma_j),$ ] provided that @xmath407 is independent of any choices made at steps 1 or 2 of the algorithm .this independence is not immediately clear since the general assumptions on the cost function @xmath146 mean that there may be more than one choice of @xmath408admissible strategy . to prove the independence of @xmath407 from the choice of strategy made in the algorithm at steps 1 or 2 ,suppose that there are two @xmath408admissible strategies @xmath409 and @xmath410 which satisfy the conditions of step 1 or 2 of the algorithm .let @xmath411 denote the time @xmath407 if strategy @xmath412 is chosen at time @xmath228 and assume without loss of generality that @xmath413 the monotonicity property of lemma [ lemma : xinc ] makes it immediately clear that , if we choose @xmath409 then the algorithm will select @xmath414 for all @xmath415 hence , @xmath59 is constant over @xmath416 in particular , if there are several choices of strategy to choose at time @xmath316 and if @xmath417 is the latest time @xmath407 associated with any of the choices , then the assumption that each interval @xmath418 is maximal implies that @xmath419 regardless of the strategy chosen .this completes the proof of statement 3 .we introduce here a simple but practically relevant model of a store , which operates purely within a single wholesale market .the set of admissible powers @xmath123 is taken to be of the form @xmath420,$ ] with @xmath421 the store is assumed to have a fixed maximum energy capacity @xmath422 so that the admissible domain satisfies @xmath423 and @xmath424 for all @xmath425 we further assume that we require the store to be empty at both the start and end times , so that @xmath426 the cost functional @xmath45 is defined by @xmath427=\int_0^tw(q(t))p(t)q(t),}\end{aligned}\ ] ] where @xmath428 is the price for one unit of power at time @xmath64 here , @xmath429 is a piecewise continuous function which acts as a weighting to the cost of buying and selling electricity , and should satisfy that @xmath430if @xmath431 and @xmath432 if @xmath433 more intuitively , @xmath434 can be thought of as the result of inefficiencies during the charging and discharging processes of the store : in order to fill the store at a rate @xmath435 at time @xmath8 the storage operator would actually need to purchase a greater amount of electricity , since power is lost during the charging process .similarly , whenever an amount @xmath435 is discharged from the store , a lesser amount is actually available to sell on the market . for simplicity , we assume that @xmath434 takes on two values @xmath436 and @xmath437,$ ] so that @xmath438 in the special case where there are no additional operating constraints ( i.e. @xmath439 , this cost functional is convex and thus an optimal strategy exists. proposition [ prop : algpot2 ] states moreover that an optimal strategy can be determined via the algorithm of section [ sec : alg ] .some simple results are immediately available , as stated in the following lemma .define the parameters of a store as above and let the cost functional @xmath45 be given by ( [ eq : cexample ] ) .assume that @xmath440 for all @xmath1 $ ] and let @xmath441 be the required levels of stored energy at times @xmath442 and @xmath443 respectively .the following properties hold : 1 . if @xmath444 and @xmath445 then there is no non - zero strategy @xmath65 which yields a positive profit .the minimum cost , over all admissible strategies , is a decreasing function of @xmath446 and @xmath447 it is an increasing function of @xmath448 and @xmath449 if @xmath65 is not the zero function , and if @xmath450 is constant for all @xmath1,$ ] then @xmath75= p\int_0^tw(q(t))q(t ) \dt\geq p\int_0^tq(t ) \ dt = p\left(\ell_t-\ell_0\right)+\alpha\int_0^t\ell[q](t ) \ dt \geq 0.\end{aligned}\ ] ] this proves statement 1 . to prove statement 2 ,let @xmath28 be the optimal strategy associated with the parameters @xmath451 and write @xmath452.$ ] it is easy to see that @xmath28 is also admissible with respect to the parameters @xmath453 for any @xmath454 and any @xmath455 with @xmath456 @xmath457 and @xmath458 thus , if @xmath459 is defined by @xmath460 whenever @xmath461 and @xmath462 whenever @xmath463 then @xmath464 this proves the required growth conditions of the minimum cost with respect to @xmath465 and @xmath466 to complete the proof , let @xmath467 and denote by @xmath468(t)$ ] the level of stored energy at time @xmath12 associated with the strategy @xmath17 and the leakage rate @xmath469 set @xmath470 whenever @xmath471 or @xmath472(t)<m.$ ] for all other @xmath1,$ ] set @xmath473 then @xmath362 is admissible with respect to the new parameter vector @xmath474 hence , @xmath475=c[q']+\int_0^tw(q(t))p(t)\left(q^*(t)-q'(t)\right ) \dt\geq c[q']\geq v(\theta')\end{aligned}\ ] ] and the minimum cost function is increasing with respect to the leakage rate @xmath476 throughout this section , we employ the notation @xmath477 and @xmath478 to mean @xmath479 and @xmath480 respectively for any @xmath481 we investigate the impact of introducing a non - empty set of additional constraints @xmath42 to the simple storage model introduced above . in thissetting , lemma [ lemma : constraints ] no longer holds in general .it is therefore not clear that there will exist a method with similar time - localization properties as the algorithm of section [ sec : alg ] .we demonstrate this problem explicitly in this section .in particular , we impose a minimum switching time constraint @xmath482 on the simple storage model introduced above .this is the minimum amount of time for which the store must be idle when switching between the charging and discharging modes of operation .hence the additional constraint set @xmath42 consists of the requirement that any power output @xmath17 must satisfy that @xmath483 \\textnormal{and } \s\in(t-\tau , t+\tau)\cap[0,t].\end{aligned}\ ] ] the minimization problem of proposition [ prop : ps ] is difficult to solve in its current form , but we may rewrite the problem by adapting the lagrangian @xmath484 of ( [ relaxed ] ) to incorporate the minimum switching time constraint as follows : @xmath485:=\underbrace{\int_0^t\big(w(q(t))p(t)-e^{\alpha t}\mu(t)\big)q(t ) \dt}_{{=\mathcal l[q,\mu]}}-\int_0^t\int_t^{(t+\tau)^-}\lambda(t , s)q(s)q(t ) \ ds \ dt\ ] ] for all piecewise continuous @xmath132\to u$ ] and @xmath486 ^ 2\to[0,\infty].$ ] in order to apply proposition [ prop : ps ] , we need to reduce @xmath487 into a form which is compatible with ( [ eq : propmax ] ) . to do this , we carefully choose @xmath145 to depend on @xmath17 and @xmath135 as follows : for a given @xmath342 let @xmath488\to u$ ] solve @xmath489 since the efficiency function @xmath434 takes the simple form ( [ w ] ) , it is clear that @xmath490 takes a value in @xmath491 whenever @xmath492 and @xmath493 at @xmath1.$ ] in these cases @xmath490 is uniquely defined .if one of these relations is an equality instead , however , then @xmath490 may take any value in @xmath173 hence , there may be multiple solutions associated with @xmath494 let now @xmath145 minimize @xmath495 over all @xmath486 ^ 2\to[0,\infty]$ ] such that ( i ) @xmath496 whenever @xmath497 and ( ii ) the inequality @xmath498 holds for all @xmath1.$ ] we write @xmath499 the significance of the above construction is that an optimal strategy can now be found by searching for minimizers of @xmath500 specifically , if a strategy @xmath28 satisfies that @xmath501\leq\lambda[q,\mu^*,\lambda^*]\end{aligned}\ ] ] for all piecewise continuous @xmath7\to u,$ ] and if ( i ) @xmath502 satisfies the conditions ( [ eq : mucond1])([eq : mucond3 ] ) , ( ii ) @xmath503 is of the form @xmath504 and ( iii ) @xmath28 is admissible , then @xmath28 is an optimal strategy . to see this ,choose @xmath505 large enough that the relation @xmath506 holds .it is clear from the assumption that @xmath28 is admissible that @xmath72=0 $ ] and it follows from ( [ lambdamin ] ) that @xmath28 minimizes @xmath507+\rho^*s[q]$ ] over all admissible @xmath73 assumption ( [ rho ] ) moreover implies that @xmath28 minimizes @xmath507+\rho^*s[q]$ ] over all piecewise continuous @xmath7\to u$ ] because , for any non - admissible @xmath43 we have @xmath508+\rho^*\underbrace{s[q]}_{=1}&\geq\mathcal l[q,\mu^*]-\int_0^t\int_t^{(t+\tau)^-}\lambda^*(t , s)q(s)q(t ) \ ds \dt=\mathcal l[q^*,\mu^*]+\rho^ * \underbrace{s[q^*]}_{=0}. \end{aligned}\ ] ] thus , proposition [ prop : ps ] implies that @xmath28 is optimal. condition ( iii ) of the above argument requires that the minimizer @xmath28 of ( [ lambdamin ] ) is admissible , in order for @xmath28 to be an optimal strategy . as in the algorithm of section[ sec : alg ] , a suitable choice of @xmath59 will ensure that the capacity constraints are satisfied , whilst the power rating constraints are immediately absorbed into the minimization problem ( [ lambdamin ] ) .the following lemma ensures that , as long as a minimizer @xmath28 exists , then we may select @xmath28 so that the switching time constraints are also satisfied . [ lemma : lambda ] let @xmath509\to\re$ ] be piecewise differentiable and let @xmath145 be of the form @xmath510 for some minimizer @xmath511 of ( [ eq : qmu ] ) .let @xmath512 be the set of piecewise continuous @xmath300\to u$ ] such that @xmath513\leq \lambda[q,\mu,\lambda]$ ] for all piecewise continuous @xmath7\to u.$ ] if @xmath514 then there exists @xmath515 which satisfies the switching time constraints ( [ gamma ] ) .suppose the claim is not true .then there exists @xmath516 $ ] such that @xmath517 and @xmath518 thus @xmath519 note that @xmath520(t):=\frac{\delta \lambda}{\delta q(t)}[q,\mu,\lambda]=w(q((t))p(t)-e^{\alpha t}\mu(t)-\int_{(t-\tau)^+}^{(t+\tau)^-}\lambda(t , s)q(s ) \ ds , \end{aligned}\ ] ] where @xmath521 denotes the functional derivative of @xmath487 with respect to @xmath435 .in particular , the term @xmath522 only influences the value of @xmath28 at times @xmath12 and @xmath523 the condition that @xmath145 minimizes ( [ lambdamin ] ) subject to the constraint ( [ lambdamin2 ] ) implies that @xmath524(t)=0 $ ] or @xmath524(s)=0.$ ] in particular , we may always choose a @xmath28 so that @xmath525 or @xmath526 an interpretation of the role of @xmath503 in the minimization problem ( [ lambdamin ] ) is that the function acts to adjust @xmath527 to a new strategy which satisfies the switching time constraints .it does this by finding all pairs of times @xmath528 such that @xmath529 and @xmath530 and adjusting @xmath531 or @xmath532 to 0 .the condition that @xmath533 whenever @xmath534 ensures @xmath527 is only adjusted at times when the minimum switching time constraint ( [ gamma ] ) is broken . choosing @xmath503 to minimize ( [ lambdamin ] ) subject to the constraint ( [ lambdamin2 ] )ensures that @xmath503 makes these adjustments in such a way that , if @xmath535 is the absolute change in instantaneous cost from the original strategy @xmath527 at time @xmath8 then the accumulation of these cost changes @xmath536 is minimal .thus , an equivalent and generally easier method for determining the optimal strategy @xmath125 given @xmath124 is to find a suitable subset of times @xmath14 $ ] and set @xmath525 if @xmath537 and @xmath538 if @xmath1\setminus i.$ ] the subset @xmath539 should be chosen to minimize the associated absolute cost change integral @xmath540 if @xmath59 is given , then the minimization problem ( [ lambdamin ] ) is much easier to solve than ( [ eq : propmax ] ) .the main task in determining an optimal strategy @xmath125 therefore , is to choose the correct reference price function @xmath53 as alluded to at the beginning of section [ sec : constraints ] , this is a more difficult task now that the set of additional constraints @xmath42 is non - empty . to illustrate the complexity of the problem ,consider the following scenario : let the power ratings and capacity constraints of the store be given by @xmath541 and @xmath542 and let the minimum switching time of the store be @xmath543 assume that there is no leakage and the store is perfectly efficient so that @xmath544 and @xmath545 suppose that @xmath546 and that the price of electricity @xmath547 is a function of time given by @xmath548 for all @xmath549 and for each @xmath550 where the values @xmath551 satisfy @xmath552suppose further that @xmath553 it is not difficult to see that the optimal strategy @xmath28 satisfies @xmath554 if @xmath555 $ ] and @xmath556 if @xmath557,$ ] with @xmath525 elsewhere .the corresponding reference price @xmath59 satisfies @xmath558 if @xmath559 and @xmath560 otherwise .however , the correct choice of @xmath561 here relies on knowing the correct choice of @xmath562 unlike in the algorithm of section [ sec : alg ] , the constant sections of the reference price function can not be determined independently . to see this, we simply remark that , if we had guessed at a constant reference price function @xmath563 then we would arrive at the corresponding solution @xmath564 with @xmath565 if @xmath566 $ ] and @xmath567 otherwise .this strategy never reaches the upper or lower capacity constraint after the initial time , and so there is no chance to update our reference price function in accordance with proposition [ prop : ps ] .even though our guess @xmath568 is correct over the time interval @xmath569 it is only possible to know this when considering the value that @xmath59 attains over the time interval @xmath570.$ ] this issue prevents the methods of section [ sec : alg ] from working here .however , as long as the store is required to perform at least one full charge and discharge cycle then the determination of a suitable @xmath59 still reduces to a localized problem thanks to lemma [ lemma : constraints ] .finally , we remark that , if @xmath59 is a reference price function and if @xmath571 is the associated strategy , constructed via the minimization ( [ lambdamin ] ) , then it can easily be shown that at each time @xmath1,$ ] the level of the store @xmath572 $ ] is increasing in @xmath53 whilst the algorithm of section [ sec : alg ] no longer works in this more complicated setting , we can still use this monotonicity in order to reduce the choice of @xmath59 at each time to a suitable range , using similar arguments as in the case with no additional constraints .the following graphs plot the optimal strategies for a store which follows the model set out in section [ sec : constraints ] , with @xmath573 and @xmath574 we take @xmath575 @xmath576 and vary @xmath577 as labeled below .the bottom two figures plot the level of the store @xmath29(t)$ ] at each time @xmath12 , assuming that the optimal strategy @xmath28 is followed .the top figures both plot the price of electricity @xmath4\to[0,\infty)$ ] which the store faces in making its operating decision . here , we have chosen hourly prices from the first 230 hours of n2ex s day - ahead auction in november 2013 .( bottom left ) and @xmath578 ( bottom right).,width=566,height=226 ] an immediate observation is that the store cycles more frequently as its efficiency @xmath577 increases .less energy is lost during operation at higher efficiencies and the store is able to earn revenues over a larger proportion of elapsed time .we highlight here that , although the price of electricity follows a reasonably periodic pattern ( with periods of roughly a day in length ) , it does not follow that the level of the store is the same at the start and end of each day ( see , for example , @xcite for a proof of the potential non - periodicity of solutions in a periodic setting ) .this underlies an important motivation for choosing this method over a dynamic programming approach .using a dynamic programming method , one works backwards in time and evaluates , at each @xmath1,$ ] the value function @xmath579 where the minimization is subject to the constraints @xmath580 however , this requires information about all prices over the time @xmath0.$ ] if @xmath443 is large ( say , @xmath581 years , the average life of a pumped hydro store ) , then the number of calculations becomes infeasible .one option is to split @xmath0 $ ] into a union of smaller disjoint intervals @xmath582,$ ] for some @xmath583 and to find the optimal strategy over each of these smaller intervals .however , this requires setting an end - state for each interval , which may not be known . a reasonable guess in the above examples would be to assume that the store is empty at some off - peak time during each night but , as seen , such an assumption would lead to a sub - optimal solution . using our approach of section [ sec : alg ], the algorithm has the property that it implicitly reduces our problem to a series of new optimization problems , which are localized in time .we have presented a method for determining the operating strategy for a store to maximize its arbitrage profits .our setting allows for leakage , inefficiencies and general operating costs which are functions of the power output .this setting also allows for time - varying constraints on the power output of the store .a significant benefit associated with this method is the implicit localization in time of the solution .moreover , we have shown that there only exists an optimal strategy of the form of proposition [ prop : ps ] if the algorithm does not terminate early . as an extension, we have discussed the inclusion of more general operating constraints and proposed a method for determining optimal strategies in the presence of minimum switching time constraints .we believe that the theory put forward in this paper serves as a good starting point for evaluating the profits available to a store .the assumption that prices can be predicted over suitable periods of time is a good approximation to the situation where a store trades through bilateral contracts , or through an auctioning market .a next step would be to consider the store as a larger player , whose actions have an impact on the price of electricity , and it is believed that the approaches given in this paper can be extended to this new setting .this work is made possible by the images ( integrated market - fit affordable grid - scale energy storage ) research group .the authors would like to thank the other members of the group for their suggestions and insight .particular thanks go to monica giulietti , jihong wang , xing luo , andrew pimm and seamus garvey .the authors would also like to thank stan zachary , james cruise and richard gibbens for their many helpful discussions related to this topic .m. aunedi , n. brandon , d. jackravut , d. predrag , d. pujianto , r. sansom , g. strbac , a. sturt , f. teng , v. yufit : _ strategic assessment of the role and value of energy storage systems in the uk low carbon energy future - report for carbon trust , _ 2012 _ study of compressed air energy storage with grid and photovoltaic energy generation , draft final report - for arizona public service company , _arizona research institute for solar energy ( aztise ) , 2010

<S> we characterize profit - maximizing operating strategies , over some time horizon @xmath0,$ ] for an energy store which is trading in an arbitrage market . </S><S> our theory allows for leakage , operating inefficiencies , operating constraints and general cost functions . in the special case where the operating cost of a store depends only on its instantaneous power output ( or input ) , we present an algorithm to determine the optimal strategies . </S><S> a key feature is that this algorithm is localized in time , in the sense that the action of the store at a time @xmath1 $ ] only requires information about electricity prices over some subinterval of time @xmath2\subset[t , t].$ ] to introduce more complex storage models , we discuss methods for an example which includes minimum switching times between modes of operation . </S>

1511.08252

active galactic nuclei ( agn ) are powered by the accretion of matter onto a supermassive black hole ( smbh ) .radiatively efficient agn , which includes optically and x - ray - selected agn as well as luminous radio - loud agn ( r - agn ) , are powered by the efficient accretion of matter from the accretion disk onto the smbh .these agn may be triggered by mergers ( heckman et al .1986 , barnes & hernquist 1996 , satyapal et al .2014 ) , although the lack of an environment density - agn luminosity relation suggests that this may not be the case @xcite . while some r - agn are found in late - type galaxies @xcite , the majority of r - agn in the current universe are found in passively evolving galaxies @xcite and are powered by the radiatively inefficient accretion of hot gas via either bondi accretion or the advective cooling of gas from the halo or from evolved stars .both modes of agn may affect their host galaxies via feedback : radiatively efficient agn via powerful winds , and radiatively inefficient agn via their extended radio jets .although agn - driven feedback could in principle boost a galaxy s star formation rate ( sfr ) by inducing the collapse of cold gas clouds @xcite , its _ negative _ effect on star formation ( sf ) is more commonly discussed in the literature .this negative feedback is required for simulations and semianalytic models to match the observed properties of massive ellipticals ( e.g. granato et al .2004 ; ciotti et al . 2010 ; cattaneo et al . 2009 ) as well as prevent the overproduction of massive galaxies @xcite .many clusters whose brightest cluster galaxies ( bcgs ) host radio jets are also observed to have x - ray bubbles ( dunn & fabian 2006 ; 2008 ) , which could indicate that heating by agn balances cooling flows in these clusters .this balance may depend on the timescales and instabilities of the cooling gas @xcite .although this negative feedback may be observed in cluster environments , it is uncertain whether it takes place in individual galaxies , partly due to the greater difficulty in detecting the fainter x - ray emission from their lower - density halo gas . in this paper, we examine the effects of r - agn activity on the stellar populations of their host galaxies using a large sample of r - agn in the local universe , identified by matching radio surveys to optical , ultraviolet ( uv ) and mid - infrared ( mid - ir ) surveys .we are most interested in the effects of radio jets on recent star formation , which can be traced in the uv .mid - infrared emission traces star formation , but it may also arise from old stellar populations or nuclear activity . to disentangle the effects of these emission sources we compare the uv and mid - ir emission of a sample of r - agn , drawn from the @xcite catalog , to that of a carefully selected sample of control galaxies . since r - agn bubbles are observed in clusters but may not be present in field galaxies , we treat r - agn in different environments ( field galaxies , cluster members , bcgs ) separately .many studies find that r - agn tend to be found in denser environments ( e.g. best 2004 ; kauffmann et al .2008 ; reviglio & helfand 2006 ; pace & salim 2014 ) , but this general statement hides important details . the incidence rate of r - agn is higher in bcgs than galaxies in the field ,although the incidence rate in cluster members is comparable to that found in the field @xcite .this suggests that the r - agn incidence depends on either the properties ( the mass ) of the dark matter halo rather than the actual density of galaxies , or that bcg r - agn have an enhanced availability of gas compared to r - agn in the field due to the presence of a cooling flow .this flow may cause the agn to always be in the on " state , in contrast to r - agn in the field which are only intermittently triggered . despite the higher incidence rate of r - agn in bcgs, the majority of r - agn are found in the field . we treat r - agn with a low accretion rate , known as low - excitation radio galaxies ( lergs ) , separately from those with a high accretion rate , which are known as high - excitation radio galaxies ( hergs ) .this is because they may not interact with the intergalactic medium in the same way .in addition to their radio jets it is possible that hergs , which constitute only a few percent of the r - agn population @xcite , also affect their host galaxies via agn - driven winds .the lergs and hergs are distinguishable by their ` excitation index , ' which is derived from four optical emission lines .many previous studies have used the fanaroff - riley classification @xcite rather than the lerg / herg separation .this classification uses radio morphology and radio luminosity to separate r - agn into two categories .r - agn that are core - brightened and of lower average luminosity are referred to as fanaroff - riley type 1 ( fr1 ) sources , while r - agn that are edge - brightened and of higher average higher radio luminosity are fanaroff - riley type 2 ( fr2 ) sources .although lergs broadly overlap with fr1 s and hergs with fr2 s , a portion of fr2 sources are lergs @xcite . in our studywe use the herg / lerg classification of @xcite since it more closely matches the accretion rate and because the morphological information required to classify into fr types is generally not available for this sample . as an alternative to excitation classification, we also split all r - agn into groups according to what we call _ specific radio luminosity _ ( srl ) , the radio luminosity divided by the galaxy s stellar mass .we consider the normalized radio luminosity to be more relevant than just the radio luminosity in the context of investigating potential feedback because a jet of a given radio luminosity will probably not have the same effect on galaxies of different masses .we find that the srl is better correlated with excitation classification than just the radio luminosity .radio jets are strongly anisotropic , and for feedback to be effective they must somehow deposit their energy to the surrounding halo gas .this is believed to occur as the expanding cavities and bubbles which the jets produce efficiently transfer the jet energy to the surrounding gas and disrupt the cooling of gas from the surrounding hot halo .this feedback mode could be responsible for maintaining the low star formation rates ( sfrs ) of early - type galaxies across cosmic time , however there is a lack of evidence , not only for a causal connection between the presence of agn and the suppressed sf , but even that the two phenomena are correlated .a given passive , or nearly passive galaxy can be passive for a number of different reasons , including e.g. , the environment .the goal of this study is to search for such a correlation .the strongest evidence for r - agn heating of halo gas comes from x - ray observations of clusters .@xcite studied the x - ray spectra of 14 cooling flow clusters and found that the gas temperatures at the cluster cores are only @xmath01/3 that at large radii , which is far below what is expected if a strong cooling flow were taking place .the perseus cluster shows evidence of weak shocks and ripples in the x - ray , which could be evidence of the expanding radio bubbles heating the ambient gas @xcite .x - ray bubbles may also be seen in individual elliptical galaxies .@xcite studied a sample of 104 ellipticals and found cavities in 24 of them , which could indicate that the r - agn duty cycle is lower in individual galaxies ( @xmath025% ) than it is in cluster environments , where it may approach 100% @xcite .a few studies have directly examined the effects of feedback on the host galaxy s sfrs , with conflicting results .@xcite measured the far infrared ( fir ) sfrs of a sample of @xmath03,000 x - ray and radio - selected agns in the chandra deep field south and found that r - agn have a much higher sfr than purely x - ray selected agns , which appears to be at odds with the scenario of r - agn quenching star formation in their hosts .the difference may arise from the @xcite sample being from an earlier cosmic epoch ( @xmath3 ) . the simple picture of r - agn negatively affecting star formation is further complicated by the results of @xcite , who examined the infrared star formation rates of a sample of @xmath0300 radio sources in the north ecliptic pole ( nep ) field and found a positive correlation between the luminosity of the agn component and star formation , which suggests that greater agn activity leads to increased sf .however when the sample is binned by redshift and agn luminosity , the specific star formation rate ( ssfr ) is found to decrease with increasing radio luminosity .interactions with radio jets may affect star formation in adjacent satellite galaxies .powerful r - agn may induce star formation in their satellies , as seen in the fr2 sources pks 2250 - 41 @xcite and 3c 285 @xcite .interactions with weak r - agn may also induce star formation , as seen in the fr1 sources associated with centaurus a @xcite and minkowski s object @xcite . on the other hand, radio jets may quench star formation in satellites .@xcite found that the @xmath4 colors of satellite galaxies in the projected path of fr2 jets are redder than those outside of the path , but no such trend was found for fr1 sources .this may indicate that interactions with powerful radio jets quench star formation , while weaker jets are not capable of doing so .this is in contrast to the results of @xcite , who found no clear difference in the satellite colors of r - agn and normal galaxies , regardless of the agn accretion mechanism , which suggests that on average r - agn have little influence on star formation in their hosts . in this work, we will look for evidence of r - agn feedback on their hosts by statistically exploring the relationship between r - agn and recent star formation .this will be done by comparing the composite spectral energy distributions ( seds ) of a sample of r - agn to that of a carefully selected control sample .the composite seds are generated from uv , optical , and mid - ir photometry from the galaxy evolution explorer ( galex ) , the sloan digital sky survey ( sdss ) , and the wide - field infrared survey explorer ( wise ) , respectively .since many of the r - agn and control galaxies have weak detections in the uv and mid - ir , we adopt a bayesian approach to produce the composite seds , which removes the biases of only considering the detections above some s / n ratio threshold .this paper is organized as follows .section 2 describes the data sources used as well as how this information was combined .section 3 details the selection of the r - agn and control samples while section 4 describes our method of constructing the composite seds .section 5 presents the results of our study , and in sections 6 and 7 we discuss our findings .the cosmological parameters we have adopted in this work are @xmath5 , @xmath6 , and @xmath7 km s@xmath8 mpc@xmath8 .our analysis is based on a sample of r - agn , drawn from @xcite , that have galex coverage to a uniform depth . like the r - agn sample , the control sample is drawn from areas of sdss that are covered by the radio surveys as well by galex to the appropriate depth . because wise is an all - sky survey , both the r - agn and control galaxies are by definition covered by wise .the selection and properties of these samples is described below .a sample of 18,286 radio - emitting galaxies was created by best & heckman 2012 ( hereafter bh2012 ) by combining galaxies with spectra from the 7th sdss data release ( dr7 ; abazajian et al .2009 ) with two radio surveys : the national radio astronomy observatory ( nrao ) very large array ( vla ) sky survey ( nvss ; condon et al . 1998 ) and the faint images of the radio sky at twenty centimeters ( first ; becker , white & helfand 1995 ) survey .since our study focuses in the uv and mid - ir emission of the r - agn , we consider the portion of the bh2012 sample that is covered by the galaxy evolution explorer ( galex ; martin et al .2005 ) to a uniform depth .we therefore selected bh2012 galaxies with galex exposure times in the range of @xmath9 seconds corresponding to what is known as the medium - depth imaging survey ( mis ) .although the range of exposure times is somewhat large , the majority of observations have exposure times of @xmath01,500 s. this resulted in an initial pool of 5,400 galaxies .our study of these r - agn used data from three large - area , multiwavelength surveys .ultraviolet photometry were drawn from the dr6/dr7 galex data release , optical photometry were drawn from the tenth data release ( dr10 ; ahn et al .2014 ) of the sloan digital sky survey ( sdss ) , and infrared photometry were drawn from the allwise catalog ( wright et al .2010 , mainzer et al .the galex survey is an all - sky survey in two uv bands : the far - ultraviolet band ( fuv ) is centered at a wavelength of 1528 @xmath10 while the near - ultraviolet ( nuv ) band is centered at a wavelength of 2271 .the wise mission surveyed the entire sky in four bands , designated w1 through w4 .the w1 band is centered at 3.3 @xmath1 m , the w2 band at 4.6 @xmath1 m , the w3 band at 12@xmath1 m , and the w4 band at 22 @xmath1 m .we required the control galaxies to have the same depth in galex as the r - agn , and found that there are 361,469 sdss galaxies that are covered in galex with the required depth .fluxes , magnitudes , and the dr10 reported redshifts were extracted from the three surveys for these galaxies .a nearest - neighbor approach was used to combine information from the surveys .we used a 5 " radius to match the galex and wise sources to the sdss sources .approximately 0.02% of the r - agn had multiple sdss matches and 9% had multiple wise matches in the 5 radius , and in these situations we kept the source with the smallest separation . choosing the closest match in wise limits the fraction of potential spurious matches to @xmath11 .the sdss and galex photometry was used to derive the physical properties ( stellar mass , mass - weighted ages , sfrs ) of the galaxies via uv - optical sed fitting to stellar population synthesis models , which were then used in the selection of the control sample as well as the analysis .the fitting was performed following the methodologies of @xcite and @xcite .we perform the fitting by comparing the observed fluxes to a library of model fluxes , which are derived from stellar population synthesis models @xcite .star formation histories are combinations of exponentially declining trends ( with inverse decline rates sampled from 0 to 1 gyr@xmath8 ) , with uniformly distributed formation times , and superimposed stochastic bursts ( with burst probability of 50% over 2 gyr ) .metallicities take values uniformly from 0.02 to 2 z@xmath12 .details are given in @xcite .stellar masses and sfrs were derived using chabrier imf .the @xcite dust extinction model is applied to model seds .libraries of model fluxes are produced in redshift steps of z = 0.01 .the initial pool of bh2012 galaxies contains r - agn as well as galaxies in which the radio emission is not from a nuclear source , but is associated with sf ( i.e. , sn remnants , condon 1992 ) .a combination of three methods were used by @xcite to separate the r - agn from the normal galaxies in their sample .two of these methods ( 4000 break strength vs. ratio of radio luminosity per stellar mass ; h@xmath13 line luminosity vs. radio luminosity ) are based on the assumption that r - agn have enhanced radio emission compared to normal galaxies with the same sf , where @xmath14 and h@xmath13 serve as proxies for the latter .the third method , which is the emission - line diagnostic or ` bpt ' diagram ( baldwin et al .1981 , kauffmann et al . 2003 ) , relies on differences in excitation due to the relative strengths of the ionizing radiation of agn and o stars .while these separation methods usually agree , they may occasionally give contradictory classifications .for a full description of how a final classification was determined in such cases , see appendix a of bh2012 .the h@xmath13 line luminosity is an indicator of star formation , but is affected by dust attenuation and agn emission .furthermore the observed h@xmath13 luminosities in the sdss database are not corrected for the fixed fiber sizes of the sdss spectrograph .no dust or aperture correction was applied by bh2012 , so it is possible that the h@xmath13 emission of some of the r - agn may not reflect their true star formation rate .sed - based sfrs similar to those used here were already subjected to extensive comparison with h@xmath13 based sfrs in salim et al .it was found that for galaxies with low levels of sf the sed based sfrs are superior to those based on h@xmath13 .the specific star formation rates ( ssfrs ) , which are the sfrs divided by the stellar mass , from our sed fitting provide another measure of star formation activity , and can be used to filter out any galaxies for whom the h@xmath13 luminosity is a poor indicator of the true star formation rate .figure [ ssfr ] shows the distribution of bh2012 sources in the log ( ssfr ) vs. log(@xmath15 ) plane .the sample is clearly composed of two types of galaxies : those with high ssfrs but low radio luminosity per stellar mass ( upper left ) and those with a low ssfr but a high radio luminosity per stellar mass ( bottom ) . the dashed line at @xmath16 shows our division between star - forming galaxies and r - agn .galaxies classified as star - forming and r - agn by bh2012 are shown in blue and black , respectively .most of the bh2012 star forming galaxies lie above the line while most of the galaxies that bh2012 classify as r - agn lie below the line . vs. @xmath17 for the 5,400 galaxies @xcite galaxies that are covered in the galex mis .the dashed line shows our division between star - forming galaxies and r - agn , while the galaxies classified by @xcite as star - forming and r - agn are shown as blue and black points , respectively .those r - agn ( black points ) that lie above the line are plotted on top of the star - forming galaxies ( blue points ) while below the line the star - forming galaxies are plotted on top of the r - agn . in most casesour classification agrees with that of @xcite . , width=336 ] rcccc[h ! ] field&108&2,536&446&3,090 + member&4&201&33&238 + bcg & 3&421&83&507 + ' '' '' total&115&3,158&562&3,835 + [ mytable ] for our sample we retain those bh2012 r - agn ( black points ) that lie below the line .the sdss data pipeline assigns objects a category of either galaxy , star , or qso based on a comparison of the object s spectrum to a range of template spectra .none of the r - agn below our dividing line were classified by the sdss pipeline as star .the uv emission of quasars will include a contribution from the agn , which would affect our ability to use uv as a indicator of sfr .lergs lack an accretion disk and should therefore have weak or no uv emission while the uv from the accretion disks of hergs is blocked by the dusty obscuring structure , so any uv observed in r - agn should trace star formation . we therefore excluded 65 r - agn that are classified by the sdss dr10 pipeline as quasars , leaving a final sample of 3,835 r - agn .the magnitude range of this sample is 11.72 @xmath18 18.62 , with a redshift range of 0.01 @xmath19 0.3 .the lower redshift limit is that of the original b&h sample , and the upper redshift limit that we have adopted matches that of the sdss main galaxy sample @xcite , thus providing a large pool of candidate control galaxies .we assign an environment classification for each galaxy in this sample , following the procedure of @xcite .three environment classifications are used : field galaxy , cluster member , and bcg .the dr10 coordinates and redshifts of the r - agn sample were used to calculate the projected distance to the nearest bcg in the @xcite catalog of bcgs .those r - agn with a projected distance to the nearest bcg of less than 1.5 kpc were classified as bcgs , while those with a projected distance between 1.5 kpc and 1.5 mpc and a difference in spectroscopic redshift less than @xmath20 were classified as cluster members .the remaining galaxies were classified as field galaxies .the number of galaxies in each environment are shown in table [ tbl1 ] .because we want to address how the accretion mechanism relates to feedback effects , we have used the lerg and herg classifications provided by bh2012 .these classifications are based on the excitation index " defined by and in most cases ( 84% of the sample ) can be used to classify the galaxy as either a herg or lerg .many of the remainder could not be classified because their emission lines are undetected , which could because they are lergs , but could also be due to their faintness .table [ tbl1 ] shows that 115 or 3% of the sample are hergs , while 3,158 or 81% of the sample are lergs .table 1 shows that hergs are almost never found in high - density environments .most lergs are also found in the field , although they are more likely than hergs to be found in cluster members or bcgs .we draw the control sample from a pool of 331,145 candidate galaxies , which , like our r - agn sample , are covered by galex observations of intermediate depth , and fall within the first footprint .the latter is essential to be able to establish that the galaxy is not an r - agn . for each r -agn we identify one control galaxy , by matching by environment ( i.e. , being in the field , in a cluster , or a bcg ) and by redshift , stellar mass , population age , and the axis ratio .the purpose of these matching criteria is to ensure that the control sample has the same quality of photometry as the r - agn sample ( ensured by redshift matching ) and that the galaxies have similar evolutionary histories ( matching by stellar mass and mass - weighted population age ) .matching by population age ensures that the galaxies have roughly similar sf histories .the population age , being mass - weighted and therefore dominated by past sf , is not sensitive to the current sfr , which is what we wish to compare between r - agn and the control sample .because most radio galaxies are ellipticals ( which tend to be rounder ) , we also apply matching by axis ratio in order to reduce the occurrence of lenticulars ( s0s ) , which tend to be flatter , among the control galaxies .stellar masses and population ages of s0s may overlap with those of the ellipticals , but s0s can nevertheless have higher current levels of sf @xcite . for each r - agn ,candidate control galaxies were first selected to have the same environment type ( field candidate for field r - agn , etc . ) , then the best match from among these candidates was chosen by minimizing the metric @xmath21 here the quantities @xmath22 , @xmath23 [ @xmath24 , @xmath25 [ gyr ] , and @xmath26 represent the difference between the r - agn and control candidates in redshift , stellar mass , population age , and axis ratio , respectively .the denominators represent the 95 percentile range of these properties for the r - agn sample and are used so that all criteria have the same relative weight .this selection process sometimes results in one control galaxy being the best match to more than one r - agn .this occurs for 9% of the controls and has no effect on our analysis .figure [ control ] shows the age and stellar masses of the control sample , plotted together with those of the r - agn .the contours enclose 90% , 50% , and 25% of the samples .divisions in age and mass indicated by the dashed lines are used in the analysis in section 4 .this plot illustrates the success of our matching process , as galaxies in the control sample have the same distribution as the r - agn in age as well as stellar mass .the analysis is based on the comparison of aggregate seds of r - agn vs. the seds of control galaxies .each comparison set is a part of the full sample binned or classified by some property .for example , we may wish to compare the seds of bcg r - agn to bcg control galaxies .an sed consists of luminosities , or absolute magnitudes , at 11 wavelengths ( 2 uv , 5 optical and 4 mid - ir ) . galaxies in our sample lie at a range of redshifts ( @xmath27 ) so the observed fluxes sample seds at a range of rest - frame wavelengths . to obtain comparable sedswe construct them as if all galaxies were at @xmath28 . to achieve thiswe assume that small sections of seds can be described with a slope @xmath29 .therefore , to construct the sed we need to obtain 11 m@xmath30 and @xmath31 .the standard approach for obtaining each of the 11 sed points would be to correct the observed fluxes of galaxies in the comparison set for galactic reddening and then convert the fluxes into absolute magnitudes , to which m@xmath32 and @xmath31 are fitted .however , this procedure would be problematic for galaxies with low s / n measurements , which may sometimes even have negative measured fluxes .this is the case for many of the galaxies in our sample , usually more distant galaxies in the far uv and at 12 and 22 @xmath1 m . rather than introducing some s / n threshold which will bias m@xmath32 towards stronger detections , we take a bayesian approach to determine the sed , i.e. , we ask what m@xmath32 and @xmath31 can best explain the observed fluxes of galaxies in the comparison set .therefore , we assume some m@xmath32 and @xmath31 and predict the observed fluxes based on the distance to the galaxy and the galalactic reddening for that galaxy .we try different values of m@xmath32 and @xmath31 until the @xmath33 @xmath34 is minimized . predicting the observed flux from modelabsolute magnitude requires an additional correction .namely , the galaxies in a comparison set have a range of stellar masses and the seds will to first order differ simply because of the scaling with the mass . to remove this overall variationwe construct the sed assuming all galaxies had @xmath35 , the median mass of the sample .therefore , in order to produce a predicted flux , the model absolute magnitude is adjusted by @xmath36 to test the robustness of our technique , we perform the simulation of the bayesian estimation of the absolute magnitude in w4 ( 22 @xmath1 m ) , the band where the detections are the weakest .we assume that the true absolute magnitude in w4 at @xmath28 is -22.0 ( for runs # 1 and # 2 ) and -21.5 ( for run # 3 ) .these assumed values bracket the actual absolute magnitudes at 22 @xmath1 m .we produce 1000 ( runs # 1 and # 3 ) and 100 ( run # 2 ) such objects and distribute them over the redshift range @xmath37 , peaking at @xmath38 , similar to the actual redshift distribution .the simulated sample size ( 1001000 ) brackets the numbers in the actual analysis .we also assume the slope of the sed to be @xmath39 , a typical true value . based on the assumed absolute magnitude at @xmath28 , the assumed sed slope and the redshift , we calculate for each object its true flux and , based on it and the sensitivity of wise , the expected flux error .we add an additional 2% error to mimic calibration uncertainties .finally , we perturb true fluxes with the value drawn from a gaussian with the standard deviation equal to the flux error .we then use these simulated fluxes to reconstruct @xmath40 and @xmath41 using the bayesian approach described above .for runs with 1000 objects we retrieve the absolute magnitude within 0.01 mag and slope within 2 ( 10% ) . for the run with 100 objectsthe estimated magnitude is 0.1 mag off from the actual value .these results are shown as the rightmost points in figure [ simres ] .this figure also shows what results we would have obtained if we used a standard approach in which only the detections above some signal - to - noise ( s / n ) ratio are used in the calculation .because w4 detections are weak , only a small percent of the objects reach s / n@xmath42 . absolute magnitudes that would be obtained only using these measurements are 1 to 1.5 magnitudes brighter than the actual value , and the slopes are much steeper than what they ought to be . lowering the s / n ratio cut reduces the discrepancy , but the derived absolute magnitudes are still too high ( 0.3 - 0.5 mag ) even when all objects with positive fluxes are used ( s / n@xmath430 ) .the least biased , and , as we show , rather accurate , answer is obtained only when all of the data are used , including those having `` unphysical '' negative fluxes , which comprise 2030% of the simulated samples .if the galaxies in the comparison set were truly homogeneous or drawn from a normal distribution the above procedure would be sufficient to obtain a typical sed : 11 m@xmath32 and their errors .the knowledge of robust errors is critical to be able to establish if the seds of r - agn and the control sample differ significantly or not .however , we have noticed that in the uv and 12 and 22 @xmath1 m there is a tail of galaxies with higher fluxes than that of the bulk of the galaxies in a given comparison set .they drive the minimum @xmath33 to high values and render the absolute magnitude error from the @xmath33 analysis useless ( because it assumes the underlying distribution of fluxes to be normal ) .furthermore , the best - fit ( minimum @xmath33 ) absolute magnitude will implicitly ignore these outliers , whose presence is important for understanding the properties of the comparison set .we address these issues by determining m@xmath32 and its error in a way that is more appropriate given the circumstances .namely , we take the sed slope derived from the @xmath33 minimization , but now look for m@xmath32 such that 50% of the observed fluxes will be higher than the predicted value and 50% lower , i.e. , we obtain m@xmath32 that is the equivalent of a median , rather than the weighted average as in the case of @xmath33 minimization . note that we used the information on the sed slope from the @xmath33 analysis , as it can not be determined by this median " estimate . to determine the error of m@xmath32we perform 100 determinations of the median from bootstrap resampled data points and determine the 84th and 16th percentile ranges of such medians .thus we obtain separate upper and lower error bars .this error takes into account not only the flux errors but also the range due to the intrinsic distribution of absolute magnitudes in the comparison set . for all fluxpoints the errors of magnitudes from the aggregate sed is dominated by the range of values of individual galaxies in the sample , not the photometric error .in this paper , we study the possible effects of r - agn on their host galaxies by comparing the seds ( especially the uv and mid - ir portions ) of the r - agn and control samples .we split the sample into comparison sets by accretion mode , specific radio luminosity , environment , and mass and age .in addition to comparing the r - agn and control galaxies , we also compare the seds of the r - agn , binned according to these properties .we also evaluate the effectiveness of mid - ir colors at selecting r - agn of either accretion type as well as how mid - ir colors are related to the star formation rates and radio luminosity of the r - agn and the host galaxy .we first consider how differences in the accretion mechanism are reflected in the composite seds .for this portion of the analysis we restrict ourselves to those galaxies in the field , which make up the majority of the sample ( table [ mytable ] ) .this is done because the large - scale environment may have an effect on the seds , and we explore these effects separately .panels ( a)-(c ) of figure [ accretion ] compare the seds of the r - agn to those of their controls , while panel ( d ) shows only the r - agn .panel ( a ) shows that lergs have less uv and mid - ir emission than their controls , while only the uv emission of the unclassified r - agn in panel ( b ) is less than that of their controls .in contrast , the uv emission of the hergs in panel ( c ) is similar to that of their controls , while the mid - ir of the hergs is greatly elevated .the comparison of the r - agn with different accretion modes , in panel ( d ) of figure [ accretion ] , shows that the overall shape of the seds for the lerg and unclassified r - agn are more similar to each other than compared to the hergs .both have uv luminosities that correspond to relatively passive galaxies ( @xmath44 ) , 1.5 dex lower than the sfr of galaxies on the main sequence having the same mass . however , the seds of lergs and unclassified r - agn exhibit some differences .most notably , unclassified r - agn are fainter in @xmath45 , @xmath46 and @xmath47 .this is consistent with them being on average older ( 0.2 gyr ) and more massive ( 0.11 dex ) than the lergs .this likely a result of the ratio of unclassified r - agn to lergs increasing from @xmath48 at @xmath28 to @xmath49 at @xmath50 , such that in the higher redshift bins the mean luminosity of the unclassified r - agn is greater than that of the lergs ( malmquist bias ) . moving on to the hergswe see that their optical emission is quite similar to that of the lergs , but their uv and mid - ir emission is strongly elevated ( figure [ accretion ] d ) .this difference is greatest in the uv and longest two wise bands , although it is seen in the other two wise bands as well .hergs are therefore consistent with having higher levels of sf than lergs and/or higher levels of dust heating from an agn .the specific sfr of hergs is @xmath51 , which is 0.6 dex higher than that of lergs .panels ( a)-(c ) show the seds of the r - agn and their associated controls in the three accretion mechanism categories . both the uv and mid - ir emission of the lergs in panel ( a ) is lower ( fuv : 0.09 dex ; nuv : 0.07 dex ; 12 @xmath1 m : 0.06 dex ; 22 @xmath1 m : 0.05 dex ) than that of the control galaxies , although the difference at 22 @xmath1 m is not statistically significant .the r - agn for which bh2012 provide no classification are shown in panel ( b ) , and here the uv is lower for the r - agn , while emission in the wise bands is broadly similar to that of the controls .the herg sample , shown in panel ( c ) , only has @xmath0100 members which results in greater uncertainty in the composite seds for both the r - agn and controls .even so , there are clear differences between the two .the hergs have a very strong excess across all four mid - ir bands . because the uv emission of hergs and their controls is similar ,all of the mid - ir excess can be attributed to dust heating from the agn .these results support the picture of hergs being found in galaxies with sufficient cold gas to fuel both radiatively - efficient accretion , which is manifest as mid - ir emission from heating of the obscuring torus , and star formation , which is manifest as elevated uv emission .the fact that we see similar levels of uv emission in both hergs and their controls is still consistent with this idea because not every galaxy with a large quantity of cold gas will be an r - agn , but when it is , it will be a herg .the agn and star formation activity appear to have little effect on one another , which is evident in the similarity of the uv portions of the composite seds in panel ( c ) .the weak mid - ir signature of the lergs probably results from their lack of an accretion disk and associated obscuring torus , while their weak uv emission reflects the low star formation rates of their hosts , and is consistent with the idea that their hosts have little available cold gas .however , unlike the hergs the fact that the lergs have both suppressed uv and ir emission suggests that there is a link between star formation lerg activity .next we consider an alternative way to classify r - agn . not by excitation classification , but by the actual radio power , i.e. , the specific ( mass - normalized ) radio luminosity .this was used together with the d@xmath52 break strength by @xcite and @xcite to separate star forming galaxies from r - agn , but here we consider its utility in separating high and low - accretion rate r - agn .the radio luminosities are expressed in w hz@xmath8 and are calculated from the radio fluxes in bh2012 .we divide all field r - agn into four categories by srl .panels ( a)-(d ) of figure [ radio ] show the seds of the field r - agn and their controls in four srl bins , while panel ( e ) shows the comparison among the r - agn in the different srl bins .there is a trend of increasing uv emission with increasing srl in the seds of the three bins with the lower srl in panel ( e ) .although the trend is weak , as all three bins are within @xmath00.1 dex , it is significant , as the derived error bands do not overlap .this trend may suggest that there could be a link between the fueling of low levels of sf and fueling weak r - agn . note that having a link , or a common cause of these two processes does not eliminate the possibility that the sf in r - agn is nevertheless suppressed compared to galaxies without an agn .the bin with the highest srl is very different , with a far greater level of both uv and mid - ir emission .the elevated emission in this bin is observed in all wise bands , with the strongest difference being found in the longer two wavelength bands .the sed of the r - agn with the highest srl is similar to that of the hergs , even though not all hergs belong to this srl bin ( @xmath01/3 do ) nor are all r - agn in this srl bin hergs ( @xmath01/3 are ) .panels ( a ) and ( b ) show the composite seds of the r - agn and control galaxies in the two bins with the lower srl . in the log(l@xmath53/m@xmath54)@xmath55 binthe r - agn have less uv and mid - ir emission ( fuv : 0.12 dex , nuv : 0.08 dex ; 12 @xmath1 m : 0.07 dex ) , as do the r - agn in the @xmath56 log(l@xmath53/m@xmath54)@xmath57 bin ( fuv : 0.1 dex ; nuv : 0.06 dex ; 12 @xmath1 m : 0.06 dex ; 22 @xmath1 m : 0.12 dex ) .these deficits in uv and mid - ir suggest that the r - agn in these bins have lower sfrs than their controls .this begins to change for the r - agn in the next srl bin in panel ( c ) , where the mid - ir emission is now very similar to that of their controls , while their uv emission remains lower than that of the control galaxies .a possible explanation may be that in the mid - ir we are beginning to see agn emission , countering the 12 and 22 @xmath1 m deficit seen in the lower two srl bins .the control galaxies in this panel exhibit a statistically significant upturn in uv emission , but its origin is unknown , as it does not appear in the other panels .the trend of increasing mid - ir emission with increasing srl becomes evident in panel ( d ) , with the emission of the r - agn being strongly elevated relative to that of the controls .the larger error bands for this bin are due to its small size , as it only has @xmath0200 r - agn .the results from this section are qualitatively similar to those from the previous section where we split by excitation .we see that the uv and mid - ir deficits are characteristic of weaker r - agn , but the mid - ir deficit turns into a strong excess as the radio power increases .we next explore differences in the seds for the r - agn found in different environments : ( 1 ) bcgs , ( 2 ) cluster member other than bcg and ( 3 ) non - cluster galaxy ( field " ) . to isolate the effects of the environment, we analyze only lergs , which make up the majority of the r - agn ( 82% ) and are found in all three types of environment .panels ( a)-(c ) of figure [ enviro ] show the seds of the lergs and their controls in the three environments , while panel ( d ) shows the comparison of the seds for lergs in the three environments .the optical and uv portions of the seds in panel ( d ) are very similar i.e. , being in a cluster does not lead to a further decrease in the sfr . in the longest two wise bands the emission is decreasing as the density increases .the separation here is much stronger , with the bcgs having the weakest emission in all four wise bands .the slope of the sed at the longest two wise bands may also change with density , largely due to weaker 12 @xmath1 m emission in r - agn bcgs .we also see that the bcg controls in panel ( c ) have weak 12 @xmath1 m emission compared to the controls in the other environments .these differences in the mid - ir are not due to differences in the srl or population age , which are similar for all three categories. a possible explanation could be that bcg lergs have much less dust than field lergs .the w3 filter is also sensitive to pah emission , so another explanation could be that pahs are more easily destroyed in denser environments .panel ( a ) shows the results for the field lergs , which of course is the same as panel ( a ) of fig .[ accretion ] .as discussed above , the field lergs show a slight deficit of mid - ir emission , while the uv is more strongly suppressed relative to that of the controls . since the field lergs is the largest sample their composite seds have the narrowest 1-sigma ranges so the differences between the lergs and controls are the most robust of all the environments .there are only @xmath0200 cluster members , which explains the larger 1-sigma error bands in panel ( b).the r - agn and controls in this panel completely overlap in the uv , which suggests that being in a cluster can have a similar effect on sf as having an agn .however , other factors such as having a more massive dark matter halo or a smaller cold gas reservoir could also affect the sfr .there is 0.06 dex deficit the 12 @xmath1 m emission in the cluster member lergs , while there are no significant differences in the seds of r - agn bcgs and control bcgs , shown in panel ( c ) . like the r - agn bcgs, the control bcgs have exceptionally low mid - ir emission , especially at 12 and 22 @xmath1 m , consistent with the aforementioned possibility that bcgs are less dusty than the early type galaxies in other environments , which may be because the dust destruction timescales are shorter in the cores of galaxy clusters .overall , the broad similarity of the r - agn seds in different environments suggests that the environment does not have an additional effect on sf .however , it is possible that the environment affects the amount of dust in galaxies , which could explain the trend for galaxies in denser environments having less mid - ir emission . in future studies we will investigate whether radio jets suppress star formation in lergs in other environments , such as in groups and filaments .we also examined how the seds of r - agn differ as a function of the mass and mass - weighted age of the host galaxy . for this portion of the analysisthe effects of environment and accretion mode are controlled for by only considering lergs in the field , which comprise 66% of the sample .we have divided the field lergs into two mass and two age bins , which allows us to compare the seds of galaxies with similar evolutionary histories such that the uv and mid - ir emission should be largely independent of these two properties .the sample is split by the median mass and age , at @xmath58 gyr and @xmath59 , and these divisions are illustrated by the dashed lines in fig . [ control ] .panels ( a)-(d ) of figure [ agemass ] show the seds of the field lergs and their controls in the four mass and age bins , and are a more detailed examination of the field sample shown in panel ( a ) of fig .[ enviro ] .panel ( e ) shows the seds of the lergs in the four bins . as described in section [ methods ] , the seds are scaled by the median @xmath60 of each bin , so the differences we see between galaxies in different mass bins are not due to the scaling of fluxes with mass , but rather reflect the differences in mass - to - light ratios .the overall shapes of the seds in panel ( e ) are similar , although the galaxies with on average younger populations have stronger emission in the optical and 3.4 and 4.5 @xmath1 m bands than the galaxies with populations that are on average older .this is because the mass to light ratio of the galaxies with a younger age is lower than that of the galaxies with ages that are older .we see a dichotomy in the ir , as the lower - mass galaxies have stronger emission in this portion of their seds while the difference due to age is less .the same dichotomy is seen for the control galaxies , which is consistent with more massive early type galaxies , whether agn or not , being less dusty @xcite .the situation is less clear in the uv , as the effects of mass and age may be confounded .previously we have seen that field lergs on average have lower uv and mid - ir emission at 12 and 22 @xmath1 m than normal galaxies .we can now see which part of the sample contributes the most to this deficit . in panels( a ) and ( b ) concurrent uv and mid - ir deficits are seen in lower mass lergs , regardless of the population age .the differences for galaxies in panel ( a ) are fuv : 0.09 dex ; nuv : 0.1 dex ; 12 @xmath1 m : 0.09 dex ; 22 @xmath1 m : 0.09 dex , and for panel ( b ) they are fuv : 0.17 dex ; nuv : 0.17 dex ; 12 @xmath1 m : 0.06 dex ; 22 @xmath1 m : 0.11 dex .while the seds of the r - agn and controls in panel ( c ) are quite similar , a uv deficit may be present for the lergs in panel ( d ) as well .the mid - ir looks the same for the r - agn and controls , possibly because the emission is already at the minimum because of little dust and therefore the differences in sf are not reflected there . in future studies we will investigate whether radio jets suppress star formation in lergs in other environments , such as in groups and filaments .mid - infrared color - color diagrams are often used to identify agn ( assef et al .2010 , stern et al .2012 ) . by examining the wise mid - ir properties of x - ray selected agn@xcite suggested that a combination of the wise [ 3.4]@xmath61[4.6 ] and [ 4.6]@xmath61[12 ] colors could be used to identify high - excitation agn .the wise [ 3.4]@xmath61[4.6 ] color has been proposed to separate normal galaxies from those with elevated emission from the agn accretion disk ( assef et al .2010 , yan et al .2013 ) , while the wise [ 4.6]@xmath61[12 ] color can separate early - type galaxies with little warm dust from dusty star formers .@xcite found that the wise 22 @xmath1 m band is less effective at selecting luminous agn than the first three wise bands because of its shallower depth and because of increased contribution from star - formation activity at the wise sensitivity .@xcite also found an increased contribution from star formation at 24 @xmath1 m in a sample of 175 r - agn in the redshift range @xmath62 .since our sample of r - agn is composed of both high and low - excitation r - agn , it can be used to evaluate the effectiveness of wise colors at separating the two modes of accretion .additionally , the composite seds that we generate allow us to investigate the potential biases of agn studies that use only galaxies that have been individually detected in longer wise bands , which represent the minority of the bh2012 sample .figure [ colorcolor ] is a wise color - color plot of the field r - agn in the sample , together with the @xcite selection wedge .the dashed line shows the wise blazar strip defined by @xcite .we only plot those r - agn with a 3-sigma detection in the wise 12 @xmath1 m , all of which have 3-sigma detection at 3.4 and 4.6 @xmath1 m .the wise colors in these plots have not been dereddened , because the correction is very small .the hergs are shown in panel ( a ) , lergs in panel ( b ) , and the unclassified r - agn in panel ( c ) .the detection fraction at 12 @xmath1 m for lergs is 39% , for hergs is 81% , and for unclassified r - agn is 17% , while the overall detection fraction is 37% .we have also plotted the results of our bayesian estimations of the two wise colors as a red symbol in each panel . the location of the in all three plots, it would appear that the composite sed from the bayesian estimation is significantly bluer than most individually detected r - agn , as the red point is to the left of what appears by eye to be the average of the individual galaxies .this is due to the fact that we include all galaxies in the bayesian estimate , which pushes the composite measurement bluer than the observed colors of the higher signal - to - noise detections . while some of the hergs in panel ( a ) lie within the @xcite selection wedge , the majority ( 65 of the 88 plotted or 74% ) lie outside of it . since the composite color is outside of the @xcite selection wedge the weak and non - detections at 12@xmath1 m are probably outside as well .if these are included in counting then the fraction of hergs outside of the selection wedge increases to 77% .the lerg sample is shown in panel ( b ) , and nearly all of these r - agn lie outside of the @xcite selection wedge in the region occupied by non - agn , as most early - type galaxies have [ 3.4]@xmath61[4.6 ] colors bluer than 0.5 , regardless of whether they host an r - agn ( yan et al .2013 , toba et al .this combined with the results from panel ( a ) suggest that the @xcite selection can produce samples of hergs that are relatively free of lergs , although such samples will be incomplete . our results support those of @xcite , whose analysis of four complete samples of r - agn suggested that wise colors can not be used to select lergs .this is not surprising , since the lack of an accretion disk and associated obscuring torus in lergs would preclude strong mid - ir emission .the colors of the unclassified r - agn in panel ( c ) are similar to those of the lergs , suggesting that some of the r - agn in this category are lergs .we have found that the host galaxies of lergs have less uv and mid - ir emission than otherwise similar but radio - quiet galaxies , as shown in panel ( a ) of fig .[ accretion ] .when we subdivide the field lergs by mass and age , we see that the effect is strongest in the lower mass galaxies ( panels ( a ) and ( b ) of fig .[ agemass ] ) . in the lower mass and older age binthe difference between the control and r - agn seds in both fuv and nuv is 0.17 dex ( i.e. the sfr of the r - agn is @xmath070% of that of the controls ) .in contrast for higher mass galaxies the seds of lergs and radio - quiet galaxies are quite similar in the uv and mid - ir portions of the spectrum .one possible interpretation of this result is that in more massive galaxies it is the so called gravitational quenching @xcite that dominates , while in less massive it is the radio - mode agn .our results are consistent with the picture of lergs suppressing star formation by preventing gas from cooling onto their hosts from the cosmic web .it is quite likely that the majority of massive ellipticals go through cycles of r - agn activity .the duration of r - agn outbursts may be on the order of 10@xmath63 to 10@xmath64 yrs ( alexander & leahy 1987 , shabala et al .2008 ) , with the time between outbursts being of about the same duration ( schoenmakers et al .2000 , konar et al .therefore we do not expect that the galaxies that are currently detected as r - agn will show the full effect of r - agn feedback , as in a hypothetical case where ellipticals whose nucleus is never active are compared with those which are always ` on ' .thus , r - agn feedback may also suppress sf in more massive galaxies , but the differences between current r - agn and the control group ( currently inactive galaxies ) may be erased if the effects of feedback persist in between the active phases , e.g. , facilitated by the presence of hot , gaseous halos that are common in more massive galaxies and may store agn energy @xcite .we see little or no difference in the uv and mid - ir emission of r - agn and their controls in denser environments ( panels ( b ) and ( c ) of fig .[ enviro ] , suggesting that in clusters the gas accretion is already prevented by the hot cluster environment , regardless of the operation of the agn .clusters with strong cooling flows may lead to some bcgs having star formation rates that reach tens of solar masses per year , as shown by @xcite .our results indicate that such high sfrs due to cooling flows must be uncommon since the uv of the lerg bcgs is comparable to that of the cluster members and field galaxies , as shown in panel ( d ) of fig .[ enviro ] .in contrast to the results for lergs , we find no evidence of either positive or negative feedback in the hosts of hergs , as they have similar levels of uv emission to their controls ( panel ( c ) of fig .the herg hosts have much higher levels of star formation than lergs ( panel ( d ) of fig .6 ) , supporting the idea that they are fueled by large quantities of cold gas that also fuels star formation . our results are consistent with those of , who found that in a sample of @xmath080 nearby r - agn those that had elevated uv emission also had a mir excess .in addition to the radio jet , hergs may also impact their hosts via radiation from their accretion disks .roos et al .2015 modeled the impact of agn ionizing radiation and found that although high - excitation agn are capable of maintaining their halo gas in an ionized state , the giant molecular clouds that are the sites of star formation are too dense to be affected .our results for hergs support this picture , since we find no evidence of feedback in their hosts .we have shown that the seds of hergs and lergs are very different in the uv and ir , as shown in panel ( d ) of fig .[ accretion ] .the hergs have strongly elevated mid - ir emission from their obscuring tori .they are also more luminous at optical and uv wavelengths .this mirrors the situation at higher redshifts , where low - luminosity r - agn are hosted by massive ellipticals while more radio - luminous agn are observed to have large mid - ir and uv excesses ( baldi et al 2014 ) .our results are also in line with those of best & heckman 2012 who found that for their sample of r - agn , from which ours is drawn , hergs tend to be bluer in sdss bands than lergs .we find that this is also true for the galex and wise bands as the hergs have strongly elevated uv and mid - ir emission compared to lergs .the hergs also have a strong mid - ir excess relative to their controls , which suggests that the agn in these galaxies are of the radiatively efficient type .@xmath61[4.6 ] and [ 4.6]@xmath61[12 ] colors plotted against specific radio luminosity and specific star formation rate for lergs ( black circles ) hergs ( red stars ) , and unclassified r - agn ( blue triangles ) .only those r - agn with a 3-sigma detection are shown . the dashed line at [ 4.6]@xmath61[12]=1.5 shows the division between spirals and ellipticals , and the line at [ 3.4]@xmath61[4.6]=0.6 is the division between powerful agn and normal galaxies @xcite .these plots show that the unclassified galaxies are broadly similar to the lergs , and the [ 4.6]@xmath61[12 ] color is more effective at separating hergs from lergs.,scaledwidth=50.0% ] our results indicate that samples of hergs selected by their wise colors will be both incomplete and contaminated by non - hergs .@xcite suggested a simple cut at [ 3.4]@xmath61[4.6 ] @xmath65 0.8 could be used to select reliable and complete samples of agns .of the hergs shown in panel ( a ) , only 20 of 88 or 23% are above this line while 3 lergs and 2 unclassified r - agn would also be included in such a sample , suggesting that this selection criterion would miss most hergs .if we use the less stringent color cut at [ 3.4]@xmath61[4.6]=0.6 of @xcite , a greater number of hergs ( 33 of 88 or 38% of those plotted ) would be selected .however the contamination by non - hergs would increase , as 18 lergs and unclassified r - agn are above the [ 3.4]@xmath61[4.6 ] = 0.6 line .the composite herg [ 3.4]@xmath61[4.6 ] color is at [ 3.4]@xmath61[4.6 ] = 0.4 , which further shows that simple [ 3.4]@xmath61[4.6 ] color cuts are not effective at selecting complete and reliable samples of high - excitation agn .this is in line with the results of @xcite , who found that simple color cuts lose reliability at fainter fluxes .the wise [ 4.6]@xmath61[12 ] color has been suggested as a way to distinguish between lergs and hergs .@xcite found that nearly all of the hergs in their sample had a [ 4.6]@xmath61[12 ] color redder than 2.0 .this is also true for our sample , as 82% of the hergs are to the right of the [ 4.6]@xmath61[12 ] = 2.0 line .however the composite color in panel ( a ) falls just to the left of the line , indicating that most hergs with weak fluxes would be missed by this color cut . furthermore panels ( b ) and ( c ) show that a sample selected with the [ 4.6]@xmath61[12 ] = 2.0 cut would be composed mostly ( 73% ) of non - hergs .we conclude that using this color to select hergs results in high completeness , but very high contamination .we now consider the more sophisticated two - color selection wedge of @xcite . in panel( a ) of fig .[ colorcolor ] we see that the composite wise colors of the hergs in our sample lie outside of the wedge , suggesting that most hergs will be missed when using this selection technique .the @xcite selection wedge is based on the wise colors of x - ray selected luminous agn , so there must be some intrinsic difference in the wise colors of x - ray and radio - selected agn .this may be due to the anisotropic nature of the mid - ir emission from the obscuring torus ( maiolino & rieke 1995 , buchanan et al .2006 ) . besides being incomplete , samples of hergs selected via both wise colors will also be contaminated by lergs .only 23 of the 40 or 58% of the r - agn in fig .[ colorcolor ] that fall within the @xcite wedge are hergs , while most of the remainder ( 14 ) are lergs .we conclude that neither simple cuts in [ 3.4]@xmath61[4.6 ] and [ 4.6]@xmath61[12 ] colors nor combinations of these colors are effective at separating hergs from lergs .our results are in line with those of @xcite , who compared the wise [ 3.4]@xmath61[4.6 ] and [ 4.6]@xmath61[12 ] colors of the portion of the @xcite sample that is covered by 2mass and found that hergs and lergs occupy different regions of the diagram , with significant overlap between the two populations .while the mid - ir emission of hergs and lergs may arise from different processes , the overlap between the two makes it difficult to distinguish between the two types of r - agn using wise colors .to more directly see what drives the mid - ir colors , we plot the ssfr and srl as a function of the wise [ 3.4]@xmath61[4.6 ] and [ 4.6]@xmath61[12 ] colors , as shown in figure [ wiseradiossfr ] . in this figurethe lergs are shown as black circles , unclassified r - agn as blue triangles , and hergs as red stars . for panels ( a ) and ( c )only r - agn with 3-sigma detections in the wise 12 @xmath1 m band are shown , while in panels ( b ) and ( d ) only those with 3-sigma detections in the wise 4.6 @xmath1 m band are shown .the dotted line at [ 4.6]@xmath61[12]=1.5 shows the division between spirals and ellipticals , and the dotted line at [ 3.4]@xmath61[4.6]=0.6 is the division between powerful agn and normal galaxies @xcite . in panel( a ) we see a moderate ( spearman s @xmath66 ) correlation between star formation and the [ 4.6]@xmath61[12 ] color for all agn .this is similar to the results of @xcite , who matched wise sources brighter than 1 mjy to the sdss spectroscopic catalog to investigate the origin of the 12 @xmath1 m emission and showed that the wise [ 4.6]@xmath61[12 ] color can be used to gauge the star formation activity in a galaxy .panel ( b ) shows that there is also a moderate ( @xmath67 ) correlation between ssfr and the [ 3.4]@xmath61[4.6 ] color for hergs . in panels ( c ) and ( d )we see that the srl also shows a correlation with both wise colors , with the stronger ( @xmath68 ) correlation being to the [ 3.4]@xmath61[4.6 ] color for both lergs and unclassified r - agn .however , neither the ssfr nor the srl correlates very tightly with these mid - ir colors . in panels ( a ) and ( b )we see that although the hergs on average may have a higher ssfr than the lergs , star formation can not be used to distinguish between the two populations as they both span the range of ssfrs .this is perhaps more clearly seen in panel ( b ) , where a greater number of both hergs and lergs are plotted .both hergs and lergs also span the range of srls , as shown in panels ( c ) and ( d ) , indicating that the srl can not be used to distinguish between the two types of r - agn .we have investigated the uv to mid - ir seds of an aggregate sample of radio - loud agn and have compared them to the seds of a control sample of galaxies that have the same quality of the data , similar environments and sf histories as the r - agn sample .we have in addition re - visited the question of r - agn selection and classification based on mid - ir colors .our conclusions can be summarized as follows : 1 .field lergs in hosts whose mass is below @xmath59 have less uv and mid - ir emission than similar radio - quiet galaxies , which suggests that lergs may have a role in suppressing star formation in these galaxies , perhaps as a result of radio - agn feedback .the suppression of uv is around 0.2 dex and is statistically significant .the potential feedback that we detect may be a lower limit because we see a difference in sfr even though r - agn probably have a duty cycle , i.e , the ones we now detect as r - agn are not always on , and vice versa .in contrast for galaxies whose mass is above @xmath59 the seds of lergs and radio - quiet galaxies are quite similar in the uv and mid - ir portions of the spectrum , suggesting another suppression mechanism , e.g. the gravitational quenching , or that the effects of agn quenching persist between agn duty cycles , possible because of more massive galaxies more often have got gaseous halos .the composite seds for lerg cluster members and bcgs are very similar to that of their respective control samples , suggesting that r - agn feedback in denser environments is not necessary to maintain quiescence .the uv emission of r - agn bcgs is very similar to that of r - agn in other environments , suggesting that significantly enhanced star formation in bcgs due to cooling flows is uncommon .4 . in contrast to the lergs , we find that the uv emission in hergs is similar to that of radio - quiet galaxies , suggesting that hergs have little effect on the star formation in their hosts. the strongly elevated mid - ir emission of hergs probably originates from the obscuring torus that is present in high - excitation agn .the differences in uv and mid - ir emission between lergs and hergs are also seen when the r - agn are classified by their specific radio luminosity rather than excitation mechanism .samples of hergs selected using either the wise [ 3.4]@xmath61[4.6 ] color , the [ 4.6]@xmath61[12 ] color , or a combination of both colors will be incomplete and will be contaminated by a significant number of non - hergs . neither the mid - ir colors nor the ssfr can be used to securely discriminate between lergs and hergs .our results present strong evidence for suppressed sf in moderately massive non - cluster radio galaxies , which may be the result of agn feedback .more massive ellipticals and cluster r - agn show no suppression of sf compared to control groups , which either means that other mechanisms maintain low sf in those galaxies , or that the effect of suppression in such galaxies extends to periods in between the active phases of the central engine , perhaps due to the presence of hot gaseous halos that store the energy released by agn during the active phases ( e.g. , kormendy et al .2009 ; salim et al .our characterization of the possible level of radio - mode feedback and identification of the regimes in which it may be important should serve as a benchmark for theoretical studies and simulations .we thank the anonymous referee for numerous helpful suggestions .this work was partially supported through nasa adap award nnx12ae06 g .galex ( galaxy evolution explorer ) is a nasa small explorer , launched in april 2003 .we gratefully acknowledge nasa s support for construction , operation , and science analysis for the galex mission , developed in cooperation with the centre national detudes spatiales ( cnes ) of france and the korean ministry of science and technology .this study uses data from the sdss archive .funding for sdss - iii has been provided by the alfred p. sloan foundation , the participating institutions , the national science foundation , and the u.s .department of energy office of science .the sdss - iii web site is http://www.sdss3.org/. sdss - iii is managed by the astrophysical research consortium for the participating institutions of the sdss - iii collaboration including the university of arizona , the brazilian participation group , brookhaven national laboratory , university of cambridge , carnegie mellon university , university of florida , the french participation group , the german participation group , harvard university , the instituto de astrofisica de canarias , the michigan state / notre dame / jina participation group , johns hopkins university , lawrence berkeley national laboratory , max planck institute for astrophysics , max planck institute for extraterrestrial physics , new mexico state university , new york university , ohio state university , pennsylvania state university , university of portsmouth , princeton university , the spanish participation group , university of tokyo , university of utah , vanderbilt university , university of virginia , university of washington , and yale university .this publication makes use of data products from the wide - field infrared survey explorer , which is a joint project of the university of california , los angeles , and the jet propulsion laboratory / california institute of technology , and neowise , which is a project of the jet propulsion laboratory / california institute of technology .wise and neowise are funded by the national aeronautics and space administration .abazajian , k. n. , adelman - mccarthy , j. k. , ageros , m. a. , et al .2009 , , 182 , 543 ahn , c. p. , alexandroff , r. , allende prieto , c. , et al .2014 , , 211 , 17 alexander , p. , & leahy , j. p. 1987, , 225 , 1 allen , s. w. , dunn , r. j. h. , fabian , a. c. , taylor , g. b. , & reynolds , c. s. 2006 , , 372 , 21 assef , r. j. , kochanek , c. s. , brodwin , m. , et al .2010 , , 713 , 970 assef , r. j. , stern , d. , kochanek , c. s. , et al .2013 , , 772 , 26 baldwin , j. a. , phillips , m. m. , & terlevich , r. 1981 , , 93 , 5 barnes , j. e. , & hernquist , l. 1996 , , 471 , 115 becker , r. h. , white , r. l. , & helfand , d. j. 1995 , , 450 , 559 best , p. n. 2004 , , 351 , 70 best , p. n. , & heckman , t. m. 2012 , , 421 , 1569 best , p. n. , kauffmann , g. , heckman , t. m. , & ivezi , . 2005a , , 362 , 9 best , p. n. , kauffmann , g. , heckman , t. m. , et al . 2005b , , 362 , 25 bower , r. g. , benson , a. j. , malbon , r. , et al .2006 , , 370 , 645 bruzual , g. , & charlot , s. 2003 , , 344 , 1000 buchanan , c. l. , gallimore , j. f. , odea , c. p. , et al .2006 , , 132 , 401 burns , j. o. 1990 , , 99 , 14 buttiglione , s. , capetti , a. , celotti , a. , et al .2010 , , 509 , a6 cattaneo , a. , faber , s. m. , binney , j. , et al .2009 , , 460 , 213 charlot , s. , & fall , s. m. 2000 , , 539 , 718 ciotti , l. , ostriker , j. p. , & proga , d. 2010 , , 717 , 708 condon , j. j. 1992 , , 30 , 575 condon , j. j. , cotton , w. d. , greisen , e. w. , et al .1998 , , 115 , 1693 crockett , r. m. , shabala , s. s. , kaviraj , s. , et al .2012 , , 421 , 1603 croft , s. , van breugel , w. , de vries , w. , et al .2006 , , 647 , 1040 croton , d. j. , springel , v. , white , s. d. m. , et al .2006 , , 365 , 11 da cunha , e. , charlot , s. , & elbaz , d. 2008 , , 388 , 1595 dekel , a. , & birnboim , y. 2008 , , 383 , 119 de ruiter , h. r. , parma , p. , fanti , r. , & fanti , c. 2015 , , 581 , a33 donoso , e. , li , c. , kauffmann , g. , best , p. n. , & heckman , t. m. 2010 , , 407 , 1078 donoso , e. , yan , l. , tsai , c. , et al . 2012 , , 748 , 80 dugan , z. , bryan , s. , gaibler , v. , silk , j. , & haas , m. 2014 , , 796 , 113 dunn , r. j. h. , & fabian , a. c. 2006 , , 373 , 959 dunn , r. j. h. , & fabian , a. c. 2008 , , 385 , 757 edge , a. c. , oonk , j. b. r. , mittal , r. , et al .2010 , , 518 , l47 fabian , a. c. 1994 , , 32 , 277 fabian , a. c. 2012 , , 50 , 455 fabian , a. c. , sanders , j. s. , allen , s. w. , et al .2003 , , 344 , l43 fanaroff , b. l. , & riley , j. m. 1974 , , 167 , 31p graham , j. a. 1998 , , 502 , 245 granato , g. l. , de zotti , g. , silva , l. , bressan , a. , & danese , l. 2004 , , 600 , 580 grkan , g. , hardcastle , m. j. , & jarvis , m. j. 2014 , , 438 , 1149 hao , j. , mckay , t. a. , koester , b. p. , et al .2010 , , 191 , 254 hardcastle , m. j. , kraft , r. p. , worrall , d. m. , et al .2007 , , 662 , 166 heckman , t. m. , & best , p. n. 2014 , , 52 , 589 heckman , t. m. , smith , e p. , baum , s. a. , et al .1986 , , 311 , 526 ho , l. c. 2008 , , 46 , 475 inskip , k. j. , villar - martn , m. , tadhunter , c. n. , et al .2008 , , 386 , 1797 karouzos , m. , jarvis , m. j. , & bonfield , d. 2014 , , 439 , 861 karouzos , m. , i m , m. , trichas , m. , et al .2014 , , 784 , 137 kauffmann , g. , heckman , t. m. , & best , p. n. 2008 , , 384 , 953 kauffmann , g. , heckman , t. m. , tremonti , c. , et al .2003 , , 346 , 1055 kaviraj , s. , shabala , s. s. , deller , a. t. , & middelberg , e. 2015 , , 454 , 1595 konar , c. , hardcastle , m. j. , jamrozy , m. , & croston , j. h. 2013 , , 430 , 2137 kormendy , j. , fisher , d. b. , cornell , m. e. , & bender , r. 2009 , , 182 , 216 laing , r. a. , jenkins , c. r. , wall , j. v. , & unger , s. w. 1994 , the physics of active galaxies , 54 , 201 lonsdale persson , c. j. , & helou , g. 1987 , , 314 , 13 magliocchetti , m. , lutz , d. , rosario , d. , et al .2014 , , 442 , 682 mainzer , a. , bauer , j. , grav , t. , et al .2011 , , 731 , 53 maiolino , r. , & rieke , g. h. 1995 , , 454 , 95 martin , d. c. , fanson , j. , schiminovich , d. , et al .2005 , , 619 , l1 massaro , f. , dabrusco , r. , tosti , g. , et al .2012 , , 750 , 138 mateos , s. , alonso - herrero , a. , carrera , f. j. , et al .2012 , , 426 , 3271 mathews , w. g. , & brighenti , f. 2003 , , 41 , 191 meisenheimer , k. , haas , m. , mller , s. a. h. , et al .2001 , , 372 , 719 narayan , r. , & yi , i. 1994 , , 428 , l13 nulsen , p. e. j. , jones , c. , forman , w. r. , et al .2007 , heating versus cooling in galaxies and clusters of galaxies , 210 odea , c. p. , baum , s. a. , privon , g. , et al .2008 , , 681 , 1035 pace , c. , & salim , s. 2014 , , 785 , 66 peek , j. e. g. , & schiminovich , d. 2013 , , 771 , 68 peterson , j. r. , kahn , s. m. , paerels , f. b. s. , et al .2003 , , 590 , 207 reviglio , p. , & helfand , d. j. 2006 , , 650 , 717 roos , o. , juneau , s. , bournaud , f. , & gabor , j. m. 2015 , , 800 , 19 sadler , e. m. , ekers , r. d. , mahony , e. k. , mauch , t. , & murphy , t. 2014 , , 438 , 796 salim , s. , fang , j. j. , rich , r. m. , faber , s. m. , & thilker , d. a. 2012 , , 755 , 105 salim , s. , lee , j. c. , ly , c. , et al .2014 , , 797 , 126 salim , s. , rich , r. m. , charlot , s. , et al .2007 , , 173 , 267 satyapal , s. , ellison , s. l. , mcalpine , w. , et al .2014 , , 441 , 1297 schoenmakers , a. p. , de bruyn , a. g. , rttgering , h. j. a. , van der laan , h. , & kaiser , c. r. 2000 , , 315 , 371 shabala , s. s. , ash , s. , alexander , p. , & riley , j. m. 2008 , , 388 , 625 shabala , s. s. , kaviraj , s. , & silk , j. 2011 , , 413 , 2815 silk , j. , & norman , c. 2009 , , 700 , 262 singh , v. , ishwara - chandra , c. h. , sievers , j. , et al .2015 , , 454 , 1556 smith , m. w. l. , gomez , h. l. , eales , s. a. , et al .2012 , , 748 , 123 stern , d. , assef , r. j. , benford , d. j. , et al .2012 , , 753 , 30 strauss , m. a. , weinberg , d. h. , lupton , r. h. , et al .2002 , , 124 , 181 toba , y. , oyabu , s. , matsuhara , h. , et al .2014 , , 788 , 45 van breugel , w. j. m. , & dey , a. 1993 , , 414 , 563 van breugel , w. , fragile , c. , croft , s. , et al .2004 , the interplay among black holes , stars and ism in galactic nuclei , 222 , 485 van velzen , s. , falcke , h. , schellart , p. , nierstenhfer , n. , & kampert , k .- h .2012 , , 544 , a18 voit , g. m. , bryan , g. l. , oshea , b. w. , & donahue , m. 2015 , , 808 , l30 wright , e. l. , eisenhardt , p. r. m. , mainzer , a. k. , et al .2010 , , 140 , 1868 yan , l. , donoso , e. , tsai , c .- w . , et al .2013 , , 145 , 55 yang , x .- h . , chen , p .- s ., & huang , y. 2015 , , 449 , 3191 zinn , p .- c . , middelberg , e. , norris , r. p. , & dettmar , r .- j .2013 , , 774 , 66

<S> the feedback from radio - loud active galactic nuclei ( r - agn ) may help maintain low star formation ( sf ) rates in their early - type hosts , but the observational evidence for this mechanism has been inconclusive . </S><S> we study systematic differences of aggregate spectral energy distributions ( seds ) of various subsets of @xmath04000 low - redshift r - agn from best & heckman ( 2012 ) with respect to ( currently ) inactive control samples selected to have matching redshift , stellar mass , population age , axis ratio , and environment . </S><S> aggregate seds , ranging from the ultraviolet ( uv ) through mid - infrared ( mid - ir , 22 @xmath1 m ) , were constructed using a bayesian method that eliminates biases from non - detections in _ galex _ and _ wise_. we study rare high - excitation sources separately from low - excitation ones , which we split by environment and host properties . </S><S> we find that both the uv and mid - ir emission of non - cluster r - agns ( 80% of sample ) are suppressed by @xmath00.2 dex relative to that of the control group , especially for moderately massive galaxies ( log @xmath2 11 ) . </S><S> the difference disappears for high - mass r - agn and for r - agn in clusters , where other , non - agn quenching / maintenance mechanisms may dominate , or where the suppression of sf due to agn may persist between active phases of the central engine , perhaps because of the presence of a hot gaseous halo storing agn energy . </S><S> high - excitation ( high accretion rate ) sources , which make up 2% of the r - agn sample , also show no evidence of sf suppression ( their uv is the same as in controls ) , but they exhibit a strong mid - ir excess due to agn dust heating . </S>

1004.3331

a problematic situation with the introduction in the study of the standard model of particle physics ( sm ) , which constitutes the main motivation of this article , is the fact that usually the process of formalizing the model is one of the aspects that cause difficulties to students .however , it is intended to formalize knowledge , but what is sought is based on a series of paradigms and rules for creating a solid foundation on which to work , which can be modified when necessary @xcite . in this contextis the sm which describes , with subtle precision and mathematical elegance , all known particles and interactions , except gravity .all model predictions have been corroborated , however , the large hadron collider ( lhc ) is expected to be a new physics or physics beyond the sm , in the tev scale , which can manifest itself in two ways : through signals involved in the production of new particles or deviations from sm predictions .consequently , a clear understanding of the sm and , in general , quantum field theory is of great importance for particle physics , this paper shows how to make an approach to sm to introduce students quickly in the formalism and basic concepts .everything that exists in nature is made of material particles ( called fermions ) which interact , either because they decay or because they respond to a force due to the presence of another particle .the decay of a fermion into a different one is explained by the action of a third call mediator boson .similarly , the interaction of two particles through the fields that originate , can be interpreted considering that the two particles exchange a mediator boson . in quantum mechanicsfermions or bosons can be represented by fields with some intrinsic properties .the interaction process is an exchange of energy and momentum , between fields , governed by laws of conservation of intrinsic quantities ( e.g. , some kind of charge ) . to explain the interaction between a fermion and a boson, we use a mechanical model consisting of two coupled mass - spring systems .the lagrangian of the system is given by : @xmath0 the kinetic energy of the first oscillator can represent a free fermion ( e.g. , electron ) , the second oscillator a free boson ( e.g. , electromagnetic field ) and potential energy , the fermion - boson interaction ( e.g. , electron in electromagnetic field ) . note that the interaction is characterized by a constant ( spring constant ) .therefore , we will always have three classes of lagrangians : the free fermion , the free boson and fermion - boson interaction .it is clear that the interaction occurs at a predetermined time , called mean - lifetime . in electromagnetic theory , the lagrangian for a free particle of mass @xmath1 and an free electromagnetic field represented for a potential , @xmath2 , in vacuum , is written as @xmath3 the introduction of the interaction reduces simply to the replacement of ordinary momentum by canonical momentum ( as is usual in electrodynamics ) : @xmath4 .substituting we obtain the lagrangian for a particle interacting : @xmath5 in the formalism of relativistic quantum mechanics a free particle can be represented by a field called spinor,@xmath6 . in the continuum ,the lagrangian ( @xmath7 ) is replaced by a lagrangian density ( @xmath8 ) , the generalized coordinates ( @xmath9 ) by the fields ( @xmath10 ) and the generalized velocities ( @xmath11 ) by the field gradient ( @xmath12 ) : @xmath13 . for simplicity , in what follows , we use natural units where @xmath14 and @xmath15 .the lagrangian density for a free particle and an electromagnetic field , equivalent to the equation , fully consistent with the principles of quantum mechanics and special relativity ( @xmath16 ) , is given by : @xmath17 where @xmath18 is defined as @xmath19 and @xmath20 .the first term is called the dirac lagrangian . in order to introduce the interaction we replace the ordinary momentum @xmath21 by canonical momentum @xmath22 , where @xmath9 is the coupling constant , which in this case corresponds to the charge . in language operatorthe momentum of the particle is given by @xmath23 , and @xmath24 , then , the introduction of the interaction for the relativistic quantum particle interacting with the field , translates to replace the normal derivative by the covariant derivative in the lagrangian , @xmath25 : @xmath26 here , @xmath27 is a current satisfying : @xmath28 , indicating that it is a conserved quantity @xcite . ]the interaction terms can be represented by feynman diagrams , we have six possibilities : creation and annihilation of an electron - positron pair , or absorption of a photon by an electron or a positron .consider the operator rotation , @xmath29 , where an infinitesimal rotation is defined by @xmath30 ; when we expand the right side in a taylor series around @xmath31 , we get , @xmath32 . where the @xmath33 operator is called the generator of rotations .consequently , a rotation is given by an infinite number of infinitesimal rotations @xcite : @xmath34^{n}=e^{-ij\theta}.\label{eq : rotation - operator}\ ] ] it is not difficult to demonstrate that in order to conserve the probability the @xmath29 operator must be unitary , @xmath35 ; this is true only if your generator is hermitian @xmath36 ..some quantum operators , symmetry and conservation laws .[ cols="^,^,^,^",options="header " , ] all operators of quantum mechanics can be constructed in the same way and have the form of a rotation transformation .noether s theorem establishes that if an action is invariant under a transformation group ( symmetry ) , there are one or more quantities ( constants of motion ) which are associated with these transformations , in other words , to each symmetry corresponds a conservation law @xcite .we see that each operator has a generator , and the generators belong to a particular lie group .the commutators of these generators are linear combinations that define the group algebra , meaning the commutation of two generators of the group produces a third generator of the same group ; this can be shown using the definition @xcite .we can change the phase of all the fields that represent the particles ( local phase change ) , @xmath37 note that @xmath38 is a function that depends on the coordinates @xmath39 and @xmath9 is the generator . in this casethe dirac lagrangian ceases to be invariant : @xmath40 we can demand that the lagrangian is invariant under local phase transformation , by removing how the additional term that appears . by making the particle to interact with potential @xmath41 , see equation , the lagrangian is composed of three terms , the lagrangian particle to which is added a term due to the potential and another term due to local phase change : @xmath42 we can calibrate the potential in such a way that the term of change of phase is cancelled : @xmath43 then @xmath44 we see that this lagrangian , which represents an interacting particle with a potential ; is invariant under a local phase change , if the potential is calibrated .the fields represented by such calibrated potentials are called gauge fields .consider that we have @xmath45 fields represented by potentials , @xmath46 @xmath47 , the covariant derivative takes the form : @xmath48 and gauge invariance occurs under the local transformation @xmath49 . by means of the operator of change of phase the representation of the lie groupcan be obtained doing @xmath50 , where @xmath51 corresponds to the coupling constant and @xmath52 to the group generators , i.e. , @xmath53 . in this way we associate a potential @xmath46 a group represented by its generator @xmath52 .the commutators of the generators of a lie group is a linear combination of generators : @xmath54=if_{abc}t^{c}$ ] , where the coefficients of the linear combination @xmath55 , are the structure constants of the group.by analogy with the electromagnetic field tensor , the field tensor @xmath56 is defined as generalized as : @xmath57.\label{eq : tensor - yang - mills}\ ] ] substituting the covariant derivative and commutator of the generators of the group is obtained : @xmath58 we see that the first two terms of this expression correspond to a tensor analogous to the electromagnetic field and the last term is the fact that generators are not abelian group ; this allows to create no - abelians gauge theories or yang - mills theories @xcite . with the gauge field tensoris constructed from yang - mills lagrangian for the free field : @xmath59 note its similarity to the free electromagnetic field lagrangian .to build the model for the strong interaction we propose a lagrangian invariant under local gauge transformations of the @xmath60 group , the subscript @xmath61 indicates that the transformations only act on particles with color charge .the strong interaction is represented by means eight gauge fields for gluons @xmath62 @xmath63 and each of these fields is associated generators @xmath64 , being @xmath65 the gell - mann matrices : @xmath66 .the coupling constants of the fields with gluons are @xmath67 .lie algebra , is obtained with the following rule of commutation : @xmath54=if_{ab}^{c}t_{c}$ ] , where @xmath68 it is the structure constant of the @xmath69 group .the covariant derivative is given by : @xmath70 applying the definition obtain the tensor field : @xmath71 and for the free field lagrangian is @xmath72 a very interesting aspect of sm is that the electromagnetic interactions and weak interactions are combined into a single , so called , electroweak interaction . for electroweak interactions we propose a lagrangian invariant under local gauge transformations of the @xmath73 group , the subscript @xmath74 means that these transformations only act on components of the left chiral fermions and @xmath75 is the hypercharge quantum number .electroweak interactions are represented by four gauge fields @xmath76 @xmath77 and @xmath78 .each of these fields are associated generators @xmath79 of the weak isospin group @xmath80 and @xmath81 , corresponding to the @xmath82 group , respectively . here@xmath83 are the pauli matrices , @xmath84 the @xmath85 unit matrix and @xmath75 hypercharge quantum number .the coupling constants of the flavor change with the electroweak field are @xmath86 and @xmath87 .the group algebra is given by the following commutation rules : @xmath88=i\epsilon_{ij}^{k}\tau_{k}$ ] and @xmath89=0 $ ] , where @xmath90 is the constant of structure of the @xmath91 group .the covariant derivative is given by : @xmath92 applying again the definition one obtains the tensor field : @xmath93 note that the gauge fields @xmath76 are non - abelian yang - mills fields and gauge field @xmath78 is a abelian field , because the structure constant of the group is zero . for the free fieldis @xmath94 performing the matrix operations indicated in the covariant derivative and by means the linear combination of the first two components of the gauge field , @xmath95 y @xmath96 , one can build charged intermediate vector bosons , @xmath97 y @xmath98 : @xmath99 and with the mixture of the third component of the gauge field , @xmath100 with the gauge field @xmath78 , the neutral vector boson is obtained @xmath101 and the boson of the electromagnetic field , @xmath102 ( photon ) ( superscript @xmath103 indicates that it is a field with neutral electric charge ) : @xmath104 where the mixing angle , @xmath105 , called weinberg angle ; which is defined by the ratio of the electroweak coupling constants , @xmath106 and the value of the electron charge , @xmath107 , which represents the unification of the weak field with the electromagnetic @xcite . the hypercharge @xmath75 is related to the electric charge @xmath108 by the gell - mann nishijima relation @xcite : @xmath109 , here @xmath110 it is third generator of the @xmath91 group and @xmath75 hypercharge quantum number .substituting these definitions , and in the covariant derivative , we obtain : @xmath111 where @xmath112 . fields representing fermions can be decomposed into a left and right chiral components : @xmath113 , using chiral projectors @xmath114 and @xmath115 , where @xmath116 and @xmath117 .fermions , quarks and leptons , are classified into three groups called generations .a generation is formed by quarks and an leptons with different charges , and generations are ordered from the lightest to the heaviest .the left component of the leptons is given by a doublet @xmath118 , where @xmath119 and @xmath120 ; for quarks @xmath121 , where @xmath122 ; @xmath123 with @xmath124 an element of the cabibbo - kobayashi - maskawa ( ckm ) matrix and @xmath125 .the ckm mixing matrix is due to the fact that weak interaction does not act equally on quarks but is divided between them , however , the theoretical origin of this mixture has not been determined yet .the right component of the leptons is given by a singlet : @xmath126 , because in the sm neutrinos are massless , they do not have a right chiral component for them ; while for the quarks @xmath127 . to describe the leptons and quarks interacting with the weak and electromagnetic fields , we use the lagrangian . as the left chiral components ( @xmath128 ) and right ( @xmath129 )have different local phase transformations , the generator of the left component belongs to the @xmath91 group , while the right ones only to the @xmath130 group , the term mass @xmath131 is not gauge invariant , thus is absent in the lagrangian : @xmath132 where covariant derivatives are given by @xmath133 and @xmath134 .consequently , the lagrangian describes only massless leptons interacting with the weak and electromagnetic fields .performing operations gives : @xmath135 where @xmath136 and @xmath137 .similarly , the lagrangian of the quark sector is given by : @xmath138 to generate the mass terms , we use the higgs mechanism , which starts from the introduction of a doublet complex scalar fields @xmath139 .the real part of the neutral field can be divided into two components , @xmath140 , so that when the expected value in vacuum is nonzero : @xmath141 , there is a spontaneous breaking of symmetry ( res ) according to the scheme @xmath142 where @xmath108 is the electromagnetic charge @xcite . to generate the mass of the bosons ,we define the kinetic lagrangian : @xmath143 here @xmath144 is given by ( [ eq : derivada covariante electrodebil ] ) , for the ground state ( @xmath145 ) , the lagrangian becomes : @xmath146 where , @xmath147 , is the w boson mass ; @xmath148 , z boson mass and the @xmath41 boson remains massless .this makes the weak interaction a short range and long range electromagnetic interaction .the leptons and quarks acquire mass through interaction between fermion fields and the higgs field which is called the yukawa lagrangian : @xmath149 where @xmath150 ; mass terms are defined as @xmath151 . for the ground state ( @xmath145 ) ,the lagrangian becomes : @xmath152v_{ij}d_{j}\phi^{+}\\ & & + \frac{g}{\sqrt{2}m_{w}}\overline{d_{i}}\left[m_{d_{j}}p_{l}-m_{u_{i}}p_{r}\right]v_{ij}^{\ast}u_{j}\phi^{-}.\end{aligned}\ ] ]in this paper , using a small number of concepts , we have obtained the main standard model lagrangians . for brevity, we have omitted some issues such as the spontaneous breaking of symmetry , the higgs mechanism , the higgs potential , the faddeev - popov lagrangian and others .however , students can find and investigate some supplementary texts .direction of investigations ( din ) of the universidad pedaggica y tecnolgica de colombia ( uptc ) and the young investigator program of colciencias .

<S> so far , the standard model of particle physics ( sm ) describes the phenomenology observed in high energy physics . in the large hadron collider ( lhc ) is expected to find the higgs boson , which is an essential part of sm ; also expects to see new particles or deviations from the sm , which would be evidence of other truly fundamental theory . </S><S> consequently , a clear understanding of the sm and , in general , quantum field theory is of great importance for particle physics , however , students face a formalism and a set of concepts with which they are unfamiliar . </S><S> this paper shows how to make an approach to sm to introduce students to the formalism and some fundamental concepts . </S>

1408.4850

we consider the totally asymmetric simple exclusion process ( tasep ) on @xmath9 with particle - dependent rates .each site can be occupied by at most a particle ( exclusion constraint ) .particles try to jump to their neighboring site with a given rate and the jump occurs provided the target site is empty . depending on the initial conditions and / or choice of the jump rates , the macroscopic density of particle can have discontinuities , also called shocks . for particle - independent jump rates, it is known that the stationary measures are either blocking measures ( with no current ) or product bernoulli measures with fixed density @xcite . in this setting , naturally one considered random initial conditions with bernoulli initial conditions but density @xmath10 to the left and @xmath11 to the right of the origin with the condition @xmath12 . the shock moves with speed @xmath13 and has gaussian fluctuations in the scale @xmath14 for large time @xmath15 @xcite ( the shock position is often defined by the position of a second class particle ) .the same holds for the fluctuations of a tagged particle around the macroscopic shock position .shocks are located where the characteristics of the associated pde meet .the gaussian fluctuations and the @xmath14 scale are due to the randomness in the initial condition ( see @xcite for a physical argument ) .the reason is that at the particles at the shock are non - trivially correlated with two regions at time @xmath16 which are of order @xmath15 away from the origin ( specifically , the positions where the characteristics meeting at the shock start ) , whose fluctuations are gaussian and of order @xmath14 . since tasep belongs to the kardar - parisi - zhang ( kpz ) universality class , the fluctuations created by the dynamics until time @xmath15 are @xmath17 and therefore are not seen in the @xmath14 scale . in our previous work@xcite we analyzed one situation where shocks occur with non - random initial condition . in that case , the shock width is of order @xmath18 and in that scale the fluctuations of a tagged particle has a law that is the maximum of two random variables , because the pieces of information coming from each of the two characteristics that meet at the shock are asymptotically independent . in this paperwe consider the situation where at time @xmath19 particles occupy the whole @xmath20 , particles starting from @xmath21 have unit jump rate and the ones starting from @xmath22 have jump rate @xmath1 .we consider large time @xmath15 and the critical scaling @xmath23 .in this situation the pieces of information from the two characteristics are strongly correlated and , at the same time , the system differs from the constant density case .therefore we call this a critical scaling .we obtain the limiting process describing the fluctuations of the particles positions around the shock ( see theorem [ thmmain ] ) .we then perform a numerical study of the convergence of the distribution function to the one expected in the case of a shock with density jump of order one obtained in @xcite . for the numerical evaluation we use the matlab program developed by bornemann @xcite and we see that the convergence is surprisingly fast , see figure [ figdistr ] . as discussed in section [ sectlpp ] , a similar critical scaling , where particles starts from @xmath24 but the first @xmath25 particles havejump rate @xmath1 has been considered in the context of last passage percolation . in the large time @xmath15 limit, the distribution function of a particle that is around the origin at time @xmath15 has a bbp distribution function @xcite .the rest of the paper is organized as follows . in section [ sectmainresults ]we define the model , discuss the macroscopic behavior , provide the main result , and finally present the numerical results .we also present the result in terms of a last passage percolation model . in section [ sectasympt ]we provide the proof of the main theorem . in the appendixwe give an explicit formula in terms of airy function of the kernel appearing in the main theorem .ferrari was supported by the german research foundation via the sfb 1060b04 project .p. nejjar is grateful for the support of the bonn international graduate school ( bigs ) .we consider the totally asymmetric simple exclusion process and assign to each particle a label .we denote by @xmath26 the position of particle with label @xmath25 at time @xmath15 .the dynamics is as follows : each particle independently tries to jump to its right site and the jump occurs if that site is empty .the rate with which particle @xmath25 tries to jump from @xmath27 to @xmath28 is denoted by @xmath29 .the dynamics preserves the order of particle and we use the convention to label from right to left , i.e. , @xmath30 for any @xmath31 and time @xmath15 .the process is well - defined since in a finite time no information is coming from @xmath32 ( as one can see either from the graphical construction @xcite or from the general hille - yoshida semigroup approach for interacting particle systems , see section i.3 - 4 of @xcite ) . in this paperwe consider the following specialization of the model : @xmath33 as there are two jump rates involved , we call this model _ two - speed tasep_. further , we call particles with label @xmath34 _ @xmath1-particles _ and the ones with label @xmath35 are called _ normal particles_. we will be interested in the fluctuations of particle positions of normal particles which are around the shock created by the @xmath1-particles .first we discuss the macroscopic picture . in the main resultwe describe the fluctuations with respect to the deterministic macroscopic behavior , given as follows . under hydrodynamicscaling , the evolution of the particle density @xmath36 of the normal particles is governed by the burgers equation @xmath37 the first normal particle is , however blocked by the last @xmath1-particle , which starts at the origin and moves with an average speed @xmath38 .therefore , the macroscopic density profile of the normal particles can be obtained by solving for @xmath39 $ ] under the boundary condition .the system we consider corresponds to the initial condition @xmath40 for @xmath41 and the macroscopic density profile of the normal particles is as follows : for @xmath42 , there is a rarefaction fan given by @xmath43,\\ \max\{0,1-\alpha/2\},&\textrm{for } \xi\in [ \min\{1,\alpha-1\},\alpha/2 ] , \end{array}\right.\ ] ] while for @xmath44 there is a shock moving with speed @xmath45 , @xmath46 .\end{array}\right.\ ] ] as a consequence , the particle with label @xmath47 will be around the shock position if @xmath48 .in this paper we focus at the critical regime where the discontinuity of the density is small . for a fixed @xmath49 and a large observation time @xmath6 ,we scale the jump rate of the @xmath1-particles critically , i.e. , we consider @xmath50 in view of the macroscopic description , particles with number given by is around the shock and its position is around @xmath51 . unlike in the macroscopic shock studied in @xcite , where around the shock position particles are correlated over a @xmath18 scale , in the present situation particlesare correlated over a @xmath52 scale .therefore we consider the scaling @xmath53 and define the accordingly scaled particle position process by @xmath54 our main analytic result is the limit process .[ deflimitprocess ] define the extended kernel @xmath55 the curves can be chosen as follows .for any choice of @xmath56 satisfying @xmath57 and , we can set @xmath58 and @xmath59 ( oriented with increasing imaginary parts ) . the limit process @xmath60 is defined by its finite - dimensional distribution : for any given @xmath61 , @xmath62 where @xmath63 .an explicit expression of @xmath64 in terms of airy functions is given in appendix [ sectappendix ] .with this definition we can state our main analytic result , proven in section [ sectasympt ] .[ thmmain ] it holds @xmath65 in the sense of finite dimensional distributions . in some special cases or limitswe recover previous known processes .for example : * for @xmath66 we have the flat tasep and see the airy@xmath67 process @xcite : @xmath68 .* when @xmath69 a rarefaction fan is created . at his left edge , @xmath70 becomes the airy@xmath71 process @xcite : @xmath72 .inside the rarefaction fan @xmath70 becomes the airy@xmath73 process @xcite .for instance , in the middle of the rarefaction fan : @xmath74 . * for @xmath75 ,there is a shock and @xmath70 is a transition process between two @xmath76 processes .indeed , @xmath77 .further , when @xmath78 one should recover the macroscopic shock picture and by the result of @xcite the one - point distribution should become a product of two @xmath79 distributions , where @xmath79 is the goe tracy - widom distribution discovered first in random matrix theory @xcite .more precisely , extrapolating the result of corollary 2.5 in @xcite ( reported below for convenience ) we conjecture that @xmath80 as the kernel @xmath64 is not converging pointwise as @xmath78 ( not even after an appropriate conjugation ) we could not verify the conjecture analytically .the numerical studies presented in section [ sectnumerics ] below are in agreement with the conjecture and , moreover , show that the convergence as @xmath7 increases is quite fast .let @xmath81 for @xmath31 .for @xmath2 let @xmath82 and @xmath83 .then it holds @xmath84 with @xmath85 .note that @xmath86 as @xmath87 .the limit process @xmath70 occurs in a related last passage percolation ( lpp ) model as well .to each site @xmath88 of @xmath89 we assign an independent random variable @xmath90 with @xmath91 further , to a tasep initial condition @xmath92 we assign the line @xmath93 for a point @xmath94 on the right / above the line @xmath95 , the last passage time from @xmath96 to @xmath94 is defined by @xmath97 where the maximum is taken over all up - right paths is a sequence of points of @xmath98 such that , for @xmath99 . ] from @xmath95 to @xmath94 .the well - known connection between tasep and lpp is @xmath100 in our model , we have @xmath101 .consider the critical scaling @xmath102 and focus at the position @xmath103 define the rescaled lpp time by @xmath104}{2(2\ell)^{1/3}}.\ ] ] [ thmmainlpp ] it holds @xmath105 in the sense of finite dimensional distributions . in short , to prove theorem [ thmmainlpp ] , one starts with the relation ( [ eqlpptasep ] ) that gives the joint distributions of @xmath106 in terms of positions of tasep particles at different times , varying around @xmath107 on a @xmath108 scale only . by the slow - decorrelation phenomenon@xcite the fluctuations at different times are asymptotically the same as the fixed time fluctuations for points lying on special space - time directions ( the characteristics ) . at fixed time , the result is exactly given by theorem [ thmmain ] .the details of this procedure have been worked out for instance in @xcite . herewe numerically compute the distribution function of the process @xmath70 , given by a fredholm determiant of the kernel @xmath64 , as well as some of its basic statistics : expectation , variance , skewness , kurtosis .for the computation , we use the formula for @xmath64 given in the appendix [ sectappendix ] , since airy functions are already implemented functions in matlab , and apply bornemann s method for the evaluation of the fredholm determinants , see @xcite , which is well - adapted for analytic kernels .bornemann s algorithm also comes with an error control that we used , see section 4.4 of @xcite .( the left one ) and ( the right one ) , @xmath109 $ ] . the solid lines are the functions @xmath110 for . for @xmath66 , @xmath111 ,when @xmath7 grows larger , @xmath112 approximates the macroscopic shock distribution @xmath113 as conjectured , see .,height=226 ] for simplicity , we study the validity of the conjecture in , i.e. , we set @xmath114 . in principle, one could look also general @xmath115 , but then it has to be taken as a function of @xmath7 too ( since in the unscaled process the correlation scale of the process changes from @xmath52 to @xmath18 as @xmath1 varies from @xmath0 to a value strictly less then @xmath0 ) . to avoid to carry around a lot of @xmath116 factors, we rescale space by a factor @xmath116 so that the conjecture writes @xmath117 we denote @xmath118 .remark that the special case @xmath66 we have the airy@xmath67 kernel , @xmath119 . by ( [ eq2.9 ] )we have @xmath120 by theorem [ thmmain ] , this is the @xmath4 limit of the rescaled position of a particle in the microscopic shock .as mentioned earlier , we let @xmath7 grow large so as to recover the macroscopic shock distribution , which for large but finite time @xmath15 would correspond to the choice @xmath121 . due to the numerical limitations discussed below, we will compute @xmath122 and its basic statistics up to @xmath123 .surprisingly , already for this relatively small value of @xmath7 , one is already quite close to the asymptotic behavior .the reason for this is the following . at first approximation , from the kpz scalings , we know that the randomness that influences the statistical properties of particle positions around the shock lives in a @xmath52 neighborhood of the characteristic lines that comes together at the shock ( for a proof in a special case , see @xcite ) and the neighborhood should be quite tight to provide the super - exponential decay of the covariance for the airy@xmath67 process @xcite .further , by a closer inspection near the end - points , we discover that at @xmath18 distance from the shock , the neighborhood is only of order @xmath18 as well @xcite .these two phenomena imply that the convergence will happen on @xmath7 of order @xmath0 .the limitation to @xmath124 is due to the numerical difficulty of evaluating @xmath122 for @xmath7 large .as @xmath7 grows , @xmath125 has some terms which are of order @xmath0 and one term which is ( super- ) exponentially diverging .more precisely , one has @xmath126 where @xmath127 , and , for @xmath128 , @xmath129 and @xmath130 .this implies that when @xmath7 increases , the ratio between the bounded terms and the large term becomes smaller than @xmath131 machine precision and no reliable numerical evaluation is possible . in our case , already for @xmath132 , @xmath125 ( much less @xmath122 ) can no longer be computed in matlab .for instance , matlab computes @xmath133 for all tested @xmath134 , @xmath135 with an error @xmath136 , whereas @xmath137 , with an error @xmath138 .we present the numerical computations until @xmath123 , since for higher values the error term in the kurtosis becomes visible .generally , the computational error of @xmath112 decreases as @xmath134 increases since then @xmath139 needs to be computed only for small entries and the evaluation of @xmath140 is easier ( namely , the matrix whose determinant approximates @xmath122 gets closer to the identity matrix as @xmath134 increases , see ( 4.3 ) in @xcite ) .the statistics of @xmath122 were computed using the ` chebfun ` package ( see @xcite ) , in which @xmath122 is represented by its polynomial interpolant in chebyshev points , for our choice in @xmath141 points . in figure [ figdistr ] we plot @xmath142 , @xmath112 for @xmath143 and the conjectured @xmath78 limit , namely @xmath144 .a property which is apparent from figure [ figdistr ] is that @xmath140 monotonically decreases towards @xmath113 as @xmath7 grows .indeed , for all @xmath145 , and @xmath146 we have @xmath147 an analytic proof of this property does not seem to be trivial and is not available so far . to further quantify the difference between @xmath148 and @xmath149 we computed @xmath150 defined by . on this limited interval width for @xmath7 the convergence of the difference of the distribution functions close to exponentially fast.,height=188 ] andare compatible with the conjecture , but to have a further more reliable verification we study numerically the basis statistics too .the reason is that the distribution functions might be optically close but still be different .for example , the plots of the gue and goe tracy - widom distribution functions scaled to have both average @xmath16 and variance @xmath0 , are almost indistinguishable. however , by looking at their skewness and kurtosis one can clearly differentiate between them . in figure [ moments] we plot the basic statistics of @xmath122 and compare them with those of @xmath151 .the approximation is fastest for the expectation , and slowest for the kurtosis ( though the observation window for @xmath7 is too small to quantify the different rates of convergence ) . and of the conjectured limiting distribution , @xmath152.,height=200 ] finally ,let us resume in table [ tab ] the basic statistics of @xmath122 in comparison to @xmath149 ..data of the basic statistics and their relative difference to the conjectured limit distribution @xmath153 for a few values of @xmath7 . [ cols="<,<,<,<,<,<",options="header " , ]in section [ sectfinitekernel ] we derive the finite time kernel , whose fredholm determinant gives us the joint distributions of tasep particle positions , see proposition [ fdetform ] .for the derivation we first need to consider the case of a finite number @xmath154 of @xmath1-particles and then take the @xmath155 limit . in section [ sectscalinglimit ]we then perform the asymptotic analysis and complete the proof of theorem [ thmmain ] . taking the limit of the situation with finitely many slow particles we obtain the following result .[ fdetform ] consider two - speed tasep as defined in .then the joint distribution of the positions of @xmath156 normal particles with labels at time @xmath15 is given by @xmath157 with @xmath158 , where @xmath159}(x)$ ] and , the notation @xmath160 means a simple path anticlockwise oriented enclosing only poles of the integrand belonging to the set @xmath161 . ]@xmath162 @xmath163 and @xmath164 the system with finitely many slow particles has been already partially studied in @xcite .there , it was shown that the distribution function of particles positions is given by a fredholm determinant and the kernel was given .for a fixed @xmath165 , consider tasep with initial conditions and jump rates given by @xmath166 to distinguish this system with the one we want to study , i.e. , @xmath167 , we will index all quantities by a @xmath154 .proposition 6 of @xcite tells us that @xmath168 where the kernel kernel @xmath169 has the decomposition @xmath170 here @xmath171 and @xmath172 are as in proposition [ fdetform ] , while @xmath173 is given by @xmath174 first we note that @xmath175 this follows since @xmath176 for all @xmath177 and by the fact that in tasep the positions of the normal particles up to a fixed time @xmath15 depend only on finitely many other particles on the right with probability one , as is seen from a graphical construction of it .so it remains to show that the convergence in holds also on the level of fredholm determinants .first of all , as shown already in corollary 8 of @xcite ) , it holds @xmath178 pointwise .the reason being that for any @xmath179 the pole at @xmath180 in vanishes , in the limit of large @xmath154 we can integrate out explicitly the simple pole at @xmath181 of the kernel @xmath173 and it results in the kernel @xmath182 given in proposition [ fdetform ] . to show the convergence of fredholm determinants we use their series expansion expression , namely @xmath183_{1\leq k ,l\leq n}.\end{gathered}\ ] ] it is easy to see that @xmath184 for @xmath185 since the pole at @xmath180 vanishes after computing the residue at @xmath186 the pole at @xmath187 vanishes . similarly , @xmath188 for @xmath189 since the pole at @xmath180 vanishes .further , in the term @xmath190 , if @xmath191 is bounded from below , then for @xmath192 small enough this term is also zero .this implies that the @xmath193 determinant in is strictly equal to zero if @xmath194 . the physical reason for thisis that if we consider the system with particle numbers bounded from above by @xmath195 , then by tasep dynamics particles can be present only in the region on the right of @xmath196 .consequently , the sums are finite and the by hadamard bound @xmath197_{1\leq k , l\leq n}|\leq c^n n^{n/2}$ ] for some finite constant @xmath198 .thus by dominated convergence we can take the limit @xmath155 inside the sum and the proof is completed . with the finite time formula of proposition [ fdetform ] at hand, we can now proceed to prove the main result .the proof is identical to the one of theorem 2.5 in @xcite , given that the we have convergence of the ( properly rescaled ) kernel in a bounded set ( proposition [ proppointwiseconv ] ) , and good enough bounds to control the convergence of the fredholm determinant by the use of dominated convergence .the strategy , nowadays standard , was first used by tracy , widom and gravner in @xcite .the bounds are contained in propositions [ propboundphi ] , [ propmoddev ] , and [ proplargedev ] below .from now on , the @xmath199 are some fixed real values .we first prove convergence to the limit kernel @xmath64 and then provide integrable bounds .we consider the scaling @xmath200 the @xmath201 measure the fluctuations in the @xmath202 scale with respect to the macroscopic approximation given in .accordingly , we define the rescaled kernel @xmath203 and similarly for each component of the kernel .[ proppointwiseconv ] for any fixed @xmath204 , we have @xmath205 uniformly for @xmath206 in@xmath207 $ ] .here @xmath64 is the kernel from theorem [ thmmain ] .we start with @xmath171 .the residue at @xmath16 can be easily computed expanding @xmath208 with the binomial formula and one readily obtains that @xmath209 .it is then an easy computation to show that ( see e.g. proposition 7 of @xcite ) @xmath210 next we consider @xmath211 .we make the change of variables @xmath212 , rename @xmath213 , and set @xmath214 and @xmath215 .then , @xmath216 equals the kernel @xmath217 in ( 3.7 ) of @xcite .the convergence of @xmath217 to the @xmath218 transition kernel is proven in proposition 4 of @xcite , giving the first double integral of @xmath64 , i.e. , + in the integral representation of appendix [ sectappendix ] .finally consider @xmath182 .we have @xmath219 with @xmath220 the poles and order of integration are different , but the exponential part equals again the exponential part of @xmath217 in ( 3.7 ) of @xcite , so let us focus on the differences .the critical point of @xmath221 is @xmath222 , and in proposition 4 of @xcite @xmath223 is divided in four regions @xmath224 depending on the sign of @xmath225 and of @xmath226 , see figure [ figzeros ] .[ c]@xmath227 [ c]@xmath228 [ c]@xmath16 [ c]@xmath229 [ c]@xmath230 [ c]@xmath231 [ c]@xmath232 [ c]@xmath233 [ c]@xmath234 [ c]@xmath235 [ c]@xmath236 is positive in @xmath232 and @xmath234 and negative in @xmath231 and @xmath233.,title="fig:",height=226 ] for @xmath237 we may choose any simple anticlockwise oriented closed path passing through @xmath222 and staying in @xmath233 .@xmath238 is restricted to stay in @xmath232 except for a local modification in a of the critical point in order to satisfy .more precisely , @xmath238 passes through @xmath239 for some @xmath240 , see figure [ figpaths1 ] .[ l]@xmath241 [ l]@xmath242 [ c]@xmath243 [ l]@xmath238 [ l]@xmath237 and @xmath237 used for the pointwise convergence .the point in the middle is @xmath244 .the vertical piece in @xmath238 is of length of order @xmath245.,title="fig:",height=188 ] we will take @xmath237 to arrive in @xmath222 with an angle @xmath246 .define for @xmath247 the segments @xmath248 and @xmath249 .denote by @xmath250 the part of the contours where @xmath251 and/or @xmath252 .then the integral is on @xmath253 on @xmath250 there exists a @xmath254 that @xmath255 and/or @xmath256 . further @xmath257 . hence the contribution coming from @xmath258 and @xmath259is bounded by @xmath260 .furthermore , on @xmath250 , @xmath261 with @xmath262 depending only on @xmath263 .hence we may bound the overall contribution of @xmath250 by @xmath264 for some finite constant @xmath265 . as we will show below, the contribution coming from @xmath266 and @xmath267 is of order one , therefore the contribution of the integrals over @xmath250 is negligible in the @xmath4 limit . next consider the contribution from the integral over @xmath268 .consider the change of variables @xmath269 and denote @xmath270 .then by taylor expansion we obtain @xmath271 the control of the error term in is ( almost ) identical to the one given in proposition 4 of @xcite , we therefore omit it .the error term is of order @xmath272 . for the remaining part ,denote @xmath273 and @xmath274 . any extension of finite length @xmath275 gives an error of order . for @xmath276 large , @xmath277 ( resp .@xmath278 ) decays linearly along @xmath279 ( resp .@xmath280 ) . therefore also extending the curves to infinity creates an error of order @xmath281 for some @xmath282 .we denote the resulting curves by @xmath283 and we are thus left with @xmath284 the integration paths can be deformed as in definition [ deflimitprocess ] without errors ( the minus factors comes from the change of orientation of one of the paths ) . for @xmath171, an integrable bound was already obtained in @xcite ( with @xmath171 as binomial coefficient , see the beginning of the proof of proposition [ proppointwiseconv ] ) .[ propboundphi ] for any @xmath206 in @xmath285 and @xmath286 fixed , there exist a finite constants @xmath198 and @xmath287 , such that for all @xmath288 , @xmath289 [ propmoddev ] for any @xmath290 large enough , there are @xmath291 , @xmath287 such that for all @xmath292 , @xmath288 , there exists a finite constant @xmath198 such that @xmath293 for all @xmath294\setminus[-l , l]$ ] .for @xmath216 the statement is proposition 5 in @xcite . for @xmath295 , we follow a similar strategy , but let us give the details .define @xmath296 $ ] and denote the integrand by @xmath297 .let @xmath298 be an interval on which @xmath299 are parametrized .the analysis of proposition [ proppointwiseconv ] shows that for a constant @xmath198 @xmath300 if @xmath301 , we have an additional factor @xmath302 in the integrand of . as we shall show in ( b ) , ( c ) , ( e ) , ( f ) below , if we are not close to @xmath222 , then @xmath303 and thus get the bound @xmath304 .close to @xmath222 , we do a modification of one of the contours , depending on whether @xmath305 or @xmath306 , and then get the needed decay for . in the @xmath307 case , we modify @xmath238 near the critical point @xmath222 and show that in the unmodified region the decay is the same as in the case @xmath308 case times an integrable factor .we then deal with the modified region and provide the needed decay there too .if @xmath309 , we integrate out the residue at @xmath310 and show the needed decay for it by modifying @xmath238 . in the remaining integralwe may then deform the contour @xmath237 to get the desired decay .* case * @xmath306 .the paths @xmath237 and @xmath238 are as in figure [ figpaths1 ] except that the distance of the vertical piece of @xmath238 with respect to @xmath222 is @xmath311 instead of @xmath243 .near @xmath222 we modify @xmath238 by a vertical part @xmath312 that passes through @xmath313 with @xmath314 ( see ) and which is symmetric w.r.t . the real line . as in proposition[ proppointwiseconv ] let @xmath315 be the angle with which @xmath237 leaves @xmath222 and let @xmath316 .the region @xmath231 in figure [ figzeros ] leaves @xmath222 with angle @xmath317 .consequently , for @xmath312 to end outside @xmath231 and satisfy @xmath318 we can choose ( for @xmath15 large enough ) its length as @xmath319 for some @xmath320 .hence we define @xmath321\}.\ ] ] \(a ) the choice of contours is such that @xmath322 for some constant @xmath323 .this is at least the same order as for the contours in proposition [ proppointwiseconv ] where we had @xmath324 hence ( as in the case ) the @xmath325 term does not create problems .\(b ) the contour @xmath237 can be chosen such that @xmath326 reaches its minimum at @xmath327 so we can simply bound @xmath328 \(c ) let @xmath329 . in the following ,we set which is just a shift in the variable @xmath330 .@xmath238 can be chosen such that on @xmath331 the maximum of @xmath332 is reached at @xmath333 . for @xmath334 small enough @xmath335therefore it holds @xmath336 for @xmath337 small enough .\(d ) for the integral on @xmath338 , it is an integral on @xmath339 $ ] in the variable @xmath340 . since @xmath341 , this term multiplied by the @xmath18 prefactor gives a term @xmath342 .so it suffices to have a bound on the integrand that controls it .on @xmath312 we use taylor expansion around @xmath222 ( from which @xmath312 is at most @xmath343 far away ) .the @xmath344dependant part of the exponential term becomes @xmath345 now we take real parts in the exponent .we see that for @xmath290 large and @xmath334 small enough we have @xmath346 and @xmath347 .we get the upper bound @xmath348 the @xmath349 cancels exactly with the contribution coming from the integrand in the @xmath350 variable .finally note that for @xmath290 large enough @xmath351 * case * @xmath305 . herewe integrate out the residue @xmath352 and obtain @xmath353 where @xmath354 the contours @xmath355 and @xmath238 in the double integral in satisfy and @xmath238 passes through the critical point @xmath222 . to provide the integrable bound for the double integral in, one does the same analysis as in the @xmath306 case , except that the roles of @xmath238 and @xmath355 , and @xmath356 and @xmath357 are reversed : we modify @xmath355 by a vertical part with distance @xmath358 to @xmath222 and then go through the steps ( a ) to ( d ) .we have for @xmath238 as in figure [ figpaths1 ] , @xmath308 and @xmath15 large enough the bound @xmath359 the bound follows from the identity in and the fact that respective bounds hold for @xmath295 and the double integral in . for @xmath360we modify @xmath238 near @xmath222 by a vertical piece @xmath361\}\ ] ] where @xmath362 .compared to @xmath308 , the integrand has the additional factor @xmath363 ( e ) we can choose the contour @xmath238 such that @xmath364 for all @xmath365 . in particular for @xmath329 we may simply bound @xmath366 ( f ) furthermore , @xmath238 may be chosen such that for @xmath329 the minimum of @xmath367 is reached at @xmath333 . for this @xmath115we have @xmath368 so that we get the bound for @xmath290 large and @xmath334 small @xmath369 now we deal with @xmath312 . we write @xmath370 with @xmath371 and @xmath372 .next we do taylor around @xmath222 in @xmath221 and we first obtain @xmath373 we compute @xmath374 for @xmath290 large , the last term in dominates , thus reaches its maximum at @xmath375 .so we may bound @xmath376 a constant . as for @xmath377 , by taylor expansion around @xmath222we obtain @xmath378 thus , for @xmath290 large we can bound @xmath379 .finally , for @xmath380 we obtain @xmath381 with @xmath382 and @xmath383 some constants .therefore , for @xmath290 large and @xmath334 small enough we obtain latexmath:[\[\label{final3 } @xmath290 large and @xmath334 small enough .so , putting together ( e ) , ( f ) and , , , we obtain the desired bound for @xmath385 .[ proplargedev ] let @xmath386 .then , there is a finite constant @xmath198 such that for @xmath15 large enough we have @xmath387 for @xmath388 .the estimate for @xmath389 is contained in proposition 6 of @xcite . as in proposition[ propmoddev ] we denote @xmath390 and distinguish the cases @xmath306 and @xmath305 .* case * @xmath306 .we choose the same contours as in the moderate deviations regime for @xmath306 , here with @xmath391 ( the additional shift by @xmath392 in @xmath338 is however unnecessary for @xmath15 large enough ) .we write so that @xmath393 thus , compared to the @xmath391 case we have the additional factor @xmath394 it suffices to bound @xmath395 because the integrand for @xmath391 is ( uniformely for @xmath15 large enough ) bounded in @xmath396 .the choice of contours is such that @xmath326 reaches its minimum at @xmath327 and @xmath332 its maximum at @xmath397 .using further @xmath398 , we may bound @xmath399 for some constant @xmath400 .* case * @xmath305 .we again choose the same contours as in the moderate deviations regime for @xmath305 , with @xmath401 ( again the additional shift by @xmath392 is unnecessary for @xmath15 large enough ) .we again integrate out the residue at @xmath402 . in the doubleintegral , with respect to @xmath403 we get the same additional factor , which can now be bounded @xmath404 as for the residue , compared to @xmath403 we have the additional term @xmath405 where the inequality holds since @xmath406 and .here we give the explicit form of @xmath64 that we used for the numerical evaluation of @xmath122 and its statistics .

<S> we consider tasep with two types of particles starting at every second site . </S><S> particles to the left of the origin have jump rate @xmath0 , while particles to the right have jump rate @xmath1 . when @xmath2 there is a formation of a shock where the density jumps to @xmath3 . for @xmath2 </S><S> fixed , the statistics of the associated height functions around the shock is asymptotically ( as time @xmath4 ) a maximum of two independent random variables as shown in @xcite . in this paper </S><S> we consider the critical scaling when @xmath5 , where @xmath6 is the observation time . in that case </S><S> the decoupling does not occur anymore . </S><S> we determine the limiting distributions of the shock and numerically study its convergence as a function of @xmath7 . </S><S> we see that the convergence to @xmath8 occurs quite rapidly as @xmath7 increases . </S><S> the critical scaling is analogue to the one used in the last passage percolation to obtain the bbp transition processes @xcite . </S>

cond-mat0510551

acoustic imaging of buried objects in a non - linear medium like nominally dry soil is still an open problem @xcite . amongst its main applications, we can mention the problem of locating antipersonnel land mines , human remains for forensic investigations , hidden underground structures of archaeological importance etc .it has been shown that gentle mechanical impulses @xcite can be used to detect buried objects at depths of a meter or so in nominally dry sand beds .imaging requires the understanding of pulse propagation in nonlinear media . besides the underlying non - linearity due to the hertzian contact forces between the grains ( see , for example , @xcite ) ,gravitational loading is an added feature that affects the dynamics .before we can analyze the gravitational effects in 3d systems , it is reasonable first to look at the problem in 1d .further motivation for 1d studies can be found in @xcite .( _ a_)-(_c _ ) , @xmath1 and we see solitary wave propagation . in fig .[ prop](_d_)-(_f _ ) , @xmath2 and in fig .[ prop](_g_)-(_i _ ) , @xmath3 . in both cases ,open boundary conditions are employed at the top with closed boundary conditions at the bottom .we see solitary wave like front with a long tail for non - zero @xmath0 .the text in the lower left hand corner indicates the times at which the snapshots are taken and , in brackets , is given the roundtrip time , i.e. the time taken by the pulse to travel up and down the chain once .the @xmath4-axis ranges have been marked in the leftmost panel for each row of panels while the @xmath5-axis range is the same in all and labelled in the top left panel , fig .[ prop](_a _ ) , where @xmath6 is the position of the particle in the chain and @xmath7 its velocity . ]impulse propagation in a gravitationally loaded chain with @xmath8 grains and a perfectly reflecting boundary at the bottom has been studied @xcite .when @xmath1 , one finds solitary wave propagation @xcite as shown via velocity vs. position plots made at different times in figs .[ prop](_a_)-(_c _ ) ( see later for simulation details ) . in all the panels of fig .[ prop ] , the numbers in the bottom left corner indicate the times at which the snapshots have been taken and , in brackets , the round trip time of the pulse is given . the velocity of the solitary wave is a function of its amplitude @xcite .each solitary wave is reflected at the bottom boundary . during reflection , the solitary wave forms secondary solitary waves @xcite .further , secondary solitary waves are generated from collisions with other solitary waves @xcite .the formation of secondary solitary waves progressively reduces the amplitude of the solitary waves in the system . in fig .[ prop](_c _ ) , we find that the amplitude of the primary solitary wave has diminished considerably after @xmath9 roundtrips . when @xmath10 , one finds a solitary wave like front with a tail that elongates with time @xcite .in figs .[ prop](_d_)-(_f _ ) , we have shown snapshots from the propagation of the impulse in a system with @xmath2 and with open boundary condition at the top .we have the same boundary conditions in figs .[ prop](_g_)-(_i _ ) with @xmath3 .we see that the impulse propagates faster with increased @xmath0 .the scheme of this paper is as follows .section ii surveys briefly the @xmath11-@xmath12 model with the simulation details .our focus is to probe impulse propagation and backscattering by simply considering the dynamics of the surface grain . in section iii ,we construct a version of the toy model and recover the results obtained in section ii .section iv summarizes the conclusions for the @xmath11d study and assesses the usefulness of this study for further @xmath13d analysis .we model the granular chain as a collection of @xmath14 spherical grains which are placed in contact with one another and loaded by a gravitational field .the interaction between every pair of spheres of radius @xmath15 is driven by hertz law @xcite , _i.e. _ , it is assumed that spheres labeled as @xmath16 and @xmath17 are interacting with a potential proportional to the overlaps , @xmath18 , in the contact region , @xmath19 where @xmath20 is a constant that depends upon the material characteristics of the grains , @xmath21 is the separation distance between the centers of the grains @xmath22 and @xmath17 and @xmath23 is the grain overlap .the equation of motion for grain @xmath16 can be written as @xcite,@xmath24 where @xmath25 and @xmath0 are the mass of the grain and the value of gravity , respectively .the system dynamics is obtained by time integration of the coupled newtonian equations of motion via the velocity - verlet algorithm @xcite .we ignore the role of dissipation in the present study ; dissipation effects can be built in later @xcite .we set @xmath26 and @xmath25 to @xmath11 , and @xmath20 is set equal to @xmath27 .we find that an integration time step of @xmath28 provides excellent enrgy conservation ; decreasing the time step further does not improve the accuracy of our results .particles are labeled starting from the bottom so that the @xmath8@xmath29 particle is at the top . while the bottom boundary condition has been kept perfectly reflecting in all the simulations , both open and closed boundary conditionshave been employed at the top in different simulations to investigate the differences induced by the particular choice .( _ a _ ) , @xmath1 , in fig .[ bcksc](_b _ ) @xmath2 , and in figs .[ bcksc](_c_)-(_d _ ) , @xmath3 . figs .[ bcksc](_a_)-(_c _ ) have closed boundary conditions at the top while fig .[ bcksc](_d _ ) has open boundary condition at the top .typical integrated backscattered energy , @xmath30 , plots are shown in figs .[ bcksc](_e_)-(_f _ ) where fig .[ bcksc](_e _ ) is for @xmath31 and fig .[ bcksc](_f _ ) for @xmath2 with top closed boundary condition ( cf . fig .[ bcksc](_b _ ) ) .the center of mass , @xmath32 , oscillations typically found in these systems are given in the third row of panels .[ bcksc](_g _ ) is for @xmath31 while fig .[ bcksc](_h _ ) is for the top closed boundary condition ( @xmath2 ; cf . fig .[ bcksc](_b _ ) and fig .[ bcksc](_f _ ) ) .[ bcksc](_i _ ) is for the top open boundary condition ( cf . fig .[ bcksc](_d _ ) ) .the @xmath5-axis remains the same in all panels as labelled in fig .[ bcksc](_a _ ) with the time , @xmath33 , given in arbitrary units ( a.u . ) of the simulation .@xmath4-axis ranges for the first row is given in the leftmost panel , and separately labelled for each panel in the bottom row .for the middle row , the first panel on the left has a different @xmath4-axis range as shown and , for the last two panels , it is as shown for the middle panel . ]the dynamics is initiated via an initial velocity perturbation at the top of the chain , @xmath34 , with @xmath35 for @xmath36 ; we have employed @xmath37 in the studies reported here .we subsequently monitor the backscattered energy received at the surface , @xmath38 .the time integrated backscattered energy , @xmath30 , has also been computed by summing the entire sequence of energy packets received at the surface , i.e. @xmath39 . in fig .[ bcksc](_a_-_d _ ) , we show plots of @xmath40 vs. @xmath33 . fig .[ bcksc](_a _ ) is obtained with @xmath31 .we have two prominent peaks at early times followed by a continuous distribution of much smaller peaks .clearly , the smaller peaks originate in the breakdown of the solitary waves at the boundaries and through mutual interactions ( this issue is discussed in detail in @xcite ) .the two large peaks at early times are separated by roundtrip times of the system . fig .[ bcksc](_b _ ) corresponds to @xmath2 , with reflecting boundary conditions at the top in which the surface grain is only allowed to move in one direction into the chain . in this case, there are large peaks separated by roundtrip times of the system , but there is much activity in between them indicating the presence of acoustic like rattling throughout the system . increasing @xmath0 to @xmath41 ( fig .[ bcksc](_c _ ) ) , with the same boundary conditions , does not reveal any noticeable change in the pattern except for increasing the density of peaks , caused by the smaller roundtrip times . if we keep @xmath42 and change to open boundary conditions , meaning that the surface grain is allowed to move up and down , we get the pattern shown in fig .[ bcksc](_d _ ) where we notice that the number of larger peaks are reduced .we attribute this to increased energy sharing caused by the open boundary at the top .the @xmath30 vs. @xmath33 plots show a distinct difference between the ballistic and acoustic type propagation cases . while fig .[ bcksc](_e _ ) is for the @xmath1 case , fig .[ bcksc](_f _ ) is typical of @xmath10 cases ( here , @xmath2 with closed boundary at the top ) .the increase is linear for non - zero @xmath0 whereas it only approaches linear for @xmath31 in the later stages .the steps in @xmath30 indicate the arrival of peaks with the quiescent period in between marked by plateaus ; the length of the plateau indicates the time period between peaks .the pattern of arrival of large peaks introduce modulations in the amplitudes of @xmath40 . nevertheless , there are no simple patterns in the distribution of peaks of @xmath40 vs. @xmath33 in figs .[ bcksc](_a_-_d _ ) .we note that the boundary conditions affect @xmath40 .a possible modulating mechanism could be due to the center of mass ( com ) oscillations of the system .we show these oscillations ( the breathing of the chain ) in the third row of fig .[ bcksc ] for three typical cases . with @xmath1 ( fig .[ bcksc](_g _ ) ) , the significant com oscillations appear only after the solitary waves have broken down . for non - zero @xmath0 ,the com oscillations are affected by the surface boundary .[ bcksc](_h _ ) is for @xmath2 and with closed boundary conditions at the top while fig .[ bcksc](_i _ ) is typical of open boundary conditions at the top ( in this case , @xmath3 ) .we now turn to the toy model to see whether these patterns in @xmath40 can be reproduced .we had introduced a toy model in an earlier paper where the energy propagation in a vertical alignment of masses was considered in a simplified manner @xcite . at time @xmath43 , we set initial energy @xmath44 for layer one and zero for the rest . at @xmath45 , the first layer in the vertical chain transfers @xmath46 ( @xmath47 ) of the impulse energy to the second layer , and retains @xmath48 .at subsequent times , the impulse will propagate in the same fashion all the way down the chain .after each transfer of energy , the phase of the mass reverses so that it will interact with its adjacent layer in the opposite direction in the following time step .the interaction between two adjacent layers occurs in the following two ways : ( i ) _ equipartition _ case , and ( ii ) _ exchange _ case . in the _ equipartition _case , the two interacting layers will come away from the interaction with equal amounts of energy ; we add up the individual energies of the two layers and divide the sum equally between them . in the _ exchange _ case , we let the layers exchange their energies ; the two interacting layers , after the interaction , come away with the energy of the other .( _ a _ ) is for the case of a single equipartition center in the toy model located at particle no .200 out of a total of 500 particles in the chain , and fig .[ fig3](_b _ ) for the case , similarly , with equipartition center at particle no .350 ; the exchange coefficient @xmath49 .data for these two cases have been averaged , along with data for cases with single equipartition centers located at 300 and 390 respectively , and presented in fig .[ fig3](_c _ ) . in fig .[ fig3](_d _ ) , we have multiple equipartition centers located at particle nos .125 , 230 and 390 with @xmath49 . ] in fig .[ fig3 ] , we show the results where the model has , for the first time , both equipartition and exchange .our goal is to develop the propagation patterns seen in figs .[ bcksc](_b_-_d _ ) .the model is based on the exchange case , but we allow equipartition centers to develop , as time goes by , at different points in the chain . in fig .[ fig3](_a _ ) , we have shown a case where a single equipartition center is kept fixed at layer number 200 in a chain of 500 layers . fig .[ fig3](_b _ ) is for a similar case but with the equipartition center located at layer number 350 .we see that the patterns in @xmath40 do change slightly depending on the location of the isolated equipartition center , but , in both cases , it is seen that the patterns are broadly similar to the non - zero @xmath0 cases seen in figs .[ bcksc](_b_)-(_d _ ) . this has prompted us to try averaging over the patterns resulting from differing locations of the isolated equipartition centers and , indeed , the plot of fig .[ fig3](_c _ ) shows that such averaging does retain the similarity with plots of figs .[ bcksc](_b_)-(_d _ ) . to obtain this figure , we averaged over , along with the cases given in fig .[ fig3](_a _ ) and fig .[ fig3](_b _ ) , two other cases with isolated equipartition centers , in each case , located at layer numbers 300 and 390 respectively .if we vary the exchange coefficient @xmath46 , we get similar patterns ( not shown here ) but with the difference that the growth in @xmath30 is proportional to the @xmath46 value .this indicates that @xmath46 could also be used , along with the equipartition centers , to capture the transition from ballistic propagation to acoustic propagation . in fig .[ fig3](_d _ ) , we show results with multiple ( three , in this cases ) equipartition centers in the system , at layer numbers 125 , 230 and 390 ; @xmath49 for this case .we see that @xmath30 grows by as much as in figs .[ fig3](_a_)-(_c _ ) but the attenuation in peaks of @xmath40 is very pronounced and similar to the pattern seen in fig .[ bcksc](_d _ ) .in this paper , we have studied the problem of impulse propagation in a gravitationally loaded granular chain .we have focused our attention on the backscattered energy received at the surface after an impulse has been initiated into the system .our studies have been carried out using two different approaches ( i ) using newtonian dynamics in a non - disspative system to describe backscattering at the surface of the system and , ( ii ) using the toy model @xcite to recover the behavior in ( i ) .our results show that the toy model is capable of reproducing the correct form of the backscattered energy as a function of time in a gravitationally loaded chain .gravitational loading introduces acoustic - like oscillations in the system .such oscillations vanish and the system ends up propagating solitary waves when gravitational loading is zero . to achieve these descriptions , one must generalize the earlier version of the toy model @xcite and incorporate both exchange ( represents non - linear ) and equipartition ( represents acoustic ) effects .we find that the role of the equipartition ( acoustic ) effect in the toy model is rather dominant and only limited equipartitioning of energy gives the correct backscattering behavior .our studies confirm that mechanical propagation in a granular chain is strongly nonlinear even in the presence of gravitational loading .ea acknowledges the fulbright foundation for support .trkm and ss have been supported by the army research office and nasa .99 g. baker , c. schmeissner , d. w. steeples and r. g. plumb , geophys . res* 26 * , 279 ( 1999 ) ; s. hostler , ph.d .thesis , mechanical engineering , california institute of technology ( 2005 ) ; s. r. hostler and c. e. brennen , phys .e * 72 * , 031303 ( 2005 ) ; _ ibid ._ * 72 * , 031304 ( 2005 ) ; d. p. visco , jr . , s. swaminathan , t. r. krishna mohan , a. sokolow and s. sen , phys .e * 70 * , 051306 ( 2004 ) .s. sen , t. r. krishna mohan , d. visco , jr ., s. swaminathan , a. sokolow , e. avalos and m. nakagawa , int . j. mod* 19 * , 2951 ( 2005 ) .v. nesterenko , j. appl .phys . * 5 * , 733 ( 1983 ) ; a. lazaridi and v. nesterenko , _ ibid . _* 26 * , 405 ( 1985 ) ; c. coste , e. falcon and s. fauve , phys . rev .e * 56 * , 6104 ( 1997 ) ; a. chatterjee , phys .e * 59 * , 5912 ( 1999 ) . o. r. walton and r. l. braun , j. rheol . *30 * , 949 ( 1986 ) ; n. v. brilliantov , f. spahn , j. m. hertzsch and t. pschel , phys .e * 53 * , 5382 ( 1996 ) ; s. sen , m. manciu and a. j. hurd , physica d * 157 * , 226 ( 2001 ) .

<S> we recently introduced a simple toy model to describe energy propagation and backscattering in complex layered media ( t.r . </S><S> krishna mohan and s. sen , phys . </S><S> rev . </S><S> e * 67 * , 060301(r ) ( 2003 ) ) . </S><S> the model provides good qualitative description of energy propagation and backscattering in real soils . here </S><S> we present a dynamical study of energy propagation and backscattering in a gravitationally loaded granular chain and compare our results with those obtained using the toy model . </S><S> the propagation is ballistic for low @xmath0 values and acquires characteristics of acoustic propagation as @xmath0 is increased . </S><S> we focus on the dynamics of the surface grain and examine the backscattered energy at the surface . </S><S> as we shall see , excellent agreement between the two models is achieved when we consider the simultaneous presence of acoustic and nonlinear behavior in the toy model . </S><S> our study serves as a first step towards using the toy model to describe impulse propagation in gravitationally loaded soils . </S>

0708.3382

obtaining reliable knowledge about the structure and distance of nearby ob associations plays a critical role in the overall study of the milky way morphology near the sun . unlike the external galaxies where the star - forming fields are generally evident from direct imaging , the study of the spiral structure of our own galaxyis largely grounded in distance determinations of young stellar tracers . at present , sufficiently accurate astrometric data ( parallaxes and proper motions ) are available for few star - forming regions within @xmath15 pc .more comprehensive and representative studies of the local history and dynamics of star formation have to rely on the photometric method of distance determination and stellar evolution theory . the completion of the @xmath0 catalog ( esa 1997 ) offered a possibility for a major improvement of the membership of young moving groups near the sun and refining the distance scale to nearby open clusters and ob associations. however , the mean @xmath0 parallaxes for some galactic clusters are in disagreement with ground - based determinations by various methods .statistically significant discrepancies between the @xmath0 trigonometric and traditional photometric , spectroscopic and interferometric results have been reported in the literature for selected small - scale fields , most notably for the pleiades open cluster ( pinsonneault et al .1998 , soderblom et al .1998 , narayanan & gould 1999 , stello & nissen 2001 , makarov 2002 , pan et al .2004 , percival et al .2005 , soderblom et al . 2005 ) .platais et al .( 2007 ) found a similar offset in the @xmath0 mean parallax for the young open cluster ic 2391 .a discrepancy was reported by kaltcheva et al .( 2005 ) for the open cluster ic 2602 as well .the cause for these inconsistencies is most likely due to a faulty data reduction algorithm used in @xmath0 , which allowed highly correlated errors of along - scan attitude parameters to propagate into the fitted astrometric parameters .an alternative data reduction approach has been suggested and successfully tested by makarov ( 2002 , 2003 ) .the region of cr 121 is another example of this discrepancy . since the discovery of a compact group at @xmath16= @xmath17 by collinder ( 1931 ) , both the cluster and the larger @xmath18 x @xmath18 field have been extensively studied by ubv and @xmath4 photometry .this area includes one of the 12 ob associations within 1 kpc from the sun with fairly detailed kinematical information and membership determined from @xmath0 .the @xmath0 proper motions reveal a moving group of 103 stars between @xmath19 and @xmath20 , identifying the compact cluster cr 121 with an unbound extended ob association at a distance of 592@xmath2128 pc , similar to sco ob2 ( de zeeuw et al .robichon et al .( 1999 ) selected 13 @xmath0 members of cr 121 and found a mean @xmath0 parallax of [email protected] mas ( @xmath22 pc ) .in contrast to these results from @xmath0 , the latest @xmath4 photometric study ( kaltcheva 2000 ) concluded that a compact stellar group apparently identical to the genuine cluster ( cr 121 ) is situated at 1085@xmath2141 pc and the closest members of the loose association are found at an average distance of 660 - 730 pc , in agreement with most of the previous photometric investigations .since the @xmath4 photometry is arguably the best photometric system in use to provide accurate photometric distances , the origin of the discrepancy was suggested to be in the @xmath0 parallaxes for the cr 121 members .burningham et al .( 2003 ) studied the low - mass pre - main sequence stars toward cr 121 and also came to conclusions consistent with the photometric distance determinations . in this letter, we consider a sample of probable members of the extended association around cr 121 selected by de zeeuw et al .( 1999 ) for which accurate @xmath4 photometry is available . the astrometric parameters of these stars are recomputed from the @xmath0 intermediate astrometry data by the method proposed by makarov ( 2002 ) .the recomputed parallaxes allow us to resolve the controversy about the distance and dimensions of the ob association in this field .our sample contains all 44 early - type stars with @xmath0 parallaxes listed by de zeeuw et al .( 1999 ) as probable members of the cr 121 moving group for which @xmath4 photometry is available .homogeneous photometric @xmath4 distances are calculated for 43 of them ( kaltcheva 2000 ) .table 1 presents the sample , where the @xmath0 identification numbers are given in the first column , followed by the @xmath0 parallaxes and their errors , recalculated parallaxes and their errors , photometric @xmath4 distances and mk spectral classification .the stars are formally divided into field stars ( or possible association members ) , spread over a @xmath23 area around the center of cr 121 and 6 photometrically selected members of the dense cluster cr 121 ( kaltcheva 2000 and references therein ) . as follows from the data in table 1there is a statistically significant difference between the mean @xmath0 parallax of @xmath24 mas and the mean recomputed parallax of @xmath25 mas .the errors provided here are the formal standard deviation of the mean computed from the formal errors of parallaxes .1 shows the original @xmath0 parallaxes ( left plot ) and our recomputed parallaxes ( right plot ) versus the photometric parallaxes for the sample of 43 stars in table 1 .the @xmath0 parallaxes are on average larger than the photometric values by [email protected] mas , where the quoted error is the sample standard error of the mean .this is a statistically significant difference of the same order as those found for the pleiades and a few other galactic clusters . on the other hand, the agreement is excellent between the mean photometric parallax and the mean corrected parallax ( @xmath26 mas ) .this supports our main conclusion that the @xmath0 parallaxes are systematically overestimated in this area of the sky . but fig .1 also reveals another strange property of the original parallaxes .while the recomputed parallaxes are scattered fairly symmetrically around the line of unit slope in the right plot and their dispersion is in good agreement with the measurement errors , the original parallaxes are grouped tightly around the mean ( @xmath27 mas ) with a standard deviation of only @xmath28 mas .this value is much too small for the estimated formal errors ( mean @xmath29 mas , _ rms _ @xmath30 mas ) .we attribute this result to a strong selection effect in the method employed by de zeeuw et al .( 1999 ) , which preferentially accepted stars with large measured parallaxes , i.e. , mostly stars with positive errors `` observed minus true '' . in combination with the correlated error of the mean parallax, this selection bias gives rise to doubt about the completeness and reliability of the present membership list .the significant dispersion of both photometric and recomputed parallaxes also implies a complex morphology of this moving group having a considerable depth , as opposed to the previous conjecture of an association compressed in the radial dimension , similar to the nearby sco ob2 association , as concluded by de zeeuw et al .the group also appears to be located at a larger distance of @xmath31 pc , rather than at @xmath1 pc as follows from the mean @xmath0 parallax of the sample in table 1 .based on a larger photometric sample it has already been pointed out that the loose nearby structure defined by de zeeuw et al .( 1999 ) to be located at 592@xmath32 pc photometrically appears to be more distant by about 100 pc ( kaltcheva 2000 ) .the parallaxes recalculated here support the photometric findings .our result implies that the problem of inaccurate mean parallaxes in @xmath0 affects more regions , and of larger angular area , than just a few small patches occupied by dense open clusters .this is not an irreversible situation , because the method of astrometric solution of the available @xmath0 data used in this paper proves once again successful in correcting this error , despite its limitations .a more systematic and thorough comparison of @xmath0 data with distances from precision multi - band photometry will probably reveal more problematic areas .it is not clear at present how widely spread the parallax error is , and whether a global astrometric solution will have a significant impact on the present knowledge of distances and morphology for many of the ob associations represented in the catalog , but it is evident that the @xmath0-based census of some of the moving groups near the sun should be critically reconsidered .burningham , b. , naylor , t. , jeffries , r. d. , devey , c. r. , 2003 , mnras , 346 , 1143 collinder , p. , 1931lund , 2 , 1 de zeeuw , p.t ., hoogerwerf , r. , de bruijne , j.h.j . ,brown , a.g.a . , & blaauw , a. , 1999 , , 117 , 354 esa , 1997 , the hipparcos catalogue , esa sp-1200 , vol . 1 - 17 kaltcheva , n.t ., 2000 , mnras , 318 , 1023 kaltcheva , n.t . , jaeger , s. , kaba bah , m. , & briley , m.m .2005 , an , 326 , 738 makarov , v.v . , 2002 , aj , 124 , 3299 makarov , v.v . , 2003 , aj,126 , 2408 pan , x. , shao , m. , & kulkarni , s.r . 2004 , nature , 427 , 326 percival , s.m . , salaris , m. , & groenewegen , m.a.t .2005 , , 429 , 887 pinsonneault , m.h . , stauffer , j. , soderblom , d.r . , king , j.r . ,& hanson , r.b .1998 , , 504 , 170 platais , i. , melo , c. , mermilliod , j .- c ., kozhurina - platais , v. , fulbright , j.p . ,mendez , r.a . , altmann , m. , & sperauskas , j. , 2007 , , 461 , 509 robichon , n. , arenou , f. , mermilliod , j.c . , & turon , c. , 1999 , a&a , 345 , 471 soderblom , d.r . , king , j.r . , hanson , r. b. , jones , b.f ., fisher , d. , stauffer , j.r . , & pinsonneault , m. , 1998 , , 504 , 192 soderblom , d.r ., nelan , e. , benedict , g. f. , mcarthur , b. , ramirez , i. , spiesman , w. , & jones , b. f. , 2005 , , 129 , 1616 stello , d. , & nissen , p.e .2001 , , 374 , 105 .the sample : identifications from the @xmath0 catalog , followed by the @xmath0 parallaxes @xmath33 and their formal errors @xmath34 , the recalculated parallaxes @xmath35 and their formal errors @xmath36 , the photometric @xmath4 distances r and the mk classification .[ cols= " < , > , > , > , > , > , < " , ] we are grateful to the referee dr . r. hanson for a number of valuable comments .a uw oshkosh vander putten award is acknowledged .the research described in this paper was in part carried out at the jet propulsion laboratory , california institute of technology , under a contract with the national aeronautics and space administration .

<S> we present further arguments that the @xmath0 parallaxes for some of the clusters and associations represented in the @xmath0 catalog should be used with caution in the study of the galactic structure . </S><S> it has been already shown that the discrepancy between the @xmath0 and ground based parallaxes for several clusters including the pleiades , coma ber and ngc 6231 can be resolved by recomputing the @xmath0 astrometric solutions with an improved algorithm diminishing correlated errors in the attitude parameters . here </S><S> we present new parallaxes obtained with this algorithm for another group of stars with discrepant data - the galactic cluster cr 121 . </S><S> the original @xmath0 parallaxes led de zeeuw et al . to conclude that cr 121 and the surrounding association of ob stars form a relatively compact and coherent moving group at a distance of @xmath1 600 pc . </S><S> our corrected parallaxes reveal a different spatial distribution of young stellar populace in this area . </S><S> both the cluster cr 121 and the extended ob association are considerably more distant ( @xmath2 @xmath3 pc ) , and the latter has a large depth probably extending beyond 1 kpc . </S><S> therefore , not only are the recalculated parallaxes in complete agreement with the photometric @xmath4 parallaxes , but the structure of the field they reveal is no longer in discrepancy with that found by the photometric method . </S><S> 0@xmath5 -y@xmath6 1@xmath7 1@xmath8 1@xmath9 $ ] 1@xmath10 $ ] @xmath11 0@xmath12 0@xmath13 0@xmath14 </S>

hep-ex9903055

polarized deep inelastic scattering is a powerful tool for studying the internal spin structure of the nucleon @xcite .early results obtained by experiments at slac @xcite and cern @xcite indicated that quarks contribute very little to the nucleon s spin .this contradicted expectations of the naive quark - parton model , and led to the so - called spin crisis .additional theoretical and experimental studies followed , resulting in a richer view of the nucleon .experiments at cern @xcite , slac @xcite , and desy @xcite have provided increasing levels of precision and enabled the determination of polarized parton distribution functions using next - to - leading order perturbative qcd calculations and the ( altarelli - parisi or dglap ) evolution equations @xcite .the main goal of the e155 experiment at slac was , for both the proton and deuteron , to make a precision measurement of @xmath0 covering a wide kinematic range .to study the internal spin structure of the nucleons , the slac experiments have used the inclusive deep inelastic scattering of a polarized electron beam incident on a fixed polarized nucleon target .the kinematics of a scattering event are determined by the incident electron energy ( @xmath10 ) and the energy ( @xmath11 ) and angle ( @xmath12 ) of the scattered electron with respect to the incident beam direction .the event kinematics are described by the 4-momentum transfer squared @xmath13 , and the bjorken scaling variable @xmath14 .the bjorken scaling variable gives the fraction of the nucleon momentum carried by the struck quark . in the unpolarized case , the cross - section for inclusive measurements is related to the unpolarized structure functions @xmath15 and @xmath16 . in the polarized case ,the difference between cross - sections for anti - aligned versus aligned electron spins ( denoted @xmath17 ) and nucleon spins ( denoted @xmath18 ) is given by @xmath19 \label{eq : parallel_diff}\ ] ] for the case of longitudinal polarization of the target , and by @xmath20 \label{eq : perp_diff}\ ] ] for the case of transverse polarization of the target , where @xmath21 is the nucleon mass and @xmath22 is the fine structure constant .the notation @xmath23 ( and similarly for other cases ) is used below .equations ( [ eq : parallel_diff ] ) and ( [ eq : perp_diff ] ) introduce the spin structure functions @xmath24 and @xmath25 .these functions differ for the various targets ( proton @xmath26 , neutron @xmath27 , and deuteron @xmath28 ) , but are related by @xmath29 .the factor @xmath30 corrects for the d - state probability of the deuteron ,@xmath31 = [email protected] .this probability is treated as a constant with respect to bjorken @xmath8 . since in both cases on the left hand side of eqs .( [ eq : parallel_diff ] ) and ( [ eq : perp_diff ] ) the cross - sections are nearly equal , to measure @xmath0 and @xmath1 by measuring the cross - sections and taking the differences would require detailed and accurate knowledge of detector acceptance and efficiency .instead , the spin structure functions are obtained by measuring asymmetries , where these common factors divide out . measuring cross - section asymmetries @xmath33 and @xmath34 , @xmath35 yields the polarized structure functions via @xmath36 \hspace{3em}{\mathrm and}\hspace{3em } g_2 = { f_1 \over d^\prime } { y \over 2 \sin\theta } \left(- a_\| \sin\theta + a_\bot { e_0 + e^\prime \cos \theta \over e^\prime}\right ) \ , .\ ] ] the quantities @xmath0 , @xmath1 , @xmath15 , @xmath33 , and @xmath34 above are all functions of @xmath8 and @xmath9 .the kinematic variables used in the above expression are determined from @xmath10 and either @xmath37 or @xmath38 as follows : @xmath39 , @xmath40 , @xmath41 $ ] , @xmath42 , and @xmath43 $ ] .@xmath15 was calculated from fits to world data on @xmath16 @xcite and @xmath44 @xcite . in leading order @xmath0is only sensitive to the net quark polarizations .however , in next - to - leading order ( nlo ) perturbative qcd ( pqcd ) @xmath0 is also sensitive to the polarized gluon distribution @xmath45 through the @xmath46 evolution of the polarized parton distributions .this evolution is governed by the altarelli - parisi ( or dglap ) evolution equations@xcite and reflects the increasing resolution of deep inelastic scattering with increasing @xmath46 .one important consequence of this evolution is to link the quark distributions to the gluon distributions through g @xmath47 gq splitting .this makes it possible to indirectly measure the gluon distributions by measuring the quark distributions over a wide range in @xmath46 .given sufficient statistical power over a wide range in @xmath8 and @xmath46 , nlo pqcd analysis of the @xmath46 dependence of @xmath48 can be used to constrain the net polarization of the gluon .although the naive picture of the @xmath1 structure function is unclear , twist-2 calculations from wandzura and wilczek @xcite provide a relation between @xmath1 and @xmath0 : @xmath49 more generally , @xmath25 can be written as @xmath50 , where the second and third terms are twist-2 and twist-3 , respectively .the experiment e155 ran in early 1997 and collected @xmath51180 million deep inelastic scattering electron events .keys to the success of e155 included : the high intensity , highly - polarized beam ; the use of lithium deuteride as the deuteron target material ; and the construction of a third spectrometer .longitudinally polarized 48.3 gev electron beam pulses @xcite of up to 400 ns duration were produced at 120 hz by a circularly polarized laser beam illuminating a strained gaas photocathode . the beam polarization @xmath52=0.810@xmath32 0.020 was determined using mller scattering from 20154 @xmath53 m thick fe - co - v polarized foils @xcite .two polarized nucleon target materials were used in e155 : ammonia ( @xmath2nh@xmath54 ) as a proton target ; and lithium deuteride ( @xmath4li@xmath7h , or @xmath4lid ) @xcite as a deuteron target . as in e143 @xcite ,polarization of the target materials was obtained using the technique of dynamic nuclear polarization ( dnp ) . in this technique , a combination of microwaves ( @xmath51140 ghz ) , low temperature ( 1 k ) , and a high magnetic field ( 5 t ) polarizes paramagnetic electron centers in the target material , and transfers that polarization to the nucleons .nuclear magnetic resonance ( nmr ) measurements were made using coils embedded in the target materials to determine the polarization at regular intervals .these measurements were calibrated to the signal measured at thermal equilibrium near 1.6 k @xcite .the polarizations slowly decreased over time due to radiation damage , and were restored by periodic annealing at about 80 k for the nh@xmath3 target and 185 k for the lid . as compared to deuterated ammonia ( @xmath2nd@xmath54 ) , lithium deuteride provides a larger ratio of polarized ( effective ) deuterons to the total number of nucleons as well as higher radiation resistance .the @xmath4li can , to first order , be treated as a polarized deuteron plus an unpolarized alpha particle , and therefore half of the nucleons in @xmath4lid are the desired polarizable species .average polarizations of @xmath55 80% for the proton target and @xmath55 22% for the deuteron target were achieved .overall relative uncertainties on p@xmath56 of 6% ( preliminary ) and 4% were obtained for the proton and deuteron respectively .scattered electrons were detected in three independent magnetic spectrometers at central angles of 2.75@xmath57 , 5.5@xmath57 , and 10.5@xmath57 with respect to the incident beam , as shown in fig .[ fig : spectrometers ] .the spectrometers at 2.75@xmath58 and 5.5@xmath58 were used previously in the experiment e154 @xcite . in each spectrometer , electrons were identified by threshold gas cherenkov counters and a total absorbing lead glass electromagnetic calorimeter .particle momenta and scattering angles were measured with sets of scintillator hodoscopes .the 10.5@xmath57 spectrometer was added for e155 to double the @xmath46 range of the experiment .the experimental asymmetries were determined from the numbers of scattered electrons per incident beam charge for negative and positive beam helicity for each target , divided by the beam and target polarizations .the rates were corrected for contributions from charge symmetric background processes , which were measured by reversing the spectrometer polarities .corrections were also included for mis - identified hadrons , which were typically 2% or less of the electron candidates .also , a dilution factor accounted for the fraction of events originating from the polarizable protons or free deuterons , as opposed to the other materials in the target .the average dilution factor for the proton target was 0.15 , for the deuteron target 0.19 .nuclear correction factors accounted for the presence of several polarizable nuclear species in the target . for the proton , a small correction was made for the polarization of the nitrogen nuclei .for the deuteron , the measured asymmetry included contributions from the free deuterons and effective deuterons in @xmath59li .the effective deuteron in @xmath4li has a net polarization of 87% of the @xmath4li polarization @xcite , which leads to an effective dilution factor of @xmath510.36 for @xmath59lid , as compared with @xmath510.22 for @xmath2nd@xmath54 .the measured deuteron asymmetries also were corrected for small contributions from polarized protons in li@xmath60h and @xmath61li .both internal @xcite and external @xcite radiative corrections were obtained using an iterative global fit of all data , including e155 .previous slac data were recorrected in a manner consistent with the e155 corrections .the e155 proton results are plotted as @xmath62 vs.@xmath63@xmath9 in fig . [fig : fanplot ] . there is a remarkable agreement between the many experiments .the world data on @xmath62 exhibit a clear @xmath9 dependence , and the scaling violation is similar in character to that of the unpolarized structure functions . at low @xmath8 ,the spin structure function increases with @xmath9 , and at high @xmath8 it decreases with increasing @xmath9 .the e155 deuteron results @xcite are presented as @xmath64/@xmath65 vs.@xmath63@xmath46 in fig .[ fig : q2dep ] .they are also in good agreement with world data , and the data from the three spectrometers provide both large @xmath46 coverage and good statistical resolution in the mid-@xmath8 region .there is no significant @xmath46 dependence for @xmath66 , and likewise @xmath67 , indicating that the polarized and unpolarized structure functions evolve similarly .evolving the data to a common @xmath9 by use of a functional fit to world @xmath0 data , and integrating over the kinematic @xmath8 range of e155 yields preliminary results for @xmath68 ( gev / c)@xmath7 : @xmath69 and @xmath70 , where the first error is statistical and the second is systematic .the extrapolations for the unmeasured high @xmath8 region are negligible .however , the contribution from the low @xmath8 region below 0.014 does not converge for the functional fit , reinforcing the need for additional data at very low @xmath8 .using the e154 @xcite pqcd fit to calculate the contribution for the low @xmath8 region gives values : @xmath71 and @xmath72 , where again the first error is statistical , the second is systematic , and the third error is a theory error .these values can be combined to obtain a preliminary e155 result for the bjorken sum rule @xcite , @xmath73 .this is consistent with the theoretical prediction at @xmath68 ( gev / c)@xmath7 , @xmath74 , which includes third order qcd corrections @xcite , using @xmath75 @xcite . using the published nlo pqcd fit methods of e154 @xcite , but including final smc results @xcite and preliminary e155 results in the data set , polarized parton distributions were extracted .the eight - parameter fit had a @xmath76/d.o.f .of 206/210 . from the parton distributions ,the following preliminary integral results are obtained at @xmath68 ( gev / c)@xmath7 , in the @xmath77 scheme : @xmath78 = @xmath79 , @xmath80 = @xmath81 , and @xmath82 = @xmath83 , where the first errors are statistical and the second systematic , and similarly sized theory errors are ignored .the bjorken sum rule result is again consistent with the theoretical prediction .additionally , the fit polarized parton distributions yield the preliminary results : @xmath84 = @xmath85 and @xmath86 = @xmath87 .the data indicate that the gluon contribution to nucleon spin @xmath45 is positive , but do not constrain the value well .the data do well constrain the quark spin contribution @xmath88 , and it is smaller than predicted naively or by the ellis - jaffe sum rule @xcite .e155 ran for a short period during the 1997 run at 38.8 gev beam energy with the target polarization transverse to the beam direction .results for the @xmath1 structure function obtained from this data @xcite are shown ( as @xmath89 ) in fig [ fig : xg2 ] .the data from the three spectrometers show no @xmath46 dependence to the virtual photon - nucleon asymmetry @xmath90 within uncertainties , so the data from the individual spectrometers have been combined .the data are consistent with @xmath91 , and satisfy the burkhardt - cottingham sum rule @xcite within errors .however , the current results lack the power to differentiate among @xmath91 , model predictions @xcite , and @xmath92 .this will be addressed by the much larger data set planned for the e155 extension run ( march - may 1999 ) , which will focus on measurements with the target polarization perpendicular to the beam direction in order to make a high precision measurement of the @xmath1 structure function .the data from this run will significantly improve knowledge of @xmath1 and will definitively distinguish between @xmath1 = 0 and @xmath91 .

<S> the experiment e155 at slac measured the spin structure functions @xmath0 and @xmath1 of the proton and deuteron . </S><S> the experiment used deep inelastic scattering of 48.3 gev longitudinally polarized electrons incident on polarized solid @xmath2nh@xmath3 and @xmath4lid targets . </S><S> the data taken by three independent spectrometers covered a kinematic range of 0.014 @xmath5 0.9 and 1 ( gev / c)@xmath6 40 ( gev / c)@xmath7 . due to the high luminosity and polarization available at slac the data on @xmath0 are to date the most precise in this kinematic range . </S><S> the @xmath8 and @xmath9 dependence of @xmath0 has been studied using nlo pqcd fits , allowing extraction of values for the bjorken sum rule and quark and gluon spin contributions to the nucleon . </S><S> results are presented for @xmath0 and @xmath1 for the proton and deuteron . </S>

1701.07708

jem - euso @xcite is the concept of a space - borne fluorescence telescope to be hosted on the international space station ( iss ) .one of its main goals is a high statistical observation of eecrs with primary particle energies above 5@xmath0 .the telescope consists of three fresnel lenses and a focal surface consisting of 0.3 m pixels of uv sensitive detectors with a fast readout system , which enables a single photon counting measurement with a wide fov of @xmath1 with a spatial resolution of @xmath2 @xmath3 on the ground , covering roughly 60 @xmath4 of the entire surface of the earth in its flight on the iss orbit .looking towards the earth from space , jem - euso will reveal these particles as well as very high energy neutrinos by observing the fluorescence signal from the generated extensive air showers ( eas ) during their passage through the atmosphere .it will also contribute to the investigation of atmospheric phenomena such as transient luminous events ( tles ) and meteors .+ the jem - euso project is carried out by an international collaboration consisting of currently 88 institutions from 16 countries . in parallel to the jem - euso development , several pathfinders such as euso - balloon @xcite , launched in august 2014 , and a ground - based pathfinder ta - euso @xcite , currently in operation at the black rock mesa in utah , us have been developed .also , a space - borne pathfinder mini - euso as well as an advanced balloon - borne pathfinder with nasa super pressure balloon , euso - spb , are currently being developed to be launched in 2017 .+ the iss is flying at a speed of 7.5 km / s , which makes 15.5 orbits per day , and every 30 to 45 minutes it passes by the night region . while the iss flies on the orbit , it passes by many kinds of sceneries such as oceans , clouds , city light , airglows , forests , lightning , and so on from the altitude of 400 km .[ ebresult ] [ eusoflt ] the fig .[ ebresult ] shows the uv intensity measured during the five hours of an euso - balloon flight in august 2014 .the balloon was flying from left to right , from timmins airport to the dark region where mostly lakes and forests are located .the plot shows the averaged intensity in reference area in logarithmic scale as a function of time ( utc ) .the high intensities at the beginning of the flight ( left part in the plot ) appeared when the balloon was flying above cities like timmins and neighbourhoods , mines and airports , where there are full of artificial lights .one may see the uv background level is increasing by a factor of 10 comparing to other areas mainly covered by forests and lakes .jem - euso focal surface is consisting of hamamatsu 64-ch multi - anode photomultiplier tubes ( mapmts ) .one elementary cell unit ( ec_unit ) consists of 4 pmts , and 9 ec_units form a photo - detector module ( pdm ) . in total , the entire focal surface consists of 137 pdms with about 5,000 pmts with 0.3 m pixels .the output of pmts are readout by an elementary cell asic board ( ec_asic ) which consists of 6 spaciroc asics @xcite .a pdm board is an interface board between a pdm and the latter part of data processing system , which sends slow control commands and processes first level triggers ( flts ) .jem - euso will have to deal with a huge amount of data .therefore , it is essential to develop an efficient flt which detects uhecrs in a continuously varying background . the fig .[ eusoflt ] shows a conceptual figure of euso flt ( top ) with an example of an euso - balloon laser event @xcite ( bottom ) ; i.e. , an integrated image during 320 @xmath5s ( left ) and 3@xmath63 pixel counts as a function of frame ( gate time unit , gtu=2.5@xmath5s , right ) .the flt is a persistence trigger which scans counts in every 3@xmath63 pixel - boxes in a pmt , checking the excess against the pixel threshold which is automatically adjusted in every 320 @xmath5s .flt sets a proper threshold automatically as indicated by the blue dotted line in the plot to avoid triggering on the background , but only on the eas - like signals .[ vasca ] [ 9minimages ] [ fltoffline ]to test our electronics , as well as study and develop the flt in such a dynamic condition , we got an idea to use a big rotation tank of turlab @xcite , which is located in the fourth basement of the physics department building of university of turin .+ turlab is a laboratory for geo - fluid - dynamics studies , where rotation is a key parameter such as coriolis force and rossby number . with using inks or particles , based on the fluid - dynamics , key phenomena such as planetary atmospheric and fluid instabilities can be reproduced in the turlab water tank .the tank has 5 m diameter with a capability of the rotation at a speed of 3 s to 20 min per rotation .also , as it is located in a very dark environment , the light intensity can be controlled artificially .by means of the rotating tank with the capabilities mentioned above , we have been testing euso electronics such as its basic performance as well as the flt for cosmic rays in the various background and sceneries which are transiting from one to another .+ when we set an euso ec_unit camera , a 2@xmath62 array of 64-ch mapmts on the ceiling at the height of 2 meters above the turlab tank , we know that a pixel watches a fov of 5@xmath7 @xmath8 from the past measurements .this fov corresponds to a solid angle of 6.25@xmath9 which is comparable to the one for jem - euso .therefore , considering the altitude and the speed of iss , we learned that we can reproduce the jem - euso observation in turlab .the fig .[ vasca ] shows the turlab tank , light sources and materials used to mimic the various kinds of phenomena and albedos that jem - euso will encounter .we use an euso ec_unit with a lens as a camera , readout by the euso front - end electronics with a test board , observing several materials such as meteor , cosmic ray , cloud , ground glass , passing from one to another in the fov in a constant background light produced by a high power led lamp diffused on the ceiling .high voltages , dc power supplies , function generators and monitoring oscilloscopes are on the desk by the side of the tank with a pc with labview interface for the test board to send commands for slow control and data acquisition , and root programs for monitoring and analysis . fig .[ 9minimages ] shows the examples of uv images obtained by the ec_unit camera during the full tank rotation .the top panel of fig .[ fltoffline ] shows the result of uv intensities in a full rotation of the tank with various materials .the plot shows the summed counts of a pmt ( @xmath1064 pixels ) as a function of time in a unit of frame .afterward , we analysed the data and processed the flt offline .the middle panel shows the averaged counts per pixel which is used to set the flt threshold , while the bottom shows when flts were issued based on signals in that pmt , as a function of frame ( gtu ) respectively .almost all triggers coincide with passing over the arduino driven led chain , which mimics cosmic - ray like events , as it should be ; the one that is not is due to a specific location near one of the two bridges crossing the tank where the variations of light reflection were still too fast to be compensated by the slower rotation of the tank .jem - euso and its pathfinders will observe various cosmic / atmospheric phenomena in uv from the space / atmosphere .turlab is a unique facility in turin with interdisciplinary experts ( waves , geophysics , atmospheric science , meteors , astroparticle / cosmic ray physics , etc . ) , capable of providing an ideal condition to test euso electronics in a controlled environment against various sceneries which the telescope will encounter .the turlab system has been used to verify / implement flt in use of euso - spb .analysis and development of flt for mini - euso are currently ongoing .+ 9 j. h. adams et al .( jem - euso coll . ) , exp .astronomy 40 ( 2015 ) 3 .j. h. adams et al .( jem - euso coll . ) , exp .astronomy 40 ( 2015 ) 281 .j. h. adams et al .( jem - euso coll . ) , exp .astronomy 40 ( 2015 ) 301 .h. miyamoto et al . , pos tipp2014 ( 2015 ) 362 .j. eser et al .( jem - euso coll . ) , proc .34th icrc , den haag , @xmath110860 ( 2015 ) m. bertaina et al .( jem - euso coll . ) , epj web of conferences 89 ( 2015 ) 03003 .

<S> the turlab facility is a laboratory , equipped with a 5 m diameter and 1 m depth rotating tank , located in the physics department of the university of turin . </S><S> the tank has been built mainly to study problems where system rotation plays a key role in the fluid behaviour such as in atmospheric and oceanic flows at different scales . </S><S> the tank can be filled with different fluids of variable density , which enables studies in layered conditions such as sea waves . </S><S> + the tank can be also used to simulate the terrestrial surface with the optical characteristics of different environments such as snow , grass , ocean , land with soil , stones etc . , fogs and clouds . </S><S> as it is located in an extremely dark place , the light intensity can be controlled artificially . </S><S> such capabilities of the turlab facility are applied to perform experiments related to the observation of extreme energy cosmic rays ( eecrs ) from space using the fluorescence technique , as in the case of the jem - euso mission , where the diffuse night brightness and artificial light sources can vary significantly in time and space inside the field of view ( fov ) of the telescope . </S><S> + here we will report the currently ongoing activity at the turlab facility in the framework of the jem - euso mission ( euso@turlab ) . </S>

cond-mat0306493

microscopic models with long - range ( lr ) interactions decaying as a power law , i.e. , as @xmath1 , have aroused renewed interest during the last decade . beyond their fundamental relevance to the understanding of critical phenomena ,they have started playing a seminal role in the modeling of neural networks @xcite and spin glasses with ruderman - kittel - kasuya - yosida ( rkky ) interactions @xcite , systems undergoing phase separation , e.g. , highly ionic systems @xcite and model alloys @xcite , and more widely in a large class of chemical or biological models where electrostatic interactions , polarization , or van der waals forces play a central role .they have also attracted much attention in the framework of nonextensive thermodynamics , where a possible equivalence with short - ranged ( sr ) models is under consideration @xcite . since the very early work of ruelle @xcite , lr spin models in particular have been extensively studied . in one - dimensional models , it has been widely shown that long - range order occurs at finite temperature if and only if @xmath6 @xcite , and this is in strong contrast to the sr case where no phase transition exists at finite temperature .fisher and co - workers @xcite have shown that the upper critical dimension is reduced to @xmath7 , whereby one - dimensional lr models exhibit mean - field - like behavior for @xmath8 , with the critical exponents taking on their classical values @xmath9 and @xmath10 provided the phase transition is continuous .conversely , the critical behavior for @xmath11 yields nontrivial exponents , and lr models in effect go through a variety of universality classes as @xmath12 is varied within this range , thus exhibiting rich critical behavior . due to the ability to continuously vary the range of interaction , which in effect alters the effective dimension of the model , one - dimensional lr models are therefore a powerful paradigm for studying the dependence of critical properties on dimensionality , e.g. , in systems above their upper critical dimension @xcite . while significant emphasis has been placed on the ising chain ( see , e.g. , @xcite for a review ) , specific studies of the lr @xmath0-state potts model are less numerous and rather recent .these include a transfer matrix study combined with finite - range scaling ( frs ) @xcite , renormalization group ( rg ) analyses based on wilson s momentum - shell method @xcite or a real - space procedure @xcite , a cluster mean - field approach @xcite , and monte carlo ( mc ) simulations @xcite .the last , however , mostly focused on the case @xmath13 , and led to numerical estimates of critical exponents and temperatures showing some discrepancies .due to the higher ground state degeneracy , this model reveals a phase diagram markedly richer than that of the ising chain .it has been shown in the sr case that the transition turns from a continuous to a first - order one as the number of states @xmath0 is increased beyond a threshold value @xmath14 depending on the dimensionality of the model .for instance , @xmath15 and @xmath16 @xcite ( see also ref .@xcite for a complete review ) . as for the lr case in @xmath17 , glumac and uzelachave shown from mc studies of the three- and five - state potts model @xcite that the same sort of behavior occurs , i.e. , there is a so - called tricritical point at some value @xmath2 depending on @xmath0 , and the transition is continuous for @xmath18 .this qualitative picture was later reinforced in @xcite for @xmath19 and in @xcite for @xmath13 , both relying on mc studies , and in @xcite using a graph - weights approach .on the other hand , it is noteworthy that rg analyses dedicated to lr models have remained thus far rather inconclusive , where distinguishing between first- and second - order transitions is concerned @xcite .although it is now believed that @xmath20 depends continuously on the range of interaction for this class of models , the exact location of the tricritical line separating both regions is still fairly controversial .the biggest hurdle for a precise and reliable determination of this borderline actually stems from the weakening of the discontinuous transition as @xmath21 is approached , even if exceptionally large lattice sizes are simulated , e.g. , using the efficient luijten - blte cluster algorithm @xcite . while for @xmath13 @xmath21 was claimed to lie between @xmath22 and @xmath23 in @xcite , krech and luijten pointed out that @xmath24 still belongs to the first - order regime , and that the second - order regime may set in for @xmath25 only @xcite .the situation with @xmath26 turns out to be even worse , with numerical estimates available only within fairly large ranges : a lower boundary value of @xmath27 was reported in @xcite , whereas @xmath28 according to @xcite .these results have not yielded a very precise phase diagram as yet , with the only reliable assertion being that @xmath2 increases with @xmath0 .the marginal case @xmath29 raises another set of thorny questions : in @xcite it was reported that the phase transition changes from a second - order to a first - order one for @xmath30 , while it has been shown by kosterlitz , using a model with a continuum of states @xcite , and later on by cardy , using a discrete model @xcite , that inverse square interactions give rise to a kosterlitz - thouless ( kt ) transition , i.e. , one governed by topological defects @xcite .it is worth mentioning that both hypotheses may be reconciled , at least partially , by following a scenario similar to the one devised in @xcite , whereby for @xmath31-like models with nonlinear nearest - neighbor interactions , the kt - like transition is preempted by a first - order transition whenever the nonlinearity becomes strong enough .while the recent work of luijten and messingfeld on the three - state potts model @xcite lends further support to cardy s assertion , the controversy still appears unsettled , however , and in this view a determination of the asymptotic behavior of @xmath2 as @xmath32 seems of major interest indeed .we wish to shed light on some of these contradictory results using mc simulations in generalized ensembles , with particular emphasis put on the first - order regime .the aim of this work is thus twofold .first , we propose an implementation of the multicanonical algorithm dedicated to the numerical study of lr models .this algorithm , devised by berg and neuhaus a decade ago @xcite , has been successfully applied in the past to sr models undergoing first - order transitions .as numerical studies of models exhibiting first - order transitions are dramatically hampered by huge tunneling times when using standard metropolis update mechanisms @xcite , a multicanonical approach is indeed an appropriate choice for both the determination of the location of the tricritical line and the estimation of critical couplings in the first - order region @xmath33 .our purpose is therefore to adapt the scheme initially proposed for sr models so as to make it suitable for a large class of lr models .second , by relying on an extensive study for @xmath34 and a wide range of @xmath12 values , we arrive at convincing conclusions regarding the location of the tricritical line , the range of validity of the mean - field - like behavior , which we find much larger than in previous studies , and the crossover from the lr to the sr regime , although the last was investigated for the three - state model only .we show that our multicanonical implementation yields numerical estimates which are in agreement with and often better than those found in previous studies , although our simulations were performed by relying on medium lattice sizes , i.e. , @xmath35 spins . in particular , we obtain the following estimates for @xmath2 : @xmath36,@xmath37 , @xmath38 and @xmath39 , and these results are highly precise .we also offer convincing evidence that the phase transition in the limiting case @xmath40 is not of the first order for all values of @xmath0 , by virtue of an unusual finite - size effect . a detailed finite - size scaling ( fss ) analysis conducted for @xmath41 showsthat , while the transition belongs to the first - order regime at finite lattice size , its first - order nature wanes quickly enough as size is increased so that the transition tends to a continuous one in the thermodynamic limit .we give a qualitative account of this behavior in terms of correlation lengths , and by raising some open questions regarding the dynamics of first - order transitions in the lr case , we try to challenge the usual picture inherited from sr models . finally , by relying on the shape of the specific heat and computing several moments of the magnetization, we conclude that a crossover between lr and sr regimes occurs inside a narrow window @xmath5 the layout of this article is as follows . in sec .[ sec : model_theory ] , we first review some prominent features of the lr potts model through a mean - field ( mf ) analysis .special emphasis is given to the calculation of the location of spinodal points , a feature we will use in sec . [ sec : results ] for estimating @xmath2 .section [ sec : muca_algo ] is devoted to implementation details of the multicanonical algorithm specific to lr models .we discuss the iteration procedure used to obtain the best estimate for the density of states , the choice of an efficient predictor , and a reliable convergence criterion .improvements over the original algorithm are made in order to work out the algorithm instability due to low energy levels being unequally spaced .numerical results regarding both first- and second - order regimes are then presented in sec .[ sec : results ] .since we do not know of any previous implementation of a generalized ensemble algorithm in the case of lr spin models , we pay particular attention to comparison with other standard mc algorithms , i.e. , in terms of dynamical exponents , tunneling times , and accuracy of numerical estimates of critical couplings .throughout this work we consider a ferromagnetic potts model incorporating lr interactions in @xmath17 .this model is derived from a generalized @xmath0-state potts hamiltonian , i.e. , @xmath42 where the potts spin @xmath43 at site @xmath44 can take on the values @xmath45 , the first sum runs over all pairs of sites , and @xmath46 is an external aligning field favoring condensation in state @xmath47 .incorporation of lr interactions is carried out by setting @xmath48 where @xmath17 throughout this study , and @xmath12 is an adjustable parameter which can be related to the effective dimension of the model .as @xmath12 falls off to @xmath49 , this model tends to the mean - field case where all interactions have equal strength , whereas the limiting case @xmath50 corresponds to a pure sr model .crossover from lr to sr behavior should actually take place at @xmath40 @xcite , yet no numerical evidence has been given so far for this model which would reinforce this assertion .the thermodynamics of the model is studied numerically by way of the following order parameter : @xmath51 where @xmath52 , and @xmath53 is the density of potts spins in state @xmath54 , which varies between @xmath55 at infinite temperature and @xmath56 in the ground state . on a lattice of size @xmath57 , numerical implementation is carried out by using periodic boundary conditions , i.e. , one adds up interactions between all the spins of the original lattice only , and replaces the bare coupling constant @xmath58 by @xmath59 .retaining only interactions such that @xmath60 leads indeed to strong shifts in energy and critical couplings for low @xmath12 values , especially when finite - size scaling is to be used with medium lattice sizes .for the purpose of numerical evaluation , this sum may be reexpressed as @xmath61,\end{aligned}\ ] ] where @xmath62 denotes the hurwitz zeta function .the self - energy will be omitted since it is just an additive constant to the total energy .mean - field behavior can be readily obtained by using a variational mf approach ( see , for instance , @xcite ) , which relies on the minimization of the following functional @xmath63 = { \operatorname{tr}}\rho h + kt { \operatorname{tr}}\rho\ln\rho\ ] ] with respect to a trial density matrix @xmath64 . herethe trace operation means a sum over all spin configurations , and the dependence of @xmath65 and @xmath64 on the spin configuration is implied .$ ] reaches a minimum whenever @xmath67 , i.e. , in the case of a canonical gibbs distribution , and this minimum yields the free energy of the system .the mean - field approximation allows us to express the density matrix @xmath64 of the whole system as a product of one - site density matrices @xmath68 which depend solely on the spin variable at site @xmath44 .we may thus rewrite the trace operation as a sum involving traces on single spin variables , namely , @xmath69 & = kt \sum_i { \operatorname{tr}}_i \rho_i \ln\rho_i -\sum_i h_i { \operatorname{tr}}_i \rho_i \delta_{\sigma_i,\sigma_0 } \\ & -\frac{1}{2 } \sum_{i\neq j } { \operatorname{tr}}_i \rho_i { \operatorname{tr}}_j \rho_j j_{ij } \delta_{\sigma_i,\sigma_j}.\end{aligned}\ ] ] for further comparison with numerical results , we are mainly interested in expressing the free energy as a function of an order parameter which is as similar as possible to the one defined above .this is carried out by parametrizing the trial density matrix @xmath68 in terms of the following order parameter field : @xmath70 where the average is weighted by the trial density matrix @xmath68 .seeing that all states but state @xmath47 are equivalent , the constraint @xmath71 thus yields @xmath72 considering a uniform external field @xmath73 , we have @xmath74 for all sites ; hence the free energy per spin @xmath75 reduces to @xmath76 \}\end{aligned}\ ] ] where we dropped terms which are constant in @xmath77 so that @xmath78 , and @xmath79 is the riemann zeta function .this function expands as @xmath80 around @xmath81 ; hence transition temperatures are expected to vary as @xmath80 in the vicinity of the mf regime .equilibrium values of the order parameter are located at minima of the free energy , and it can be seen that @xmath82 is a stable minimum for @xmath83 . for @xmath84 , there is no third - order term in the series expansion of @xmath75 ; hence a second - order transition occurs at @xmath85 . for @xmath86 , the negative coefficient in the third - order term of the series expansion creates a second minimum , which physically corresponds to a first - order transition . at the transition temperature ,the free energy has the same value at both minima . following @xcite ,the exact transition temperature @xmath87 may be computed by simultaneously solving @xmath88 and yields @xmath89 and @xmath90 together with the reduced transition temperature @xmath91 , as a function of @xmath0 in the mf approximation.,title="fig : " ] similarly , spinodal points are computed by jointly solving @xmath92 , giving temperature points at which either one of the two minima vanishes .these equations possess one trivial solution , namely , @xmath93 corresponding to the extrema at @xmath82 becoming unstable , and a nontrivial solution @xmath94 which may be obtained numerically by solving the following equation , @xmath95 where @xmath96 and we have set @xmath97 . alternatively, one may also express @xmath98 as a function of the mf energy @xmath99 and impose @xmath100 .while these equations yield the same @xmath101 and @xmath94 as above , the two expressions of @xmath98 obviously do not have the same shape .spinodal points are sketched in fig .[ fig : theory_potts_model_metastable_states ] for @xmath0 between @xmath102 and @xmath103 .these correspond to the limit of metastability for each subphase , respectively . for temperature points lying inside this temperature range ,there exist two values of the order parameter corresponding to a null curvature of the free energy , a feature which is known to induce a long - ranged ( i.e. , low wave number ) instability .this in turn triggers a phase transition through the so - called spinodal decomposition @xcite . as expected , the width of the spinodal curve @xmath104 shrinks to zero as @xmath105 , and accounts for the second - order nature of the transition at @xmath84 , since in this limit the two minima merge into a single large minimum responsible for the well - known divergence of fluctuations at a continuous transition .the metropolis algorithm ( hereafter denoted as belonging to the class of _ canonical _ algorithms , i.e. , relying on a boltzmann weighting ) has long been considered the paradigm for monte carlo simulations in statistical physics , yet this method faces some severe drawbacks in situations where the sequence of states created by the markovian chain leads to very repetitive dynamics , i.e. , dramatically low acceptance rates and exponentially diverging autocorrelation times : this makes it necessary to simulate systems over exceedingly long runs in order to obtain good statistics and reliable estimates of thermodynamical averages ( see , for example , @xcite and the contribution by krauth in @xcite for an introductory review ) .this is the case when one comes to simulating systems with rugged free energy landscapes , e.g. , polymers , proteins , and disordered systems including spin - glasses , for the dynamics may then get trapped in one of numerous local minima , especially at low temperature .one experiences similar behavior when simulating first - order phase transitions ( the so - called supercritical slowing down @xcite ) , where the tunneling time between coexisting phases grows exponentially with the system size , due to the increasingly high free energy barrier to be overcome ( e.g. , @xcite ) .since slow dynamics mainly results from the combination of weighting the markovian chain with boltzmann weights and using local updates , there have been several attempts to devise efficient update algorithms based on _ global updates _ , e.g. , cluster algorithms , which in the case of continuous transition decrease critical slowing down by several orders of magnitude ( see @xcite ; also a lr implementation in @xcite ) . on the contrary , multicanonical methods @xcite are based on random walks in the energy landscape , irrespective of the particular move update utilized , whereby a _ flat _energy distribution is now sampled .first , this results in the algorithm quickly sampling a much wider phase space than in the canonical case , by allowing the system to cross any free energy barrier .second , this allows the density of states to be computed over the whole energy axis , thus extending the reliability of reweighting procedures over a much wider range of temperature than in the case of standard histogram methods @xcite , where poor histogram sampling at low - energy usually induces strong statistical bias .as opposed to multihistogramming @xcite , a single run is needed to cover the energy range of interest .once a reliable estimate of the density of states has been obtained , it is then straightforward to compute thermodynamical functions otherwise hardly within reach of canonical simulations , e.g. , canonical entropy and free energy .it is noteworthy that this simulation technique actually belongs to a larger class of algorithms called _ generalized - ensemble _ algorithms , which encompasses variants based on random walk in the entropic variable ( `` @xmath106 ensemble '' or `` entropic sampling '' algorithms @xcite ) , or the temperature variable ( e.g. , `` simulated tempering '' @xcite ) .the rationale behind the multicanonical algorithm is the generation of a markovian chain of states @xmath107 , whose weights @xmath108 are tweaked so that one eventually gets a flat energy histogram , i.e. if @xmath109 denotes the probability in energy and @xmath110 is the density of states , @xmath111 since @xmath110 usually increases drastically with energy , low - energy states are thus sampled much more often than high - energy ones .following berg in @xcite , we compute @xmath112 through an iterative procedure , starting from an initial canonical simulation at inverse temperature @xmath113 .@xmath113 indirectly sets the energy below which the energy histogram is to be flat , i.e. @xmath114 .thus , @xmath115 must be chosen high enough to ensure that the final energy histogram spans a suitably large energy range upward , e.g. , reaches the energy of the disordered phase in the case of a first - order transition , and extends even further away if one wants to observe with satisfying accuracy the free energy plateaus signaling the limit of metastability . for convenience , we now define an effective hamiltonian @xmath116 , so that @xmath117 hence , multicanonical simulation can be envisioned as a canonical simulation at inverse temperature @xmath113 with the usual boltzmann weight , provided the original hamiltonian is replaced by an effective hamiltonian to be determined iteratively . as a side note , a cluster implementation in the framework of the multicanonical algorithm is thus far less straightforward , since this effective hamiltonian has fundamentally a global nature , whereas canonical simulations explicitly preserve the locality of the original hamiltonian ( see , e.g. , the multibond approach in @xcite ) . denoting @xmath118 as the true estimate of the effective hamiltonian , we may thus write @xmath119 the microcanonical inverse temperature @xmath120 may be easily related to @xmath118 , as we have ( assuming @xmath121 ) @xmath122 since the dynamics of the markovian chain is governed by the transition rate @xmath123\})$ ] , we may write , for two states infinitely close in energy ,i.e. , whenever @xmath124 , @xmath125)$ ] .hence it is the _ microcanonical temperature _ which is the relevant quantity where the dynamics ( e.g. , the acceptance rate ) of the multicanonical algorithm is concerned .we initially set @xmath126 , or equivalently @xmath127 , as this indeed corresponds to a canonical simulation at temperature @xmath128 . at step @xmath44 ,a simulation is performed using a boltzmann weight with effective hamiltonian @xmath129 ; then an energy histogram @xmath130 is eventually computed using independent samples .incidentally , taking truly independent samples proves useful during the late stages of the iteration scheme only , where the aim is then to refine a nearly flat histogram . during early iteration steps, histograms may be computed using nonindependent samples without significantly affecting the convergence .we now denote @xmath131 as the lowest energy level that was reached throughout the previous runs , including step @xmath44 : this is the energy level below which @xmath132 will have to be predicted , since no histogram data are available inside this energy range .issues regarding adequate predictor choice will be considered later on in this section .the rules for updating @xmath133 at step @xmath134 from @xmath135 at step @xmath44 are based on the following equations . for @xmath136 , @xmath137 ,i.e. , the dynamics is canonical - like at inverse temperature @xmath113 for all iteration steps . for @xmath138 , @xmath139 where @xmath140 and @xmath141 is a _raw _ inverse damping factor proportional to the reliability of the @xmath142th histogram .it has been shown in @xcite , following an error calculation argument , that @xmath143 provides an estimator proportional to the inverse of the variance of @xmath144 .once @xmath144 is known , @xmath145 is derived by a mere integration starting from the initial condition @xmath146 .finally , for @xmath147 , @xmath132 will have to be computed using a suitably chosen predictor , until at last @xmath131 becomes equal to the ground state energy .a cubic spline is then fitted to @xmath116 at every bin center , and this curve is used to compute acceptance probabilities during the next run .it can be seen that eq .( [ equ : update_beta_g0cum ] ) leads to a steady state whenever @xmath148 is constant over the energy range of interest .writing a recursion equation involving @xmath120 instead of @xmath116 , together with the inclusion of a damping factor , allows one to handle the situation where some bins have null entries , a case which otherwise leads to a fairly spiky graph for @xmath116 and inconsistent dynamics ._ accidental _ null entries at energy values @xmath149 or @xmath150 will simply leave @xmath120 unchanged , and the corresponding parts of @xmath116 thus move as a block .since acceptance rates hinge on the microcanonical temperature , this in effect drastically reduces bias on the dynamics . considering a small set of histogram bins that are copiously filled for the first time during a given iteration run ( e.g. , high - energy bins during the early iteration runs whenever we begin with a canonical simulation ), we see that the related _ cumulative _ inverse damping factor first soars and produces a great amount of change in @xmath120 in the couple of runs that follow , and then decays progressively to zero as these bins continue to be filled . by taking into account all the data that have been sampled up to step @xmath44, this modified recursion both clearly stabilizes the algorithm and reduces relative errors due to poor histogram sampling . choosing the most appropriate value of the histogram bin width results from a trade - off between resolution and computation time . a higher resolution on the one hand guarantees good histogram flatness , and is especially crucial at low energy levels , where the density of states displays a rugged graphon the other hand , we impose a fixed number of independent samples per histogram bin , so as to give the histogram variance an acceptably low value ; hence a low @xmath151 implies more simulation steps per iteration .our approach is thus to first choose a fairly high @xmath151 , e.g. , one yielding around @xmath152 bins , during the early stages of the iteration process in order to obtain a rough picture of the density of states , and then to progressively reduce @xmath151 once the ground state has been reached . as will become obvious in sec .[ sec : free_energy ] , the ultimate value of @xmath151 deeply affects the attainable precision on the computation of spinodal points , since the latter is based on a precise location of free energy plateaus , and this indeed entails having enough bins belonging to a given plateau . as a rule of thumb ,the best compromise is then to obtain between @xmath153 and @xmath154 histogram bins in the final stage , with the number of bins increasing as the @xmath12 value corresponding to the second - order regime is approached ., computed by sorting energy samples from a long simulation run .each level is drawn as a horizontal line.,title="fig : " ] in this view , the unequal spacing of energy levels in lr spin models deserves specific attention . as witnessed in fig .[ fig : first - energy - levels - q5s05n400 ] , large energy gaps separate isolated energy levels or tiny groups thereof in the vicinity of the ground state , whereas the distribution gradually turns into a near continuum above @xmath155 .setting a low @xmath151 value leads in turn to _nonaccidental _ null entries in those bins located inside energy gaps , whereby @xmath120 never gets updated at isolated energy levels and @xmath156 is always zero . since the graph of the density of stateslooks indeed fairly wrinkled near the ground state , and the dynamics there is noticeably sensitive to even the smallest departure of @xmath116 from the ideal line , we would then observe a sharp steady peak in the lowest part of the energy histogram , which the present recursion would not be able to suppress .one could trivially think of working this out by implementing variable - width bins that would span energy gaps .this is , however , impracticable since the distribution of energy levels is not known prior to starting the iteration process ( for this is precisely what we intend to compute with the density of states ) . to circumvent this limitation ,we have modified the previous recursion so that null entries are always skipped , however accidental or nonaccidental they may be . denoting by @xmath157 and @xmath158 , with @xmath159 , the centers of histogram bins located on each side of a set of contiguous empty bins , we have @xmath160 where @xmath161 and we now impose @xmath162 hence @xmath156 can never be zero . in order to avoid losing details of the shape of @xmath116 for @xmath163 that were possibly collected during previous runs , we update @xmath116 through a linear difference scheme , @xmath164 where @xmath165 . while this has obviously no effect where nonaccidental null entries are concerned , this favors quicker convergence during the early runs where the inadequate shape of @xmath116 is more likely to produce empty bins .the iteration process stops whenever the energy histogram has become suitably flat over the energy range of interest , namely , between the ground state energy and @xmath166 for our purpose .we evaluate this property by computing the standard deviation of histogram entries , as well as the same quantity for the logarithm of histogram entries restricted to nonempty bins .the latter seems to be a better indicator since it is sensitive to both poorly populated bins and histogram peaks , whereas the former increases only with rather spiky histograms .in addition , we estimate the degree of convergence of the algorithm by computing the mean square distance between @xmath167 and @xmath132 after the ground state has been reached .we then compute a threshold value for each indicator by trial and error , based on a couple of short runs for various lattice sizes and bin widths .once @xmath116 has been satisfactorily computed , a long production run is performed using this effective hamiltonian in place of the original one , and then estimates of thermodynamical quantities of interest at inverse temperature @xmath168 are computed using a reweighting scheme , i.e. , formally , @xmath169 where @xmath170 denotes the microcanonical average of @xmath171 at energy @xmath149 , and the partition function is given by @xmath172 . the best estimate for the density of states @xmath110is provided by @xmath173 , where @xmath148 stands for the number of bin entries at energy @xmath149 computed from the production run . in order to avoid numerical overflows , as well as to suppress bias resulting from possibly strong variance on microcanonical averages , we found it more appropriate to compute @xmath174 from a sum running over samples instead of energy bins , i.e. , @xmath175 , where @xmath176 .@xmath177 is then determined so as to avoid both numerator and denominator overflows . providing that the histogram sampled during the production run is flat to a good approximation ,the maximum in @xmath178 is reached whenever @xmath179 , which yields the energy value at which @xmath177 is to be computed .in addition , since the reweighting scheme involves an exponential contribution of @xmath116 , the resulting curve @xmath178 is strongly peaked around the maximum ; hence it is clear that only histogram points in the vicinity of this maximum contribute to @xmath174 . in effect, we found that the existence of two distinct maxima , or equivalently of two energy values for which @xmath120 has the same value , coincides with the occurrence of a first - order phase transition .following the same reweighting procedure , we compute partial free energy functions , i.e. , @xmath180 where @xmath77 is the order parameter , and reweighted histograms of the energy , i.e. , @xmath181 .the partial partition function is straightforwardly derived from a partial sum over samples having the prescribed order parameter , @xmath182 which then yields @xmath183 .similarly , a reweighted histogram of the energy is obtained from @xmath184 .we now discuss some issues related to the choice of an efficient predictor for @xmath185 . for small lattice sizes, we initially feed the algorithm with an effective hamiltonian @xmath186 , and the objective is then to find an appropriate trade - off between speeding up the convergence of @xmath131 toward the ground state and avoiding algorithm instability . while the former demands that @xmath129 have a sufficiently high slope below @xmath131 , the latter still requires that the algorithm remain ergodic to a suitable extent .our implementation relies on a first - order predictor , @xmath187 , and we impose continuity on @xmath116 at @xmath188 .the simplest approach is then to choose a predictor slope so that continuity on @xmath189 is enforced at @xmath190 , i.e. , @xmath191 . while @xmath188 reaches the ground state rather quickly using this predictor , the dynamics often gets locked in very low energy levels due to the particularly steep slope of @xmath116 in the vicinity of the ground state .the time needed by the iteration scheme to recover from this deadlock and obtain a flat histogram thus becomes prohibitive .on the other hand , choosing @xmath192 leads to the smoothest yet slowest convergence , and avoids deadlock issues .an efficient compromise is thus to ensure a `` weak '' continuity at @xmath188 , i.e. , by computing the slope of the predictor using a least - square scheme based e.g. , on the first 10% of points above @xmath188 . for large lattice sizeswhere reaching the ground state energy can become time consuming , we resort to a `` scaling trick '' whereby @xmath116 is initially guessed from the density of states obtained at a smaller lattice size .this approach was initially mentioned by berg and neuhaus @xcite , and claimed to work perfectly within the framework of a study of the two - dimensional ten - state potts model with nearest - neighbor interactions where the energy is additive to a perfect extent .the presence of lr interactions , however , slightly worsens the case , especially at low @xmath12 .the scaled density of states is computed as follows .let us consider , for the sake of simplicity , two systems @xmath193 and @xmath194 with respective lattice sizes @xmath195 and @xmath196 , and let us divide the latter into two subsystems @xmath197 and @xmath198 of equal size @xmath57 . since @xmath199 , where @xmath110 stands for the density of states , we have to compute @xmath200 for @xmath194 as a function of @xmath110 for @xmath193 . neglecting the interaction between subsystems @xmath197 and @xmath198 , and denoting by @xmath201 the energy of @xmath197 , the density of states for @xmath194 just reads @xmath202 , which yields @xmath203}\\ & \sim \ln \frac{1}{\delta e}\int de_1 e^{\beta_0 [ h_{mu}(e_1)+h_{mu}(e - e_1)]},\end{aligned}\ ] ] where @xmath151 is the energy histogram bin width . providing that @xmath110 is a monotonic and rapidly increasing function of @xmath149 , we may use a saddle - point approximation to evaluate the former sum .the maximum of @xmath204 is reached whenever @xmath205 ; hence @xmath206 that were fed into the iteration scheme at @xmath207 , @xmath26 , and @xmath208 .each initial guess was computed using eq .( [ equ : scaling - trick ] ) , i.e. , by scaling a true estimate obtained at @xmath209 .solid lines show true estimates @xmath116 as obtained after the whole iteration scheme at @xmath207 converged .the straight dashed line sketches the original hamiltonian , i.e. , @xmath186.,title="fig : " ] spins , using eq .( [ equ : scaling - trick ] ) to compute the initial effective hamiltonian @xmath210 from a previous run at @xmath209 spins .labeling on @xmath211 axis indicates normalized probabilities.,title="fig : " ] this expression may be readily extended to lattice sizes that are any multiple of the original size .figure [ fig : scaling - predictor ] sketches results obtained for @xmath26 and @xmath212 and @xmath213 . a series of iteration runswas first conducted with @xmath209 spins in order to obtain an estimate of @xmath116 for this lattice size , then this estimate was scaled using eq .( [ equ : scaling - trick ] ) and used as the initial guess @xmath210 for the next series of iteration runs at @xmath207 .equation ( [ equ : scaling - trick ] ) yields a very acceptable guess for @xmath214 , and the two curves are hardly distinguishable from each other . as illustrated in fig .[ fig : flat_histo ] , the energy histogram becomes nearly flat within five iterations . for @xmath215 and @xmath216 ,the agreement remains quite satisfying , yet the initial guess falls slightly below the true estimate at low energy levels , and the lowest - energy bins are exceedingly enhanced during the first iteration runs .more iteration runs are thus required to obtain a perfectly flat histogram as @xmath12 is decreased , and the benefit of this approach in effect becomes negligible for @xmath217 . indeed , the algorithm then spends a great number of iteration steps being trapped in low energy levels , seeking to rectify the shape of the density of states in this energy region until convergence is obtained : starting from an initial canonical effective hamiltonian actually yields better performances . since , for systems with lr interactions , computation time scales with @xmath218 , using this `` scaling trick '' thus greatly reduces the time needed for proper convergence , at least for @xmath219 , and partially compensates for the lack of a hybrid multicanonical - cluster algorithm dedicated to lr models . in order to measure the performance of our implementation ,we have computed a dynamical exponent @xmath220 defined as the scaling exponent of a relevant characteristic time @xmath221 of the simulation , i.e. , @xmath222 , where @xmath57 denotes the lattice size : while for second - order transitions it is widely known that the integrated autocorrelation time represents such a relevant time , for first - order transitions the tunneling time through the energy barrier ( @xmath223 ) proves to be a more meaningful indicator @xcite .we define the latter as one half of the average number of monte carlo steps per spin ( mcs ) needed to travel from one peak of the reweighted energy histogram ( @xmath181 ) to the other , with the temperature being set to the transition temperature .tunneling time is expected to grow exponentially with @xmath57 for canonical algorithms , and to scale as a power law of @xmath57 for multicanonical algorithms @xcite . in both cases, it appears that the chosen characteristic time is a good indicator of how quickly the demands in cpu time should grow with increasing lattice size : for second - order transitions , this is the time needed to generate truly independent samples , while for first - order transitions , this tells us at what rate the dynamics spreads out over the energy barrier and thus to what extent samples get efficiently picked from the two phases in coexistence .the integrated autocorrelation time is computed by using the well - known time - displaced correlation function which displays an exponential - like short - time behavior , namely , @xmath224 ; @xmath221 is then derived from a simple integration scheme . since the latter function makes sense within equilibrium only , we first discard @xmath54 thermalization steps , where @xmath54 is obtained by using the nonlinear relaxation function that describes the approach to equilibrium @xcite and averaging over several dry runs .an interesting point regarding multicanonical simulations is that , since they are random walks in energy space , `` thermalization '' ( although this term is no longer appropriate as far as generalized ensemble algorithms are concerned ) always occurs rather rapidly ; simulations based on a nearly flat histogram have shown that a value of @xmath225 mcs is indeed appropriate on average . vs lattice size @xmath57 for @xmath226 and @xmath227 .dynamic exponents computed from a fit to @xmath228 are @xmath229 , respectively ., title="fig : " ] results for @xmath226 and @xmath12 lying between @xmath230 and @xmath27 are shown in fig .[ fig : autocorrel_times ] for integrated autocorrelation times , and in fig .[ fig : tun_time ] for tunneling times .the slight dispersion in the power - law fits arises from the fact that simulations at larger sizes were conducted with a higher number of mcs between measurements in order to reduce memory overhead . where computing tunneling times is concerned , this results in some tunneling events being possibly skipped and the average tunneling time being overestimated . both figures show , however , that a power - law behavior is perfectly plausible . in the case of first - order transitions ,the reduction in simulation costs is thus drastic in comparison with standard canonical algorithms . for both indicators ,we obtain an average @xmath220 slightly above @xmath231 for @xmath232 , yet @xmath220 increases smoothly with decreasing range of interaction .this may be accounted for by the fact that spatial and time correlations grow as we depart from the mf regime and approach the sr one .as for tunneling times , the prefactor turns out to be slightly higher near the mf regime , and @xmath220 increases at a lower rate with increasing @xmath12 than is the case for autocorrelation times . vs lattice size @xmath57 for @xmath226 and @xmath227 .dynamic exponents computed from a fit to @xmath228 are @xmath233 , respectively ., title="fig : " ] since there are no other numerical studies of lr models based on multicanonical algorithms to our knowledge , comparison is limited to estimates obtained for sr models .for the three - state potts model , canonical simulations using local updates led to @xmath234 @xcite , while swendsen and wang obtained @xmath235 using their percolation cluster algorithm @xcite . for further comparison ,the metropolis algorithm applied to a sr ising chain in @xmath236 and @xmath237 yielded a value of @xmath220 slightly above @xmath102 @xcite , whereas wolff s cluster algorithm led to @xmath238 @xcite .while our value is slightly greater than in the case of cluster implementations , it is worth underlining that our multicanonical implementation yields reliable statistics within a single mc sweep , whereas several are needed in the case of a standard canonical simulation , whatever reweighting procedure may be used .we have conducted multicanonical simulations for @xmath239 $ ] , using for each value of @xmath0 an appropriate set of @xmath12 parameters between @xmath240 and @xmath213 , so that we could observe strong and weak first - order transitions , as well as continuous ones . for @xmath13 ,some simulations were performed with @xmath241 in order to investigate the crossover from lr to sr regimes .once the density of states had been determined using the iteration process described above , a production run was performed for lattice sizes between @xmath242 and @xmath207 .the number of mc sweeps needed for each production run was computed so as to yield approximately @xmath243 truly independent samples . in this view , rapidly increasing autocorrelation times in effect restricted our work to lattice sizes @xmath244 . as already stated in the introduction ,a precise determination of the tricritical value @xmath2 is a real challenge , due to the weakening of the first - order transition as @xmath21 is approached from below .this makes traditional indicators e.g. , latent heats or energy jumps , fairly inefficient , since observing clear jumps in the vicinity of the tricritical value entails simulating huge lattices .glumac and uzelac in @xcite used three less traditional estimators , namely , the interface free energy , the specific heat , and the reduced fourth - order binder cumulant , which all turned out to be less sensitive to this weakening : in particular , the last quantity defined as @xmath245 is expected to reach a nontrivial constant @xmath246 as @xmath247 at a first - order transition only @xcite ; by extrapolating to the thermodynamic limit from measures taken at different lattice sizes , they found @xmath21 to fall between @xmath22 and @xmath23 for the three - state model .still and all , this approach imposes simulating fairly large lattices ( around @xmath248 ) for the extrapolation procedure to be reliable , let alone the fact that binder cumulants may experience uncontrollable crossover effects @xcite . due to the modest lattice sizes that are within reach of our local update algorithm, we rather resort to an approach based on the location of spinodal points , which may be accurately determined already for medium lattice sizes .in marked contrast to multihistogram techniques , the multicanonical method indeed allows one to obtain partial free energy functions ( or , equally , reweighted histograms of the energy ) over a range of temperature which extends fairly far away from the transition temperature , with remarkably modest numerical effort .the basis of our method relies on the fact that the temperature difference between the two spinodal points will tend to zero as @xmath21 is approached , since there are no metastable states in the case of continuous transitions. stated differently , the conditions under which metastability occurs , i.e. , both first and second derivatives of the partial free energy are zero , are met only at the critical point for a continuous transition : hence metastable states merge into a single large minimum as the first - order transition turns into a second - order one .such behavior has indeed been widely observed , e.g. , for liquid - vapor transitions near the critical point , and is borne out by our mf calculation . for a given set of @xmath249 parameters ,we determine the location of the spinodal points by first computing a partial free energy function of the order parameter [ @xmath250 , see eq .[ equ : free_energy_mag_reweighting ] ] over a large temperature range .alternatively , we make use of a similar function of the energy , i.e. , @xmath251 , where @xmath252 denotes the reweighted histogram of the energy .while the latter function plays the same role as the partial free energy of the magnetization , it yields a higher precision at low @xmath0 , as we will witness in a moment .the limit of metastability at finite lattice size is then defined by @xmath253 , or alternatively @xmath254 : for a first - order transition , this condition is met at two temperatures @xmath255 and @xmath256 which satisfy the inequality @xmath257 , where @xmath258 denotes the transition temperature . since these free energy functions usually have rather rugged graphs , we first filter out rapid oscillations by means of a linear smoothing filter , whose order is computed so that we are left with at most three extrema over the whole temperature range of interest . by continuously varying @xmath259 within this range , we determine the temperature of each metastable state by monitoring the change in the number of minima ( see fig . [fig : rew_histo_q5s4n400 ] ) .in contrast to @xcite , the transition temperature @xmath260 is then obtained by imposing that the number of bin entries in @xmath261 be the same below and above the energy corresponding to the maximum of @xmath262 .this corresponds to the so - called equal - weights condition as proposed by lee and kosterlitz in @xcite , and is equivalent to the condition that the average energy be the arithmetic mean of the energy of each phase .for the sake of comparison , however , we also compute the temperature @xmath263 at which both minima of @xmath262 have the same value .we then proceed with the computation of similar quantities using @xmath250 , and we estimate statistical errors using a bootstrap procedure . for @xmath264 , at four characteristic temperatures : @xmath265 , and @xmath266 denote the temperatures of the two metastable states , and the temperature of equal peaks heights , and that of equal peak weights , respectively .@xmath267 denotes the energy per spin.,title="fig : " ] , computed over both phases ( solid line ) , ordered phase only ( lower dashed line ) , and disordered phase only ( upper dashed line ) .vertical dotted lines indicate the four characteristic temperatures : from left to right , lower limit of metastability ( @xmath101 ) , transition temperatures ( equal heights @xmath268 , then equal weights @xmath269 ) , and upper limit of metastability ( @xmath94 ) ., title="fig : " ] graphs of the free energy @xmath262 in fig .[ fig : rew_histo_q5s4n400 ] show that the peak and the plateau corresponding to the disordered phase are much narrower than those of the ordered phase . as a result ,the precision in the determination of the temperature @xmath255 of the lowest metastable state is fairly lower than that of the upper metastable state ( @xmath256 ). this asymmetry increases with increasing @xmath0 , and in effect precludes the use of reweighted histograms for the estimation of spinodal points at @xmath270 . for @xmath41, we thus relied on the extrema of the partial free energy @xmath250 , since this function then becomes nearly symmetric and displays peaks that are well separated .incidentally , the asymmetric shape of @xmath262 can be accounted for by the fact that specific heats have a different magnitude in the disordered and ordered phases , since this thermodynamic quantity is simply proportional to the standard deviation of the associated gaussian peak @xcite .this may be readily observed by reweighting thermodynamical averages over a single phase at a time , once the maximum of @xmath262 which separates the two phases has been located .figure [ fig : mean_ene_per_phase ] shows how this procedure was applied to the computation of the mean energy per spin of each subphase for @xmath271 , and @xmath207 spins .a simple visual inspection allows one to assess a much lower specific heat for the disordered phase than for the ordered phase .at finite lattice size , however , all these temperatures experience a distinct shift proportional to the distance from the thermodynamic limit . assuming that the fss theory developed in @xcite for first - order transitions is also valid in the lr case, we therefore compute temperatures at infinite lattice size by assuming power - law corrections in @xmath272 .we also expect temperatures defining the limit of metastability to obey the same scaling behavior , although the phenomenological theory proposed in @xcite does not explicitly handle them .the inclusion of a second - order term proves necessary in order to to obtain satisfying fits , due to the presence of small lattice sizes in our set of data . yet , interestingly enough , fitting finite - size temperatures to a power law of the form @xmath273 yields very similar extrapolated values , with discrepancies smaller than @xmath274 , i.e. , within our range of uncertainty .in addition , we observed that @xmath262 and @xmath250 led to distinct finite - size shifts , with the latter function easily allowing one to drop second - order correction terms without much affecting the final result . for the sake of completeness , we also compute transition temperatures by relying on two other estimators , namely , the magnetic susceptibility , which for magnetic systems has more pronounced peaks than the specific heat , and binder cumulants of the magnetization defined as @xmath275 .the latter are known to cross at a critical fixed point @xmath276 defining the true critical temperature , yet , since the crossing point drifts smoothly over our range of lattice sizes , we assume a power law of the form @xmath277 for @xmath278 , with an unknown exponent @xmath279 @xcite .in addition , these two quantities are advantageously used to obtain critical temperatures in the second - order regime as well ( see sec . [sec : second_order ] for more details on this issue ) .results for all temperature estimates are summarized in table [ table : all_temperatures ] for @xmath19 , and sketched in fig .[ fig : spinodal_q_05 ] for @xmath26 . as expectedaccording to fss theory , both definitions of the transition temperature , i.e. , using equal peak weights vs equal peak heights , lead within error bars to the same estimates at infinite lattice size .other quantities @xmath280 and @xmath281 yield very similar results , with a discrepancy never exceeding 1% . [ cols="<,<,<,<,<,<,<,<,<,<",options="header " , ] [ table : exponents_allq ] if the relation @xmath282 derived in @xcite is indeed exact for @xmath283 , we should thus observe the simple behavior @xmath284 in the second - order regime . as illustrated in the fifth column of table [ table : exponents_allq ] , the qualitative behavior follows the conjecture , yet clearly @xmath285 , and the discrepancy is remarkably higher for @xmath26 than for @xmath13 .moreover , while it appears to shrink to @xmath286 as @xmath287 , it is unclear whether @xmath288 varies linearly with @xmath12 , considering the small number of points available . in order to get a deeper insight into the crossover to the sr regime, we then conducted several simulations at @xmath13 for @xmath12 above the borderline value @xmath289 .this value has been reported to play the role of a critical range of interaction beyond which a crossover from lr to sr behavior sets in . according to @xcite , @xmath290 , where @xmath291 denotes the value of the @xmath292 exponent in the sr case . since @xmath293 for all values of @xmath0 in the sr case , @xmath294 , and this indeed leads to @xmath289 .it should be noted , however , that this definition , as initially proposed by sak in @xcite on theoretical grounds , as well as the exact location of @xmath295 within the interval @xmath296 $ ] , is still controversial .as shown in table [ table : exponents_allq ] , @xmath288 indeed appears to reach its sr value as @xmath297 , yet this ratio proves no longer reliable above the borderline value , as we will witness in a moment , and reliance on other quantities becomes necessary .we first review some exact results concerning the sr regime , which we obtained using an exact transfer matrix method . for @xmath13 ,the transfer matrix is a @xmath298 matrix having three eigenvalues , which in zero external field read @xmath299 , @xmath300 , where @xmath301 . by retaining the largest eigenvalue @xmath302 only , and taking the limit @xmath303 ,we successively obtain the free energy per spin @xmath304 and the specific heat @xmath305 from there on , the correlation length is then computed using the standard formula @xmath306 , which then yields @xmath307^{-1}\ ] ] finally , the magnetic susceptibility is obtained using the fluctuation - dissipation relation , which gives @xmath308 it is then straightforward to show that @xmath309 . however , evaluating this ratio at finite inverse temperature , i.e. , for a finite correlation length as imposed by a finite lattice size , yields a greatly overestimated result .for instance , we obtain @xmath310 for @xmath207 , a feature which is supported by our simulation results , e.g. , @xmath311 , @xmath312 , and @xmath313 for @xmath314 , @xmath315 , and @xmath316 , respectively .since the last two values are clearly overestimated , this in effect indicates the presence of exponential divergences and as a by - product drastically slow convergence of the correction to scaling .this analysis was corroborated by a study of the shape of the specific heat , which turns out to provide the most tractable approach at medium lattice sizes where distinguishing between the sr and the lr regime is concerned . in the thermodynamic limit , @xmath317 admits a maximum @xmath318 at @xmath319 .it is enlightening to investigate the nonmonotonic behavior of this maximum at finite @xmath57 , and this may be carried out by computing @xmath320 and then @xmath321 while retaining all three eigenvalues . since the calculation is fairly involved , and the final result admits no simple expression , we shall hereafter simply refer to the corresponding curve sketched in fig .[ fig : cv_allsigma_vs_sr ] .when @xmath57 is increased , the peak of the specific heat first increases to a maximum , and then graphs of @xmath322 collapse and merge gently as the thermodynamic limit is approached . whenever it is witnessed in graphs obtained from simulation data ,this feature thus signals a sr - like behavior . andthe pure sr case ( obtained using a transfer matrix method ) , from right to left . data for other values of @xmath12 have been omitted in order to preserve the clarity of the figure .@xmath322 was computed using the fluctuation - dissipation relation @xmath323 .solid , dashed , dotted and long - dashed styles refer to @xmath324 , and @xmath325 respectively , except for the sr case where they refer to @xmath326 ., title="fig : " ] simulations were performed for @xmath327 for various lattice sizes between @xmath242 and @xmath207 , and we set the initial canonical temperature to @xmath328 so that the maximum of @xmath322 would be clearly visible within the whole range @xmath329 . as appears obvious from a glance at fig . [fig : cv_allsigma_vs_sr ] , the cases @xmath40 and @xmath314 , on the one hand , and @xmath330 , on the other hand , display fairly distinct qualitative behaviors . for @xmath40 ,the specific heat reaches its maximum monotically , at least for the lattice sizes that were investigated .the slowing down in the increase rate as @xmath331 allows one to assess a finite maximum in the thermodynamic limit , and this clearly shows that @xmath322 is a nondivergent quantity , thus bringing support to cardy s scenario whereby the transition has a kt - like nature on the borderline @xmath40 .the same behavior is observed for @xmath314 . on the contrary ,the qualitative behavior is clearly different for @xmath330 , where the maximum of @xmath322 first decreases with increasing lattice size , and then quickly reaches a plateau reminiscent of the exact sr behavior investigated above . while this plateau only slowly reaches the exact sr value as @xmath332 ( see fig .[ fig : cvmax_allsigma_vs_sr ] ) , we can however conclude that the behavior is already sr - like . from top to bottom .the solid line is a reminder for the ( exact ) sr case in the thermodynamic limit .other lines are guides to the eyes.,title="fig : " ] this assertion can be further confirmed by considering the magnetization , as sketched in fig .[ fig : mag_all_sigma ] .graphs of this quantity clearly merge slightly above @xmath82 , whenever @xmath330 ; hence there is no transition at finite temperature . while for @xmath314 there remains some ambiguity due to statistical errors , for @xmath40 the curves now clearly intersect around @xmath333 , which at least shows that the behavior is no longer sr - like. for @xmath334 , and @xmath335 from rightsolid , dashed , dotted and long - dashed styles refer to @xmath324 , and @xmath325 respectively ., title="fig : " ] we finally compute critical temperatures from the crossing points of binder cumulants of the magnetization .we obtain @xmath336 , @xmath337 , and @xmath338 for @xmath314 , @xmath339 , and @xmath315 . as for @xmath340 and @xmath341 , cumulants no longer cross except at @xmath342 within statistical error ( the latter case yielding @xmath343 between @xmath344 and @xmath345 , yet with excessive dispersion ) . while the crossover appears to take place in the very vicinity of the borderline @xmath40 , the critical temperature actually dies off quite slowly to @xmath286 as @xmath12 increases .all these numerical results lend support to sak s scenario for @xmath346 , namely , that a crossover from lr to sr behavior occurs whenever @xmath290 .nonetheless , it is worth mentioning that we found this crossover to occur within the finite , yet narrow range @xmath5 , and the pure sr case to be reached in the limit @xmath347 only .we feel strongly that this is consistent with the rg scenario of theumann and gusmao @xcite , whereby the crossover actually results from a competition between sr and lr fixed points .this competition , as seems obvious to us , may not resolve instantly whenever @xmath12 crosses the borderline , and may thus blur this borderline over some finite region .we have studied some critical properties of the long - ranged potts model using a multicanonical implementation of generalized ensemble algorithms .our implementation of the iteration procedure needed to obtain the density of states was shown to yield satisfying estimates of this quantity over a large range of energy and with much quicker and more stable convergence than with the initial historical algorithm .the multicanonical algorithm allows one to efficiently circumvent the slowing down traditionally experienced at first - order transitions , and at the same time makes the reweighting approach a fairly straightforward way of examining thermodynamic quantities over a large range of temperature with strikingly modest numerical effort , i.e. , by simulating over medium lattice sizes and performing a single long simulation run .we have used this multicanonical approach to locate spinodal points in the first - order regime over a large range of @xmath0 and @xmath12 parameters .the shape of the spinodal curve in the vicinity of the change of regime then yielded precise estimates of the tricritical value @xmath2 up to two digits .in particular , the value @xmath348 is perfectly consistent with the lower bound of @xmath23 proposed by krech and luijten @xcite , yet in terms of precision this is markedly better by an order of magnitude . in this respect ,our multicanonical implementation allows us to obtain numerical results whose accuracy is at least comparable to that of previous numerical studies based on multihistogramming and the lr cluster algorithm , although our simulations were performed on lattices having fewer than @xmath325 spins . we feel strongly that this approach might be successfully applied to other spin models incorporating lr interactions , e.g. , continuous spin models or frustrated systems . in addition , our study significantly extends the range of available estimates of critical couplings and exponents . in the first - order regime , the agreement with mf predictions , and in particular with tsallis s conjecture @xmath349 in the limit @xmath350 @xcite , is exceptionally good . in the second - order regime , the relation @xmath351 , conjectured to be exact for @xmath84 , is shown to yield an increasingly high discrepancy when @xmath0 is increased , and its validity may just be reinforced in the vicinity of @xmath40 .we found however that the crossover from the lr to the sr regime occurs between @xmath40 and @xmath352 , thus lending strong support to sak s conjecture .our detailed fss analysis of the case @xmath353 yielded one of the most surprising results of this study , namely , the unexpected behavior of correlation lengths whereby the transition appears to be of the first order at finite lattice size , despite the fact that fss theory predicts a continuous transition in the thermodynamic limit .we feel strongly that this may be accounted for by the truncation of the lr potential , which artificially brings the model closer to the mf regime , yet we also pointed out that the physical meaning of the correlation length should be somewhat challenged in the case of lr models .the exact nature of the transition in the borderline case @xmath40 , however , needs further investigation , especially at large @xmath0 where no results have been made available thus far .in this view , an efficient combination of a global update scheme with a multicanonical approach would be of prior importance to reach far higher lattice sizes .is indebted to dr .a. revel for providing access to computing facilities of the etis ( ensea / ucp / cnrs , umr 8051 ) research team .`` laboratoire de physique thorique et modlisation '' is associated with cnrs ( umr 8089 ) .

<S> we investigate the critical behavior of the one - dimensional @xmath0-state potts model with long - range ( lr ) interaction @xmath1 , using a multicanonical algorithm . </S><S> the recursion scheme initially proposed by berg is improved so as to make it suitable for a large class of lr models with unequally spaced energy levels . </S><S> the choice of an efficient predictor and a reliable convergence criterion is discussed . </S><S> we obtain transition temperatures in the first - order regime which are in far better agreement with mean - field predictions than in previous monte carlo studies . by relying on the location of spinodal points and resorting to scaling arguments </S><S> , we determine the threshold value @xmath2 separating the first- and second - order regimes to two - digit precision within the range @xmath3 . </S><S> we offer convincing numerical evidence supporting @xmath4 for all @xmath0 , by virtue of an unusual finite - size effect , namely , finite - size scaling predicts a continuous transition in the thermodynamic limit , despite the first - order nature of the transition at finite size . a qualitative account in terms of correlation lengths is provided . </S><S> finally , we find the crossover between the lr and short - range regimes to occur inside a narrow window @xmath5 , thus lending strong support to sak s scenario . </S>

0810.5025

supernova 1993j ( ra = 09h55 m 24.7747s , @xmath2 = + 69@xmath3 01 13.7031 ) was discovered on 28 march 1993 in the nearby spiral galaxy m81 ( ripero et al . , 1993 ) and became the brightest radio supernova ever observed .its brightness and high declination have allowed this object to be studied in great detail at radio wavelengths .a shell - like radio structure associated with sn1993j was resolved using vlbi techniques as early as 239 days after the supernova explosion ( marcaide et al . , 1995 ) .an extensive vlbi monitoring program then followed which tracked the expansion of the shell over seven years at three frequencies ( bietenholz , bartel & rupen , 2001 ; bartel et al ., 2002 ; bietenholz , bartel & rupen , 2003 ) .these observations yielded a wealth of information about the supernova , including accurate determination of the explosion centre , measurement of the deceleration of the ejecta and radio light curves and spectra .images of the resolved shell allowed direct monitoring of its development , from its initial barely resolved state , through expansion into partial shell structures , to the complex structure apparently present in the final epochs 3000 + days later .behaviour such as apparent rotation of features in the shell in the plane of the sky are seen between epochs of vlbi observations , as are phenomena such as coherent structures dividing into a series of peaks which apparently change in azimuthal position as the remnant expands .bietenholz , bartel and rupen ( 2003 ) discuss possible physical mechanisms which could explain the structure in the shell , including rayleigh - taylor cooling instabilities , evolving magnetic field structure and interactions with inhomogeneous circumstellar material .this paper presents a new investigation of this intriguing evolution of the brightness distribution in the expanding shell .we explore whether the development of apparently complex structure may be due to incomplete sampling of the @xmath0 plane , which is problematic especially for sparse interferometer arrays such as vlbi networks .similar effects had also been a cause for concern when interpreting 6-cm merlin images of nova cassiopeiae 1995 ( heywood et al .2005 ; heywood & obrien , 2007 ) where the apparently complex development of the radio emission could be accounted for by a spherically symmetric shell model and sparse sampling of the @xmath0-plane . with this in mind we simulated 8.4-ghz vlbi observations of the radio shell of sn1993j at nine epochs , for comparison with the genuine vlbi images .further discussion regarding the non - uniqueness of deconvolved radio images ( particularly due to the clean deconvolution method ) may be found in e.g. cornwell , braun and briggs ( 1999 ) and briggs ( 1995 ) .the latter contains a particularly relevant case study of the radio shell of sn1987a whereby different deconvolution methods yield subtle variations in apparent structure .creation of the simulated vlbi images requires two components .first , a model of the assumed source brightness distribution is required .this model should represent how an ideal instrument would image the source at a given frequency .the second component of the simulation is a @xmath0-plane sampling function which accurately represents that of the real instrument .a synchrotron shell model is used as the basis for our simulated radio images .the expansion of the shell in this model is not self - similar because the radial density profile is time dependent as the physical shell boundary expands , and reverse shocks propagate back through the ejecta .the model is fully spherically symmetric however , and the density variations are purely radial with no azimuthal component .a detailed description of this model is provided by mioduszewski et al .simulated radio images were used with ages that closely matched the epochs of the vlbi observations , and were evenly spaced throughout the evolution of sn1993j .mioduszewski et al . (2001 ) also parametrized and fitted their models to the radio observations to give consistent angular sizes and flux density values , and kindly allowed us to use their models in the investigations presented here .figure [ fig : inputmodels ] shows three epochs of the simulated radio images and their corresponding flux density profiles . ( 16,11.0 ) ( -0.5,0.0 ) in addition to the model images being chosen to coincide as best as possible with the genuine vlbi observations the model array for each epoch is defined to be the same as those used during the sn1993j vlbi campaign .only stations which were used during an actual observation are included for the corresponding simulation .the aips task uvcon is used to generate the model @xmath0-data sets . by examining vlba archive files ,the start and end time for any observation are used to determine the total duration . using the relevant model image and an antenna file describing the stations present for a specific observation ,uvcon is executed , resulting in a track length equivalent to this total duration . in the case of the real observation, this track would encompass all sources and frequencies . in order to reduce the @xmath0-data set such that it only includes visibilities related to sn1993j at 8.4 ghzthree flag tables are applied , documenting the frequency switching information , the sn1993j scan times , and the automated flagging that occurred at the receivers .these flag tables are constructed from archival data files , and converted to a format that can be directly read by the uvflg task in aips .bad data that are manually removed from the true vlbi observations in the pre - calibration stage can not be accounted for as the simulations are not prone to problems such as radio frequency interference .this means that in each case the simulations are over - estimating the @xmath0-plane sampling , and that our analysis errs on the side of being conservative .multi - frequency interferometric observations often employ a technique to improve @xmath0-plane coverage , whereby the observing frequency is switched periodically .consider an 18-hour track at three frequencies .rather than observing at each frequency in three successive six - hour blocks the observing frequency is switched , for example every 20 minutes .the visibilities for a given frequency are therefore more spread out on the @xmath0-plane , facilitating the sampling of more diverse fourier components and reducing linear biases .determining this frequency switching information allows us to flag out scans which do not occur at 8.4 ghz .once the simulated @xmath0-tracks are consistent only with the scans at 8.4 ghz we flag out times where the calibration sources are being observed . from the archive data we can obtain the times whensn1993j was being observed and use this information to retain only these time ranges in the simulation .aips - readable flag tables describing data that were flagged at the time of the observation ( e.g. due to slewing or pointing errors , source changes etc . ) are also available , and this is the final flag table applied to the simulated data set . note also that a 15@xmath3 elevation cut off for each antenna is also imposed on the simulation .data are routinely flagged below this elevation because system temperatures rise as the antenna points closer to the horizon , and local geographical features can often obscure the line of sight . for a source at the declination of sn1993jthis condition only slightly affects the longest baselines in the vlba network .the uvcon task also adds random gaussian noise to the simulated data set , the level of which is governed by the parameters of each antenna in the array , such as the antenna efficiency and system temperature .the simulated @xmath0-data sets were imaged using the clean algorithm ( clark , 1980 ) using the aips task imagr . for each simulationthe circular restoring beam used is consistent with that of the nearest epoch of the vlbi observations , as reported by bietenholz , bartel and rupen ( 2003 ) .briggs ( 1995 ) weighting was utilised with the robust parameter in imagr set to zero .for the nine simulated maps presented in figure [ fig : ninemaps ] the data were reweighted by the square roots of the data weights using the uvwtfn parameter .this is a common technique for vlbi observations and is consistent with the imaging methods of bietenholz , bartel and rupen ( 2003 ) .deconvolution was carried out interactively and the clean cycle terminated when it was judged that all substantial flux has been removed . since no antenna - based or baseline - based amplitude or phase errorsare included in our simulations , this means no self - calibration procedures will be responsible for spuriously enhancing features in our simulations .our simulations will only underestimate the extent to which small compact features could become _ reinforced _ in self - calibration and cleaning cycles .contouring of the simulated images is also carried out in a way that is consistent with the images presented by bietenholz , bartel and rupen ( 2003 ) .the greyscale in each image runs from 8% to 100% of the peak brightness for epochs below 1253 days and from 16% to 100% for epochs above .contour levels are 1% , 2% , 4% , 8% , 16% , 32% , 45% , 64% and 90% of the peak brightness in each case , starting with the first contour level that is brighter than three times the root mean square background brightness levels ( 3@xmath1 ) .presenting the simulated images in a consistent way better facilitates comparison with the real observations .in this section we first present the nine simulated radio images which best recreate the genuine vlbi observations of sn1993j .we discuss aspects of the real vlbi observations which differ from our simulations .also presented in this section are residual images created by differencing the simulated radio image from a convolved input model .figure [ fig : ninemaps ] shows the resulting images from nine simulated vlbi observations of sn1993j .a summary of the relevant image parameters can be found in table [ tab : images ] .the total shell flux density values for the simulated vlbi maps are measured by using the aips task tvstat to define the region of emission . [ cols= " < ,< , < , < , < , < , < " , ]the first part of this section presents simplified simulated observations to demonstrate the coupling between the @xmath0-plane sampling and the resulting image .the second subsection takes a more realistic simulation and shows how subtle changes to the visibility data ( e.g. antenna flagging ) affect the image . as a simplified illustration of the effects of incomplete @xmath0-sampling on azimuthal structure , figure [ fig : sixhours ] shows a simulated continuous six - hour vlbi simulation for seven different hour - angle ranges . note thatthe tracks are truly continuous ; there are no gaps in the visibilities .these simulations are constructed with an antenna configuration and input model both consistent with those of day 2787 of the actual vlbi observations of sn1993j .the left hand column shows the @xmath0-coverage and the right hand column shows the resulting map .the most notable feature in this series of images is that as the @xmath0-coverage rotates with hour angle , _ so does the structure of the corresponding image_. the position of the peaks on the ring rotate , as does the central ` bar ' in the image .note that the input model is purely circularly symmetric with no angular structure variations ( see figure [ fig : inputmodels ] ) , therefore _none of the complex structure in the final images is real _ in order to quantify the azimuthal structure that is imparted to an image , and demonstrate how it is radically affected by the @xmath0-coverage , the pixel values as a function of angle are presented in figure [ fig : greyscales ] . this simulation is carried out 24 times with the hour - angle coverage drifting by 30 minutes each time , using day 2787 as the input model .the resulting images are 512 @xmath4 512 pixels in size , and the pixel values are measured around a ring with a radius of 141 pixels with respect to the image centre , which on these images corresponds to 3.8 milliarcseconds .this value was chosen as it corresponds to the radius of the brightest annulus of emission on the input model .( 16,11.9 ) ( -1.0,-0.6 ) plotting the brightness distribution in this way allows one to clearly see how the structure rotates with the @xmath0 coverage .the brightest peaks in the image appear as dark patches and the rotation manifests itself here as the diagonal striping in figure [ fig : greyscales ] . to illustrate how apparently subtle changes to the visibility datacan alter the structure in the final image , figure [ fig : uv_decay ] presents @xmath0 plots and corresponding images for three different simulated observations .these are created using the simulation corresponding to day 2525 .contour and greyscale levels , and the restoring beam , are consistent with the day 2525 simulation presented in figure [ fig : ninemaps ] .the left hand pair shows the simulation with the 15 degree elevation cut off applied . for a source at the declination of sn1993j the effect this cut offhas on the @xmath0 coverage compared to that of a zero degree elevation cut off is minimal ; in our simulation it reduced the number of visibilities by less than 3% .the effect this elevation cut off has on the plot of @xmath0-plane coverage is subtle and difficult to notice by eye .as such the complete plot is not presented here . to illustrate potential problems that may be caused by local geographical features the centre pair of images simulation has an asymmetric ( + 35@xmath3/+15@xmath3 ) elevation cut off imposed on the mauna kea station .this drastically affects the longest baselines and consequently alters the structure in the corresponding image . for the final image the st .croix station is entirely removed , which mainly affects the intermediate length baselines as shown by the two prominent gaps in the @xmath0 plot .further simulations can be found in appendix a where we present a range of simulated images corresponding to the day 2787 epoch for a range of declinations , track lengths and hour angle ranges .( 16,9.9 ) ( 0.6,-0.3 )we have demonstrated in this paper how sparse @xmath0 coverage can impart apparently complex azimuthal structure to a radio brightness distribution that is in reality morphologically simple .rotation of the @xmath0-plane sampling pattern determined by varying hour angle coverage leads to a corresponding rotation in the radio image .this is just one example of the kind of structures that are vulnerable to errors arising from sparse @xmath0-plane sampling . for a full discussion of other structures and their vulnerabilitieswe refer the reader to the works by cornwell ( 1999 ) and briggs ( 1995 ) who investigate the use of different deconvolution methods which further undermine these structures and can not completely overcome the problems caused by incomplete information . by reconstructing the @xmath0-plane distributions achieved in the sn1993j vlbi campaign and convolving these with a perfect spherically symmetric model of the supernova ,we have shown that sparse @xmath0 coverage may be wholly or partially responsible for the evolving complex azimuthal structure observed in real supernova remnant .the authors wish to thank amy mioduszewski for kindly allowing us to use her synchrotron models .we thank the referees for their comments , and for highlighting some potential issues with uvcon .we also wish to thank leonia kogan and eric greisen of nrao for rapidly addressing these issues .ih thanks the science and technology facilities council .kmb thanks the royal society for a university research fellowship .ih and kmb are very grateful to the leverhulme trust whose prize supported this work .hrk thanks the university of oxford for financial support .bartel , n. , et al .2002 , , 581 , 404 bietenholz , m. f. , bartel , n. , & rupen , m. p. 2001, , 557 , 770 bietenholz , m. f. , bartel , n. , & rupen , m. p. 2003, , 597 , 374 briggs , d. s. 1995 , bulletin of the american astronomical society , 27 , 1444 briggs , d. s. 1995 , `` high fidelity deconvolution of moderately resolved sources '' , phd thesis , new mexico institute of mining and technology clark , b. g. 1980 , , 89 , 377 cornwell , t. , braun , r. , & briggs , d. s. 1999 , synthesis imaging in radio astronomy ii , 180 , 151 heywood , i. , obrien , t. j. , eyres , s. p. s. , bode , m. f. , & davis , r. j. 2005 , , 362 , 469 heywood , i. , & obrien , t. j. 2007 , , 379 , 1453 marcaide , j. m. , et al .1995 , , 373 , 44 mioduszewski , a. j. , dwarkadas , v. v. , & ball , l. 2001 , , 562 , 869 ripero , j. , et al .1993 , , 5731 , 1the parameter space defined by factors which will affect image fidelity is enormous .both the realistic and illustrative simulations presented in this paper represent a very specific case in terms of source declination , track length and therefore @xmath0-plane coverage . in this sectionwe present simulations of the day 2846 model at declinations of 20 , 40 , 60 and 80 degrees with 6 , 9 and 12 hour continuous observations .please refer to figure [ fig : inputmodels ] to see the morphology of the input source .seven unique hour - angle ranges are simulated for each declination - track length pair resulting in 84 unique observations .simulated images are grouped by declination and are presented in figures [ fig : d20 ] , [ fig : d40 ] , [ fig : d60 ] and [ fig : d80 ] .the greyscale range in each image is from 16% to 100% of the peak value and the contours are 32% , 45% , 64% and 90% of the peak flux .the starting value of the hour - angle range is listed on the left hand side of the figures , with the end hour angle being the start value plus the track length , which is printed at the top of each page .imaging of the simulated data sets was automated in this case .cleaning was not carried out interactively and the clean cycle was terminated after reaching 1000 components .the gaussian fitted to the half power point of the central lobe of the point spread function for each image was used as the restoring beam in each case .this is plotted in the lower left of each figure .the favourable declination ( + 69 degrees ) of sn1993j facilitates the use of a circular restoring beam as the fitted beam itself is nearly circular . for lower declinationsthis is not applicable , as can be easily seen by the elongated beams of lower declination simulations .note the wide variety of image artifacts that can arise due to the @xmath0-plane sampling .very sparse coverage at low declinations leads to ` polar cap'-type features , striped gaps in the @xmath0-coverage result in dual ridge features , and central bars across the shell are a regular outcome .only the 12-hour tracks at high declination , featuring almost complete coverage , result in a ring structure with a central minimum .artificial azimuthal structure is present in each of these simulated images , even those with excellent @xmath0-coverage .

<S> we present simulated images of supernova 1993j at 8.4 ghz using very long baseline interferometry ( vlbi ) techniques . a spherically symmetric source model is convolved with realistic @xmath0-plane distributions , together with standard imaging procedures , to assess the extent of instrumental effects on the recovered brightness distribution . </S><S> in order to facilitate direct comparisons between the simulations and published vlbi images of sn1993j , the observed @xmath0-coverage is determined from actual vlbi observations made in the years following its discovery . </S><S> the underlying source model only exhibits radial variation in its density profile , with no azimuthal dependence and , even though this model is morphologically simple , the simulated vlbi observations qualitatively reproduce many of the azimuthal features of the reported vlbi observations , such as appearance and evolution of complex azimuthal structure and apparent rotation of the shell . </S><S> we demonstrate that such features are inexorably coupled to the @xmath0-plane sampling . </S><S> the brightness contrast between the peaks and the surrounding shell material are not as prominent in the simulations ( which of course assume no antenna- or baseline - based amplitude or phase errors , meaning no self - calibration procedures will have incorporated any such features in models ) . </S><S> it is conclusive that incomplete @xmath0-plane sampling has a drastic effect on the final images for observations of this nature . </S><S> difference imaging reveals residual emission up to the 8@xmath1 level . </S><S> extreme care should be taken when using interferometric observations to directly infer the structure of objects such as supernovae . </S><S> = = = = = = = = # 1 # 1 # 1 # 1 @mathgroup@group @mathgroup@normal@groupeurmn @mathgroup@bold@groupeurbn @mathgroup@group @mathgroup@normal@groupmsamn @mathgroup@bold@groupmsamn = `` 019 = ' ' 016 = `` 040 = ' ' 336 = " 33e = = = = = = = = # 1 # 1 # 1 # 1 = = = = = = = = # 1 latexl-.36em.3ex-.15em t-.1667em.7ex-.125emx [ section ] [ firstpage ] stars : supernovae supernovae : individual ( sn1993j ) radio continuum : stars techniques : interferometry </S>

cond-mat0207122

simple models of dna thermal denaturation have attracted a lot of attention for a long time @xcite . although it was pointed out rather early that excluded volume effects may be crucial in determining the order of the denaturation transition @xcite , it was only very recently that such effects were taken into account both at analytical and numerical level @xcite .even once accepted excluded volume effects as a key factor in determining the nature of the transition , one can still be interested in understanding how the relaxation of self - avoidance constraints influences the overall behavior of the system . as this can be done in different ways , learning in detail about the effects of different geometrical exclusion constraints can help in conceiving better models or more efficient approximations .the main general aim of this paper is that of producing a contribution to the debate on these general issues . in particular we analyze the model of dna denaturation recently introduced and studied by garel , monthus and orland ( gmo ) @xcite . among other things ,we show that this model is a rather ideal context in which a recently proposed representation of dna in terms of block copolymer networks @xcite can be applied .the hamiltonian for the gmo model is @xcite : @xmath2 where @xmath3 , @xmath4 describe the conformations of the two strands as functions of the curvilinear coordinate @xmath5 and @xmath6 is the attractive interaction between homologous base pairs ( @xmath7 ) . the last term on the r.h.s . of eq .( [ hamin ] ) is due to excluded volume interactions , which act only between the two chains and not within a single chain .thus the self - avoidance constraint is relaxed and each isolated strand follows a simple random walk ( rw ) statistics .the authors of ref .@xcite simplified further the hamiltonian ( [ hamin ] ) by approximating the excluded volume term by a long range interaction between the strands , which allows for an analytical solution of the problem . within such approximation , the denaturation transition turns out to be of first order type @xcite , as in models where excluded volume effects are fully incorporated @xcite .our purpose here is to clarify further this point .we start from a description _ la _ poland - sheraga ( ps ) @xcite of the gmo model and use a series of exact results on reunion exponents of mutually avoiding walks , which show indeed that the denaturation transition is first order .more surprisingly , the transition turns out to be even sharper than that occurring in models where excluded volume effects are fully taken into account .this conclusion is corroborated by approximate analytical treatments of excluded volume effects due to loop - loop interactions and by a series of numerical results obtained for polymer lattice models in @xmath8 and @xmath9 .0.2truecm in the poland - sheraga ( ps ) model @xcite the dna chain is treated approximately as a sequence of non - interacting bound double segments and denaturated loops . within this scheme the full partition function @xmath10 for a chain whose constituent strands have lengths @xmath11 , is factorized in terms of elementary partition functions for loops and segments ( see fig .[ fig01 ] ) . in the grand canonical ensemble , where a fugacity @xmath12 per lattice step is assigned, the total partition function reads @xcite : @xmath13 where @xmath14 and @xmath15 are the grand canonical partition functions for loops and double segments , respectively .the form of the numerator @xmath16 depends on boundary conditions and is not interesting for our purposes .the crucial quantity is the loop partition function , which in the most general form reads @xcite : @xmath17 where @xmath18 is the ( non - universal ) effective connectivity constant , while @xmath1 is a universal exponent .the order of the transition is related to the value of @xmath1 @xcite : at the transition temperature ( which in the ps model corresponds to @xmath19 ) loops are power - law distributed in size with probability @xmath20 , thus their average size @xmath21 is finite for @xmath0 , implying a first order denaturation , while it diverges for @xmath22 and the transition becomes continuous . for self- and mutually avoiding strands @xmath1was estimated analytically in an extended ps scheme @xcite , where excluded volume effects between a self - avoiding loop and the rest of the chain are approximately taken into account .this exponent was also obtained directly by monte carlo simulations @xcite for polymer lattice models .the agreement between the analytical estimates and simulation results for @xmath1 is very good both in @xmath8 and @xmath9 @xcite . in the next sections we will present analytical and numerical estimates of the exponent @xmath1 for the gmo model .the former are based on the general theory of copolymer networks , which is briefly reviewed in the appendix . 0.2truecma denaturated loop in the gmo model is formed by two mutually avoiding random walks of equal length @xmath23 , with common origin and endpoint ( see example of fig .[ fig02 ] ) .the probability that two random walks , with common origin , meet for the first time after @xmath23 steps decays asymptotically as @xmath20 with @xcite : @xmath24 here @xmath25 is the dimensionality of the embedding space , @xmath26 for rws and @xmath27 denotes the contribution of a vertex from which the two walks originate ( see fig . [ fig02 ] ) .we follow the notation of ref .@xcite and indicate with @xmath28 the exponent associated to a vertex from which @xmath29 solid and @xmath30 dashed lines depart , at the gaussian fixed point , which is the relevant one for our purposes . at this fixed pointsolid and dashed lines do not interact with themselves , but they avoid the lines of the other group .figure [ fig02 ] , as well as figs .[ fig03]-[fig05 ] below , are examples of so called _ copolymer networks _ , i.e. networks whose constituents are polymers of different species .the relevance of these block copolymer networks for a description of denaturating dna has been already stressed in the context of models with full excluded volume effects @xcite .a general network of this type may be made , for instance , by an arbitrary mixture of random and self - avoiding walks ( saws ) .any pair of these walks may either avoid , or be allowed to cross each other . as for the homopolymer case ,the general entropic exponent for a network with arbitrary topology follows from those of the constituents vertex exponents ( see refs .@xcite for details ) . the general rule to calculatethe entropic exponent of a network is simple and it is reviewed in the appendix .restricting to the case in which all constituent polymers have the same @xmath31 exponent , one associates to each independent loop a factor @xmath32 and to each vertex a factor @xmath33 , with @xmath34 the appropriate exponent which depends on the number of outgoing legs , and on their type .the generalization to an arbitrary mixture of polymer segments with distinct @xmath31 s , as for instance rws and saws , is also possible .coming back to eq .( [ cisolated ] ) , the vertex exponent for one solid and one dashed line is known exactly in @xmath8 @xcite .its value is @xmath35 , which leads to @xmath36 . in higher dimensions one has to resort to @xmath37 expansion renormalization group results . in @xmath9 resummation techniquesyield @xmath38 @xcite , which implies @xmath39 .thus , both in @xmath40 and @xmath41 the reunion exponent is larger than the threshold value ( @xmath0 ) , which immediately implies a first order denaturation for the gmo model , within the ps picture of non - interacting loops and bound segments . as a comparison we recall that for an isolated loop made by two self- and mutually avoiding strands the entropic exponent is @xcite : @xmath42 for a saw one has @xmath43 in @xmath8 , which implies @xmath44 , while @xmath45 in @xmath9 , implying @xmath46 .thus , within the ps picture , denaturation is a continuous transition ( @xmath47 ) when full self- avoidance within a single loop is included @xcite .it is also possible to go beyond the ps approximation in the gmo model , but it is already quite clear at this point that excluded volume effects between a single loop and the rest of the chain `` localize '' even more the loops and @xmath1 can only increase .this can be shown explicitly for some simple geometries in which the loop is embedded .the calculations follow closely those of ref .@xcite for the model in which self - avoidance is fully included and are also briefly reviewed in the appendix .we give here only the final results .0.2truecm we consider the geometry shown in fig .[ fig03 ] in which a loop of length @xmath48 is connected to two loops of length @xmath49 each . in the limit @xmath50 the partition functions of the inner loop is still of the type of eq .( [ loop ] ) and factorizes with the partition function of the rest of the chain @xcite . as explained in appendix , the exponent @xmath1 is however modified : @xmath51 0.2truecm in @xmath8 some care has to be taken in considering vertices , as the order of walks does matter .this is illustrated in fig .[ fig04 ] where the two possible distinct types of vertices with four outgoing lines , two solid and two dashed , are shown . for a vertex of type ( a ) more configurationsare possible as neighboring lines of the same type are allowed to overlap . in ( b ) , instead , the distinction between solid and dashed lines becomes irrelevant as neighboring lines avoid each other . in the latter case the lines form a vertex of four mutually avoiding walks ( maws ) @xcite .a vertex with @xmath52outgoing maws , has an associated entropic exponent @xcite : @xmath53 hence , for loops embedded in a geometry as in fig .[ fig03 ] , and formed by vertices of type ( b ) one finds @xmath54 .let us consider now vertices of type ( a ) for which we keep the notation @xmath55 to indicate the associated exponent .using the prescriptions of ref .@xcite one has @xmath56 and thus from eq .( [ cinteracting ] ) @xmath57 .notice that stronger excluded volume effects ( as for vertices of type ( b ) ) imply a higher value for @xmath1 , with a quite large difference between the two estimated values .we point out that , obviously , @xmath58 in all dimensions . in @xmath9the above subtleties do not occur .the vertex exponent @xmath59 was estimated in ref .@xcite using @xmath37 expansions and resummation techniques .thus , from eq .( [ cinteracting ] ) , one obtains @xmath60 .we consider now another geometry , where a loop is connected to two long double stranded segments , as illustrated in fig .[ fig05 ] . using the same method of ref .@xcite we find ( see appendix ) : @xmath61 where with @xmath62 we indicate the exponent associated to a vertex where one double , and two single stranded segments join .notice that the double stranded segment is actually a self avoiding walk since , if it would overlap itself in some part , the mutual avoidance condition between the two constituent strands would be violated .0.2truecm again in @xmath8 one can use the techniques of ref .@xcite to find @xmath63 which leads to @xmath64 .to our knowledge , there exists no field theoretical analysis for this type of vertices in higher dimensions , thus we are not able to make any predictions for @xmath1 in @xmath9 , at present .table [ table01 ] summarizes the exponents found in the gmo model for an isolated loop ( @xmath65 ) , for a loop interacting with other loops ( @xmath66 ) , as in the geometry of fig .[ fig03 ] , and for a loop interacting with double stranded segments ( @xmath67 ) , as in fig .[ fig05 ] .for the two dimensional case we reported both values of @xmath66 corresponding to the loops formed by the two types of vertices of fig .[ fig04 ] .we notice that , as anticipated above , the interaction with other parts of the chain has the effect of increasing the loop exponent @xmath68 as is the case for all other models studied so far @xcite .the first order character of the transition in the gmo model is therefore strengthened .it is interesting to compare the exponents obtained in this paper with the corresponding ones for the model in which strands are fully self - avoiding ( fsa ) .the latter , taken from ref .@xcite , are given in table [ table01 ] .the comparison between the two models indicates that the transition in the gmo model is generally sharper ( higher @xmath1 ) .this is always true with the exception of the exponent @xmath69 in @xmath8 for vertices of the type of fig .[ fig04](a ) for which we find @xmath70 , for the gmo model , while @xmath71 in the fsa case .this is due to the fact that in the gmo model for the type of vertex considered excluded volume effects are not so pronounced as walks have considerable freedom to overlap .we performed a series of numerical calculations for a lattice version of the gmo model both in @xmath9 and @xmath8 ( cubic and square lattices ) .we considered two random walks of length @xmath11 , described by the vectors @xmath72 and @xmath73 ( @xmath74 ) which identify the positions of the lattice monomers .the walks have common origin ( @xmath75 ) , and no specific boundary condition is imposed on the other ends . while each strand can overlap itself , mutual avoidance requires that @xmath76 for @xmath77 .overlaps at homologous points are however allowed ( @xmath78 ) ; these correspond to a bound state of complementary base pairs , to which we assign an energy @xmath79 . at sufficiently low temperature ( @xmath80 )the walks are fully bound and form an object obeying self - avoiding walk statistics .we used the perm @xcite algorithm and computed @xmath81 the probability distribution function ( pdf ) to find a loop of length @xmath48 in the chain .while at low temperatures loops are exponentially distributed in size , at the critical point the pdf has an algebraic decay @xmath20 , from which we can extract the exponent @xmath1 @xcite .figure [ fig06 ] shows a log - log plot @xmath81 vs. @xmath23 in @xmath9 at the estimated critical point ( @xmath82 ) , for chains of lengths up to @xmath83 .a linear fit of the data yields @xmath84 , not far from the @xmath69 given by eq .( [ cinteracting ] ) ( see table [ table01 ] ) .the result confirms that the transition in the gmo model is sharper than when self - avoidance is fully included .indeed , in the latter case the most accurate numerical estimate of the loop exponent is @xmath85 @xcite .0.2truecm we consider next @xmath86 , the probability that the distance between homologous points in the strands @xmath87 equals @xmath88 .this quantity was found to decay as a stretched exponential for @xmath89 @xcite , from an analysis of the hamiltonian of eq .( [ hamin ] ) .this conclusion is at odd with the algebraic decay of the loop pdf found here for the gmo model , and also in other similar models with a first order transition @xcite . for the numerical calculation of @xmath90 we registered all distances between homologous pairs along the chains @xmath87 , starting from @xmath91 and up to the highest @xmath92 for which the two strands are in contact ( @xmath93 ) .figure [ fig07 ] shows a plot of @xmath94 vs. @xmath95 at the transition temperature in @xmath96 .we can not fit the data with a stretched exponential decay , but by increasing the system size the data approach a power law decay @xmath97 with @xmath98 .the fact that @xmath99 implies that the probability is normalizable .this is a confirmation that the transition is of first order type ( see ref .notice however , that the algebraic nature of @xmath90 implies infinite moments @xmath100 starting from @xmath101 .we conclude that the exponential decay found in ref .@xcite is an artifact of the approximation introduced .we performed a series of calculations also in the two dimensional case . at the estimated critical temperature ( @xmath102 ) we find , from the decay of the probability distributions @xmath81 and @xmath90 , the estimates @xmath103 and @xmath104 , confirming the first order nature of the transition .again denaturation happens to be sharper than in the case where self - avoidance is fully incorporated , for which @xmath105 @xcite . in the perm algorithmboth types of vertices of fig .[ fig04 ] are generated , thus @xmath1 can not be directly compared with any of the value obtained analytically in the previous section . 0.2truecmin this paper we studied the model for dna denaturation recently introduced by garel , monthus and orland @xcite .our analysis is complementary to that of ref . @xcite and it is based on the calculation of the exponent @xmath1 describing the algebraic decay of the probability of denaturated loop lengths at the critical point . using exact results from mutually avoidingwalks statistics we find that @xmath0 already for isolated loops , implying that the transition is of first order type .this conclusion is in agreement with the results of ref .@xcite , but it is not based on the approximate treatments @xcite . for some specific simple geometries in which a loop is embedded in the chain we provide analytical estimates of @xmath1 , following ideas from the theory of polymer networks @xcite .these calculations show explicitly that @xmath1 increases with respect to the isolated loop value , i.e. the transition robustly maintains its first order character .more surprising is , at first sight , the fact that the denaturation in the gmo model is sharper than that occurring in models where excluded volume constraints are fully incorporated ( higher @xmath1 ) , as indicated by our analytical and numerical estimates of @xmath1 both in @xmath8 and @xmath41 .this behavior can be explained by taking into account that the self - avoidance of the single strands somehow contrasts the excluded volume effects between each loop and the rest of the molecule .the result is that , when the single strand self - avoidance is relaxed , the loop sizes are further reduced , consistently with an increase of @xmath1 . it is somehow remarkable how results of the theory of block copolymer networks , which are relevant to issues like that of establishing a link between field theory and multifractal measures ( see e.g. ref .@xcite ) , allow to draw reliable predictions on models of the denaturation transition of dna .there is another possibility of relaxing the self - avoidance which is somehow complementary to that of the gmo model .one can indeed neglect the mutual avoidance between the strands , while keeping self - avoidance within each stand .this possibility was studied in @xmath9 in ref . @xcite and the denaturation transition turns out to be continuous ( @xmath47 ) .thus , relaxing the self - avoidance constraint in two different ways within the single loop , results in a softening ( as in the latter example ) or a sharpening ( as in the gmo model ) of the transition .finally , we also considered the probability distribution of the distances between homologous monomers of the two strands .we found that this quantity decays as a power law at the transition point , in disagreement with the stretched exponential behavior found in ref .@xcite , which is probably an artifact of the approximation introduced there .the algebraic decay of various quantities at denaturation , even in the case of first order transitions , is a common feature of other similar models . finding the appropriate treatment of the hamiltonian ( [ hamin ] ) , which provides power - laws decay at the transition point is still an open question .useful discussions with david mukamel are gratefully aknowledged . 0.2truecmwe review here some known results for entropic exponents of networks of arbitrary topology .we focus mainly on networks whose constituent polymers are all of the same type , and just briefly mention copolymer networks .let us consider a network made of self - avoiding walks , all avoiding each other , as the one shown in fig .[ fig08](a ) .let @xmath106 be the total length of the network , where each segment has length @xmath107 . in the asymptotic limit @xmath108 , @xmath109 , the total number of configurations for the network scales as @xcite : @xmath110 where the factor @xmath18 is the effective connectivity constant for the walks ( see e.g. ref @xcite ) and @xmath111 is a scaling function .the universal exponent @xmath112 depends only on the number of independent loops , @xmath113 , and the number of vertices with @xmath92 outgoing segments @xmath114 as @xcite @xmath115 where @xmath116 are exponents associated to a vertex with @xmath92 outgoing legs .such exponents are known exactly in @xmath8 from conformal invariance @xcite , while they have been obtained from renormalization group and resummation techniques in @xmath9 @xcite .we notice that it is also customary to use the definition @xmath117 .this theory was recently applied to the denaturation problem @xcite . in this case oneconsiders network geometries as that of fig .[ fig08](b ) , in which a loop of total length @xmath48 is connected to two polymers of lengths @xmath118 each . following the above notation one has @xmath119 , @xmath120 and @xmath121 and thus the total number of configurations scale as : @xmath122 with @xmath123 .now in the limit @xmath124 one should recover the partition function for a self - avoiding walk of total length @xmath11 which has an entropic exponent @xmath125 .this requires that : @xmath126 in the limit @xmath127 .thus , loops of length @xmath23 embedded within a long chain of length @xmath11 ( @xmath128 ) are distributed according to the probability @xmath129 , with @xcite : @xmath130 as second example we consider a loop of length @xmath48 embedded between two long loops of total length @xmath131 , as depicted in fig . [ fig08](b ) . the total number of configurations is still given by eq .( [ eq11 ] ) with @xmath132 , as the network has now @xmath133 independent loops and two vertices with four outgoing legs , i.e. @xmath134 . in the limit @xmath124one should find the partition function of two loops joined at a single vertex .matching the scaling function in this limit yields @xmath135 for @xmath127 .one finds thus that loops are distributed according to the probability @xmath20 , with @xcite : @xmath136 finally , this theory can be generalized to the case of networks with more than one type of polymer , as for _ copolymer _ networks .( [ gammag ] ) can be generalized easily in the case that the constituting polymers have all the same metric exponents ( for instance in the case of mutually avoiding random walks ) .the only difference is that now vertex exponents will not depend only on the number of outgoing legs , but also on the type of polymers forming the vertex and , in two dimensions , also on the order .for instance , for a network made of two types of polymers , which are random walks and mutually avoiding with a geometry of fig .[ fig03 ] one arrives to a @xmath1 exponent in the limit of an inner loop much shorter than the rest of the chain as given by eq .( [ cinteracting ] ) , which is the copolymer network counterpart of eq .( [ sigma4 ] ) .notice that while for homopolymers @xmath137 , in general a vertex with two outgoing polymers of different type will have a non - vanishing exponent ( as is the case of the @xmath27 introduced in fig .[ fig02 ] ) . finally if polymers constituting the inhomogeneous network will have different @xmath31 exponent , as for networks formed by a mixture of random and self - avoiding walks than eq .( [ partfunc ] ) is still valid , but in a form involving the radii of gyrations @xmath138 of the constituent polymers , where @xmath139 , for a segment of length @xmath107 .d. poland and h. a. sheraga , j. chem . phys .* 45 * , 1456 ( 1966 ) ; * 45 * , 1464 ( 1966 ) .m. e. fisher , j. chem .* 45 * , 1469 ( 1966 ) .m. e. fisher , j. stat .* 34 * , 667 ( 1984 ) .m. peyrard and a. r. bishop , , 2755 ( 1989 ) .m. s. causo , b. coluzzi , and p. grassberger , , 3958 ( 2000 ) .y. kafri , d. mukamel , and l. peliti , , 4988 ( 2000 ) ; y. kafri , d. mukamel , and l. peliti , eur . phys. j. b * 27 * , 132 ( 2002 ) .e. carlon , e. orlandini , and a. l. stella , , 198101 ( 2002 ) .m. baiesi , e. carlon , and a. l. stella , , in press ( cond - mat/0205125 ) .t. garel , c. monthus , and h. orland , europhys . lett . *55 * , 132 ( 2001 ) .b. duplantier and k .- h .kwon , , 2514 ( 1988 ) .b. duplantier , commun .phys . * 117 * , 279 ( 1988 ) . c. von ferber and yu .holovatch , europhys .lett . * 39 * , 31 ( 1997 ) ; c. von ferber and yu .holovatch , , 6370 ( 1997 ) .b. duplantier , , 941 ( 1986 ) .notice that denaturation becomes first order for @xmath140 , since @xmath141 and thus @xmath142 .p. g. de gennes , _ scaling concepts in polymer physics _ ( cornell university press , ithaca , ny , 1979 ) . c. von ferber and yu .holovatch , , 042801 ( 2002 ) .b. duplantier , , 880 ( 1999 ) .p. grassberger , , 3682 ( 1997 ) .the approximate treatment of the excluded volume interaction as done in ref .@xcite has been the subject of recent debate .see : s. m. battacharjee , europhys .lett . * 57 * , 772 ( 2002 ) ; t. garel , c. monthus , and h. orland , europhys . lett . * 57 * , 774 ( 2002 ) . c. vanderzande , _ lattice models of polymers _( cambridge university press , cambridge , united kingdom , 1998 ) .l. schfer , c. von ferber , u. lehr , and b. duplantier , nucl .phys . * b374 * , 473 ( 1992 ) .

<S> recently garel , monthus and orland ( _ europhys . lett . _ * 55 * , 132 ( 2001 ) ) considered a model of dna denaturation in which excluded volume effects within each strand are neglected , while mutual avoidance is included . using an approximate scheme they found a first order denaturation . </S><S> we show that a first order transition for this model follows from exact results for the statistics of two mutually avoiding random walks , whose reunion exponent is @xmath0 , both in two and three dimensions . </S><S> analytical estimates of @xmath1 due to the interactions with other denaturated loops , as well as numerical calculations , indicate that the transition is even sharper than in models where excluded volume effects are fully incorporated . </S><S> the probability distribution of distances between homologous base pairs decays as a power law at the transition . 2 </S>

nucl-th9904020

the measurement of the nuclear structure function , @xmath3 , in deep inelastic scattering ( dis ) indicates that the structure function of a bound nucleon is different from a free nucleon .very precise data on the proton structure function from hera @xcite@xcite@xcite@xcite and nmc @xcite which cover a wide range of kinematics , both in @xmath4 and @xmath5 , sets a stringent demand on the understanding of nucleon structure function in the nuclear environment .the difference between the parton distributions in the free and the bound nucleon , that is , @xmath6 is attributed to the nuclear effects .these effects are characterized in the structure function ratio @xmath7 there are a host of interpretations that describe the dis data on the nuclear targets pioneered by the european muon collaboration ( emc effect ) and it is believed that various mechanisms are responsible for the behavior of @xmath8 in various kinematical intervals : + i)the shadowing effect at @xmath9 where @xmath10 + ii)anti - shadowing effect at @xmath11 where @xmath12 .+ iii)the _ emc _ effect for @xmath13 where @xmath14 .+ iv)the fermi motion for @xmath15 where @xmath12 .+ for a detailed study of these mechanisms and other theoretical models sea arneode @xcite . + currently , there are fairly accurate data on the structure function ratio @xmath16 that has been obtained by different experimental groups @xcite for a variety of nuclei ranging in mass number from @xmath1 to @xmath17 . these data ,now , permit us to investigate the @xmath4 and @xmath5 dependence of the @xmath18 .one feature of the data , almost universal , is the shape of the @xmath4-dependence of the @xmath19 and is basically the same for every nuclei , thus making it rather easy to fit in a wide range of models with totally different underlying assumptions .our interest here is to find a description for the nucleon structure which can be extended to the nuclei ; yet maintaining some universal features common to both . to this end , our starting point is the study of the internal structure of the constituent quark which is common to both nucleon and nuclei .its internal structure is universal and governed by the annihilation and pair production in quantum chromodynamics .we assert that at high enough @xmath5 , it is the structure of this constituent quark that is being probed , whereas , at low @xmath5 it behaves as a structure - less valence quark .such a model has been applied successfully to the proton structure for the entire range of kinematics in @xmath4 and @xmath5 @xcite . encouraged by this results, we think that the nuclear environment can only alter the distribution of the constituent quark , leaving its internal structure intact .the fact that valence quark distribution in nuclei receives a distortion as opposed to its distribution in a free nucleon is very similar to the nucleon swelling and the relaxation of the confinement of the constituent quark within the boundaries of a nucleon embedded in a nucleus . to put it differently , a nucleon in a nucleuswill relax its boundaries and will occupy broader space , preserving the universality of the constituent quark of the relaxed nucleon . none of the number of the constituent quarks , its internal structure , and the average momentum of the constituent will change .thus , here we will be dealing with a collection of structure - full @xmath20 , @xmath21-type constituent quarks participating in dis .to proceed quantitatively , we need to employ a model which is compatible with the deep inelastic scattering data on a free nucleon and then we impose the restrictions due to nuclear effects for the bound nucleons.the picture that comes in mind is the so - called _ valon _ model of r.c.hwa @xcite . in this modelthe nucleon is considered as a system of bound constituent quarks , which themselves have structure .the bound state problem is a nonperturbative effect which is taken into account in the distribution of constituent quarks .the structure of constituent quark is produced perturbatively and is free of bound state problem .in such a picture the structure function of a free nucleon is the convolution of constituent quark distribution in a nucleon and the structure function of the constituent quark itself : @xmath22 where @xmath23 and @xmath24 are the distribution of @xmath20 and @xmath21-type constituents in a nucleon and @xmath25 is the structure function of the constituent calculated in qcd . at sufficiently high @xmath26, @xmath27 can be described accurately in the leading - order results in qcd .they are also calculated in the next - to - leading order[8 ] ; but here we will restrict ourselves to the leading order , because , we are mainly interested in the application of the model and it describes the nuclear structure functions sufficiently accurately . the moments of @xmath27 can be expressed in terms of the evolution parameter : @xmath28 where @xmath29 and @xmath30 are scale parameters determined from the data .the moments of the singlet and non - singlet constituent quark structure function in the leading order solution of the renormalization group equation are given as[9 ] : @xmath31 the anomalous dimensions , @xmath32 , and other associated parameters are : @xmath33^{1/2 } , \\d_{ns}=\frac 1{3\pi b}[1-\frac 2{n(n+1)}+4\sum_{2}^{n}\frac 1j],\\ d_{gq}=\frac{-2}{3\pi b}\frac{2+n+n^2}{n(n^2 - 1)},\\ d_{qg}=\frac{-f}{2\pi b}\frac{2+n+n^2}{n(n+1)(n+2 ) } , \\d_{gg}=\frac{-3}{\pi b}[-\frac 1{12}+\frac 1{n(n-1)}+\frac 1{(n+1)(n+2)}-\frac f{18}-\sum _ { 2}^{n}\frac 1j ] , \\d_{\pm } = \frac 12[d_{ns}+d_{gg}\pm \delta ] , \\b=(33 - 2f)/12\pi , \end{array}\ ] ] since these moments , @xmath34 , are known , using inverse mellin transformation technique , we can obtain the distributions of various components in the constituent quark ; namely , valence , sea quarks , and gluon distributions , @xmath35 .they are as follows : @xmath36 and @xmath37 is the quark - to - gluon evolution function given by : @xmath38,\ ] ] to account for the @xmath39 asymmetry of the nucleon sea which is evident form the violation of the gottfried sum rule , we follow the same procedure as in@xcite and replace the distribution of d - quark in the sea by : @xmath40 the calculation of the moments given in eq.(4 ) is simple . instead of exhibiting the moment distribution , we present the results in parametric form .that is , for every @xmath41 we fit the moments by a form give below : @xmath42 for the valence , sea quarks , and gluon respectively .the dependence of the coefficients @xmath43 , @xmath44 and @xmath45 on @xmath5 , or rather on @xmath41 is given in the appendix . to complete the calculation of @xmath46, we also need to specify constituent quark distributions , @xmath47 and @xmath48 .the simplest approach to evaluate these distributions is to write down the inclusive momentum distribution of the constituent quark in a free nucleon .we will limit ourselves to proton here .these inclusive distributions can be written as : @xmath49 where @xmath50 and @xmath51 are two free parameters and @xmath52 refer to @xmath20-type and @xmath53 refers to the @xmath21-type constituents . by double integration over the unspecified variablewe get the exclusive distributions : @xmath54^{-1}\ y^a\ ( 1-y)^{a+b+1 } \\ g_{\frac dp}(y)=[b(b+1,2a+2)]^{-1}y^b\ ( 1-y)^{2a+1 } \end{array}\ ] ] where @xmath55 is the euler beta function .the above distributions satisfy also the following normalization conditions : @xmath56 in figure 1 , we present the result for @xmath57 for several values of @xmath4 and @xmath26 , its agreement with the experimental data in a wide range of kinematics is rather good .+ having determined the nucleon structure function , next we extend the same procedure to the nuclear medium .as we mentioned earlier , in the nuclear medium the constituent quark distribution has to be modified , but its structure function , or rather parton distribution in a constituent will not be affected by the nuclear environment .the modification to @xmath58 is achieved by introduction of a distortion factor , @xmath59 .hence , we write the inclusive momentum distribution as follows : @xmath60 which leads to : @xmath61 with : @xmath62^{-1 } \\ \alpha ^{\prime } = [ beta(b+\delta+1,2a+2\delta+2)]^{-1 } \end{array}\ ] ] we note that @xmath59 is a function of atomic number , @xmath63 . in ref.@xcite a probabilistic argument is used to find @xmath59 .here we do generalize the same result as follows : @xmath64 so far we have determined the constituent quark distribution in a bound nucleon .we need to find its distribution in a nucleus . for that matter , however , one needs also to know the nucleon distribution in the nucleus .we can use the fermi distribution and arrive at the following result : @xmath65 where @xmath66 and @xmath67 are the light cone coordinates and @xmath68 is the nucleon density in the nucleus .so , straightforward integration would yield the nucleon distribution in a nucleus .however , this also means that the bound nucleon is nearly on shell and therefore , the nuclear binding effect is neglected . to be more realistic , and include this effect , we replace @xmath69 by @xmath70 where @xmath71 and @xmath72 is the effective mass of the nucleon . with these fine tuningswe arrive at the following function for the nucleon distribution in a nucleus : @xmath73 the values of the fermi momentum , @xmath74 , and the binding energy , @xmath75 , are given by the requirements of nuclear physics .going one step further , we carry the fermi motion to the constituent quark level .this is done easily by renormalization of the distribution function give in eq.(14 ) this leads to the following constituent quark distribution per proton in a nucleus : @xmath76 @xmath77 where @xmath78 . having collected all the ingredients ,now we are in a position to evaluate the nuclear structure function ratios , @xmath16 .notice that all of our discussions about the constituent quark distribution are pertinent to proton .finally , taking the neutron excess in the nucleus into consideration , we get : @xmath79\}\ ] ] where @xmath80 and @xmath81 are the free proton and free neutron structure functions , respectively , and @xmath82 is obtained using eq.(1 ) .the results of the model is presented in figure @xmath83 for a variety of nuclei ranging from @xmath1 to @xmath2 . as it is apparent from fig.(2 ) , we see that the model calculation is reasonable only down to @xmath84 . for smaller values of @xmath4 ,@xmath85 an extra ingredient is required . an obvious and natural place to look forthe dynamics of small @xmath4 would be the shadowing effects , where the growth of parton densities are suppressed due to recombination and annihilation of the partons .there are some attempts in the literature to calculate the shadowing corrections , pioneered by qiu@xcite .however we found that such a parameterization is inadequate in describing the data .following ref.[12 ] we modify the shadowing correction of qiu in two ways : ( 1 ) we take the general idea of vector meson dominance which describes low @xmath26 virtual photon interaction , to model the approximate scaling of shadowing by pomeron exchange .it turns out that the experimental data on @xmath86 up to @xmath87 can be well described by a sub - asymptotic pomeron with the effective anomalous dimension going down logarithmically with @xmath4@xcite.(2 ) the magnitude of the shadowing effect in such models is sensitive to the value of vector meson - nucleon cross section .this sensitivity is explored by g. show @xcite and it is described by a quantity : @xmath88 this ration is particularly manifested for heavy nuclei and reflects the ratio of nuclear radius @xmath8 to the mean free path , @xmath89 , of the vector meson inside the nucleus ; which also depends on the nuclear density .that is , the shadowing effect at small @xmath4 values increases with increasing nuclear radius and density . utilizing these modifications to the original work of qiu [ 12 ] , finally we present the following modification of parton distributions in a nucleus : @xmath90 where superscript @xmath91 stands for sea ( valence ) components and @xmath92 is the shadowing factor affecting parton distributions of the constituent quark .these distributions in the nuclear medium are compared to the corresponding distribution in a free nucleon .the parameterized form of this factor is : @xmath93 and @xmath94 is given by : @xmath95 in equation ( 24 ) @xmath96 is evaluated by the @xmath97-moment equations for the parton distribution in the sea of a constituent quark@xcite : @xmath98 where @xmath99 . obviously, nuclear shadowing do not affect the distribution of valence quarks in the constituent quarks .if the momentum loss owing to the recombination process of the shadowed sea quarks is negligible , the momenta carried by each parton component would be conserved separately and approximately , @xmath100 and the following inequalities will be satisfied for @xmath101[15 ] : @xmath102 as a result , following re.@xcite we acquire : @xmath103 @xmath104 where @xmath105 is the shadowing strength and and depends on the starting scale @xmath106 for the evolution to set in .it is easy to verify that the primitive shadowing for gluon in a constituent quark is weaker than that for the sea partons[12 ] at such scale.the form of shadowing correction is in terms of evolution kernels is given as : @xmath107 where @xmath108 are the evolution kernels . defining @xmath109 , these kernel functions are[15 ] : @xmath110 \\ a_{ns}= \frac{\gamma _ { qq}}{\beta _ { \circ } } \ , a_{\pm } ^n=\frac{\gamma ^{\pm n}}{% \beta _ { \circ } } \\ \alpha_ n= \frac{\gamma _ { qq}-\gamma ^{+}}{\gamma ^{-}-\gamma ^{+ } } \\ \beta _n=\frac{% \gamma _ { qq}}{\gamma ^{-}-\gamma ^{+ } } \end{array}\ ] ]we have calculated the nucleus structure function ratio in the context of constituent quark picture , utilizing the essence of the so called valon model . as we can see from the figures ; the model describes the data rather well for a wide range of @xmath63 .it appears that the @xmath26 dependence of the ratio of structure functions is week .the shadowing effect is quite large for small value of @xmath4 , @xmath111 and dies away very rapidly as @xmath4 increases .also , the shadowing correction becomes more pronounced with increasing atomic number @xmath63 .we further note that in this model , which is based on the constituent quark structure , we have not included explicitly the anti - shadowing effects notwithstanding that there are some debates on the subject particularly pertinent to the middle range of @xmath4 .we simply did not need it to describe the data .our anticipation is that such an effect would not be very large .in this appendix we provide numerical values for the coefficients given in equation ( ) .these are the results of our @xmath26 dependence parameterization of parton distributions in a constituent quark of proton .they are calculated in the leading order .@xmath112 ; \\ a_v=-0.300 + 1.254s-0.368s^2 + 0.034s^3\\ b_v=0.196 + 1.286\exp [ -s/3.14 ] \\c_v=-0.844+.451s^{.17 } \end{array}\ ] ] h1 coll ., c. adloff , _ et al ._ , nucl . phys .b * 497 * , 3 ( 1997 ) .zeus coll ., j. breitweg , _ et al ._ , desy-98 - 121 ( 1998 ) .zeus coll ., m. adamus , _ et al ._ , phys lett .b * 407 * ( 1997 ) 432 .zeus coll ., m. derrick , _ et al ._ , z. phys .c * 72 * ( 1996 ) 399 .m. arneodo,_et al . _ ,b * 333 * ( 1990)1 ; p.amaudruz , _ et al .b * 295 * ( 1992 ) 159 .h1 collaboration , t.ahmed,_et al._,nucl .b * 439 * 471 ( 1995),_ibid _ , nucl.phys .b * 407 * ( 1993 ) 515 .zeus collaboration .m. derrick , _ et al ._ z. phys .c * 65 * , r.g.roberts and m.r .whalley , j. phys .g*17 * , ( 1991 ) d1 . 293nmc collaboration , p.amaudruz _ et al ._ phys.rev.lett . * 66 * ( 1991 ) 2712;phys.rev .d * 50 * ( 1994 ) r1 .m. arneodo , phys .rep.*240*,301 ( 1994 ) ., m. arneodo , _ et al ._ , nucl . phys .* b481 * ( 1996 ) 3 .p. amaudruz , _ et al .phys * b441 * ( 1995 ) 3 . f. arash and a. n. khorramian , _ next - to - leading order calculation of the nucleon structure function in the constituent quark model _ , submitted to _ phys .hwa , phys.rev . d*22 * ( 1980 ) 1593 .hwa , phys .rev . d*51 * ( 1995 ) 85. f. arash and l.tomio , phys .b * 401 * ( 1997 ) 207 .w. zhu and l.qian , phys .c*45 * , 1397 , ( 1992 ) ; phy .rev . d*44 * , 2762 , ( 191 ) .jian - wei qiu , nucl .phys.*b291 * , 746 , ( 1987 ) .n. armesto and m.a.branin , z. phys .c76 ( 1997 ) 81 .graham shaw , phys .d47 , r3676 ( 1993 ) .w. zhu and j. g. shin , phys .c*41 * , 1674 , ( 1990 ) .

<S> we have used the notion of the constituent quark model of nucleon , where a constituent quark carries its own internal structure , and applied it to determine nuclear structure functions ratios . </S><S> it is found that the description of experimental data require the inclusion of strong shadowing effect for @xmath0 . </S><S> using the idea of vector meson dominance model and other ingredients this effect is calculated in the context of the constituent quark model . </S><S> it is rather striking that the constituent quark model , used here , gives a good account of the data for a wide range of atomic mass number from @xmath1 to @xmath2 . </S>

hep-ph0410181

the standard model ( sm ) of high energy physics suffers from the gauge hierarchy problem and the flavor problem .the first is the fine tuning required to maintain a low electroweak mass scale ( @xmath3 ) in the theory , in the presence of a high scale , the planck scale ( @xmath4 ) .the second problem is a lack of explanation of the mass hierarchy and mixings of the quarks and leptons .supersymmetry ( susy ) eliminates the gauge hierarchy problem by introducing for each sm particle , a new particle with the same mass but different spin .for example , for each sm quark / lepton a new scalar ( squark / slepton ) , and for each sm gauge boson a new fermion ( gaugino ) , is introduced . if susy is realized in nature , the fact that we do not see such new particles , we believe , could be because susy is spontaneously broken , making the superpartners heavier than the mass ranges probed by experiments .owing to a lack of understanding of how exactly susy is broken , a phenomenologically general lagrangian , for example , the minimal supersymmetric standard model ( mssm ) , is usually considered to compare with data .various experimental searches have placed constraints on the masses and couplings in the mssm .attempts have been made to address the flavor problem by proposing various flavor symmetries . in a supersymmetric theory ,a flavor symmetry in the quark sector might imply a certain structure in the scalar sector , leading to definite predictions for flavor changing neutral current ( fcnc ) processes on which experiments have placed severe constraints . in the literature , a lot of attention has been devoted toward analyzing the minimal flavor violation ( mfv ) scenario , in which the scalar flavor structure is aligned with the quark sector so that the two are simultaneously diagonalized . in mfv , the cabibbo - kobayashi - maskawa ( ckm ) matrix describes the flavor changing interactions in the supersymmetric sector as well , and the only cp violating phase is the one in the ckm matrix . in this work , we do not assume such an alignment , and we will consider non - minimal flavor violation ( nmfv ) , which we treat as a perturbation over the mfv case . in this paperwe wish to explore in what form a supersymmetric extension of the sm , with a u(2 ) flavor symmetry , could influence k and b physics observables .we thus restrict ourselves to the quark and scalar - quark ( squark ) sectors .we consider an `` effective supersymmetry '' @xcite framework , with heavy ( tev scale ) first two generation squarks , in order to escape neutron electric dipole moment ( edm ) constraints .this allows the possibility of having large cp violating phases in the squark sector .we consider a supersymmetric u(2 ) theory @xcite , impose recent @xmath1 and @xmath0 meson experimental constraints and obtain predictions for , , , , mixing and the dilepton asymmetry in @xmath2 .though we consider a specific flavor symmetry , namely u(2 ) , our conclusions would hold for any model with a sizable off - diagonal 32 element in the squark mass matrix .some @xmath0 physics consequences in a supersymmetric u(2 ) theory have been considered in ref .large @xmath5 effects in @xmath0 decays have been carefully analyzed in ref .@xcite , but for simplicity we will restrict ourselves to the case when @xmath5 is not too large .other work along similar lines , though in more general contexts , have been presented in refs . @xcite . in this workwe will include all dominant contributions to a particular observable in order to include interference effects between various diagrams .this has not always been done in the literature .we will then study the implications of recent data from the b - factories , including the @xmath6 penguin decay mode which shows a slight deviation from the sm prediction .the paper is organized as follows : in section [ u2susy.sec ] we specify the supersymmetric u(2 ) theory we will work with , and the choices we make for the various susy and susy breaking parameters . in sections [ kmix.sec ] and [ db2.sec ] we consider @xmath7 ( kaon mixing ) and @xmath8 ( and mixing ) fcnc process , respectively . in section [ db1.sec ]we will consider the implications of such a theory to @xmath9 fcnc processes , namely , , and .we conclude in section [ concl.sec ] .we give details of various squark mixings and their diagonalization in appendix [ appmixang ] , and collect loop functions that we will need in appendix [ apploopfcn ] .the supersymmetric model that we will discuss is as described in ref .@xcite , with the first and second generation superfields ( @xmath10 , a=1,2 ) transforming as a u(2 ) doublet while the third generation superfield ( @xmath11 ) is a singlet .the most general superpotential can be written as : .thus ( @xmath12 labels generations ) , _ i h _ j _ u q_i u^c_j h_u + _ u u^c_i q_j h_u - _ d q_i d^c_j h_d - _ d d^c_i q_j h_d + ( lepton sector ) . ]= _ 1 h + _ 2 h _ a + _ a _ 3 h _b + _ a _ 4 h _b + _ a _ 5 h _b + h_u h_d , where @xmath13 is the cutoff scale below which such an effective description is valid , the @xmath14 are o(1 ) constants , and three new u(2 ) tensor fields are introduced : @xmath15 a u(2 ) doublet , @xmath16 a second rank antisymmetric u(2 ) tensor and @xmath17 a second rank symmetric u(2 ) tensor .the parameter @xmath18 could be complex and we allow for this possibility . following ref .@xcite , we assume that u(2 ) is broken spontaneously by the vacuum expectation value ( vev ) can be different from @xmath19 , but for simplicity we will assume that they are the same . also for simplicity , we take @xmath20 to be real . ]@xcite = ; = v ^ab ; = 0 , = v , with @xmath21 and @xmath22 , in order to get the correct quark masses .these vev s lead to the quark mass matrix given by ( we show only the down quark mass matrix after the @xmath23 is broken by the usual higgs mechanism ) & & -_d + h.c . , + m_d & = & v_d , where @xmath24 is the vev of the higgs field . in @xmath25 ,the @xmath26 s are o(1 ) ( complex ) coefficients , given in terms of the @xmath14 s . ref .@xcite shows that such a pattern of the mass matrix explains the quark masses and ckm elements .if u(2 ) is still a good symmetry at the susy breaking scale , and broken ( spontaneously ) only below the susy breaking scale , the susy breaking terms would have a structure dictated by u(2 ) .for our purposes it is sufficient to consider the down sector squark mass matrices , and they are given as -^2_ll - ^2_rr + + ( ^2_rl + h.c . ) , ^2_ll & = & m_d^_d + _ ll + d - term , + m^2_rr & = & m_d m_d^+ _ rr + d - term , + m^2_rl & = & ^ * _ d + v_d , [ msq.eq ] where @xmath27 and @xmath28 are determined by the susy breaking mechanism . here@xmath29 are real , while @xmath30 could be complex .we will assume that the @xmath28 are of order @xmath31 , a common mass scale .the d terms are flavor diagonal , and since we are interested in fcnc processes , we will not write them in detail , but will think of them as included in @xmath32 and @xmath33 .thus far we have presented the mass matrices in the gauge basis . in the following sections, we will work in the superkm basis in which the quark mass matrix is diagonal , and the quark field rotations that diagonalize the quark mass matrix are applied to the squarks , whose mass matrix would also have been diagonalized in the mfv scheme .since we will not assume an mfv structure , in the superkm basis , there would be small off - diagonal terms in the squark mass matrix , which we treat as perturbations .the structure of the squark mass matrix in the superkm basis is similar to that in eq .( [ msq.eq ] ) owing to the smallness of the mixing angles that diagonalize the quark mass matrix .lacking specific knowledge about the susy breaking mechanism realized in nature , we make some assumptions on the susy mass spectrum .neutron edm places strong constraints on the cp violating phases and the masses of the first two generations of scalars . to satisfy this and other collider constraints, we consider an `` effective susy '' framework in which the scalars of the first two generations are heavy , suppressing edm , and allowing for larger cp violating phases .defining the scalar mass scale , @xmath34 , we take all @xmath35 except for @xmath36 .we take @xmath37 , the gaugino mass parameter @xmath38 and charged - higgs masses to be @xmath39 and the gluino mass to be @xmath40 .we assume such a spectrum just above the weak scale without specifying what mechanism of susy breaking and mediation might actually give rise to it .as we will show later , if realized in nature such a spectrum would lead to enhancements in the processes we are considering here .the rates of various fcnc processes follow from the mass matrix that we have specified in eq .( [ msq.eq ] ) .we will work in the superkm basis .the interaction vertices in the mass basis are obtained by diagonalizing the mass matrices in eq .( [ msq.eq ] ) , and the perturbative diagonalization to leading order is shown in appendix .[ appmixang ] .the dominant nmfv susy contributions to fcnc processes would be due to the 32 and 23 entries in eq .( [ msq.eq ] ) , since they are the biggest off - diagonal terms . for conveniencewe define _ 32,23^rl , rr , ll .since we have written down an effective theory and not specified the dynamics of u(2 ) and susy breaking , we can only specify the order of magnitude of @xmath41 . to parametrize this uncertaintywe write , _ 32,23^rl = d_32,23^rl , _ 32,23^ll , rr = d_32,23^ll , rr , [ deldef.eq ] where we have denoted the unknown @xmath42 coefficients by @xmath43 .@xmath5 & @xmath44 @xmath45 & @xmath46 & @xmath18 & @xmath47 @xmath48 & @xmath49 & @xmath38 & @xmath50 @xmath51 & @xmath49 & @xmath52 & @xmath53 @xmath31 & @xmath49 & @xmath54 & @xmath50 @xmath55 & @xmath56 & @xmath57 & @xmath58 we summarize our choice of the parameters in table [ sparam.tab ] .for these values , from eq .( [ deldef.eq ] ) , the natural sizes of @xmath41 are given by _32,23^rl = 6.8210 ^ -4 d_32,23^rl , _ 32,23^ll , rr = 0.02 d_32^ll , rr .[ delnum.eq ] we will find in the rest of this paper that @xmath59 induces nmfv @xmath9 fcnc processes dominantly , while @xmath60 induces @xmath7 and @xmath8 fcnc processes . though the @xmath61 and @xmath62 elements have similar magnitudes , the @xmath61 gluino nmfv contribution to @xmath9 fcnc processes is larger , since we take @xmath63 to be much lighter than the other scalars , and the @xmath62 gluino diagrams are relatively suppressed by the heavier @xmath64 mass .therefore , in this work we will include only the dominant @xmath61 contribution .we illustrate this in fig .[ bsg3223.fig ] , where we show the gluino contribution to as an example . similarly , owing to the smaller @xmath63 mass, the @xmath65 nmfv contribution to @xmath7 and @xmath8 fcnc processes is relatively larger compared to the @xmath66 contribution .we note here that , from eq .( [ thdiagsb.eq ] ) in appendix [ appmixang ] , the sbottom mixing angle is negligibly small , and therefore , we ignore sbottom mixing effects ; stop mixing is not as small and we include its effects . in the next three sections we will discuss the implication of the u(2 ) model to @xmath7 , @xmath8 and @xmath9 fcnc processes . from this we will see that present experimental data are compatible with the values shown in table [ sparam.tab ] , and we will obtain expectations for some measurements that are forthcoming .we will present plots of different fcnc effects by varying a couple of parameters at a time , while keeping all others fixed at the values shown in table [ sparam.tab ] .the cp violation parameter @xmath67 due to mixing in the kaon sector has been measured to be @xcite |_k| = ( 2.2840.014 ) 10 ^ -3 .we wish to estimate the new physics contributions to @xmath67 in the scenario that we are considering . herewe note that even though the direct cp violation parameter @xmath68 has also been measured , large hadronic uncertainties do not permit us to constrain new physics models through this observable .kaon mixing is governed by the @xmath7 effective hamiltonian ^eff_s=2 = _i=1 ^ 5 c_i q_i + _i=1 ^ 3 _ i _ i , where , q_1 & = & q_2 & = & |d_r^s_l^|d_r^s_l^ , + q_3 & = & |d_r^s_l^|d_r^s_l^ , + q_4 & = & |d_r^s_l^|d_l^s_r^ , + q_5 & = & |d_r^s_l^|d_l^s_r^ .[ ds2ops.eq ] the operators @xmath69 ( i=1,2,3 ) are obtained by exchanging @xmath70 . in the sm and the new physics model we are considering ,the dominant contributions are to @xmath71 , as we explain later in this section .the cp violation parameter @xmath67 is then given by ( see for example ref .@xcite ) _ k = e^i/4 i m , where @xmath72 is the bag parameter and @xmath73 is the kaon decay constant .in addition to the sm @xmath74 box diagram contribution to @xmath75 , in the supersymmetric u(2 ) theory we are considering , the charged - higgs and chargino mfv contributions could be sizable .the dominant mfv contributions to @xmath75 can be written as c_1^mfv = c_1^w + c_1^h + c_1^ , which is the sum of the sm @xmath74 , the charged - higgs , and the chargino contributions , respectively . + * sm contribution * : the sm @xmath74 contribution is @xcite c_1^sm(m_t ) = c_1^w(m_t ) = ( v_td^ * v_ts ) ^2 s_0(x_t ) + .+ ( v_cd^*v_cs)^2 s_0(x_c ) + 2 ( v_td^*v_ts v_cd^*v_cs ) s_0(x_t , x_c ) , [ db2c1wmt.eq ] where the function @xmath76 is given in appendix [ apploopfcn ] , eq . ( [ boxsms0.eq ] ) , and @xmath77 , @xmath78 .the qcd correction due to renormalization group running from @xmath79 to @xmath80 gives c_1^sm(m_k ) = c_1^w(m_k ) = ( v_td^ * v_ts ) ^2 _k33 s_0(x_t ) + .+ ( v_cd^*v_cs)^2 _k22 s_0(x_c ) + 2 ( v_td^*v_ts v_cd^*v_cs ) _ k32 s_0(x_t , x_c ) , where the @xmath81 are qcd correction factors given in eq .( [ paramk.eq ] ) below , and @xmath82 are the ckm matrix elements .+ * charged - higgs contribution * : supersymmetric theories require two higgs doublets to give masses to the up and down type fermions .the higgs doublets contain the charged - higgs @xmath83 , and the dominant charged - higgs - top contribution is @xcite , which becomes important only at large @xmath5 . ]c_1^h(m_t ) & = & ( v_td^ * v_ts ) ^2 , [ db2c1hmt.eq ] + f_v^h & = & x_t^2 y_1(r_h , r_h , x_t , x_t ) + x_t^2 y_1(1,r_h , x_t , x_t ) - x_t y_2(1,r_h , x_t , x_t ) , where @xmath84 , and the functions @xmath85 and @xmath86 are given in appendix [ apploopfcn ] , eq .( [ boxhy.eq ] ) . + * chargino contribution * : the dominant chargino - right - handed - stop contribution is @xcite c_1^(m_t ) & = & ( v_td^ * v_ts ) ^2 , [ db2c1chmt.eq ] + f_v^&= & |_r^(i)|^2 |_r^(j)|^2 y_1(r_t_2,r_t_2,s_i , s_j ) , where @xmath87 , @xmath88 , and the coupling is given by _r^(i ) = ( c_r^*)_1i(c_t^*)_12 - , with the chargino and stop diagonalization matrices @xmath89 and @xmath90 given in appendix [ appmixang ] , eqs .( [ cdiagch.eq ] ) and ( [ cdiagsb.eq ] ) , respectively . taking into account renormalization group running, we have c_1^h,(m_k ) _ k33 c_1^h,(m_t ) . :in general , the nmfv gluino contributions induce many operators shown in eq .( [ ds2ops.eq ] ) , but in the model we are considering , these are not significant due to a suppression from the heavy @xmath91 and @xmath92 masses , glashow - iliopoulos - maiani ( gim ) suppression owing to their approximate degeneracy ( split only by @xmath93 , cf .( [ msq.eq ] ) ) , and the contribution from the relatively light right - handed sbottom being suppressed by its small mixing to the first two generations .moreover , owing to the structure of the mass matrix , eq .( [ msq.eq ] ) , the gluino contribution is real , and hence does not contribute to @xmath67 . in our numerical analysis ,we take the following values for the various parameters @xcite : _ k33 & = & 0.57 , _ k22 = 1.38 , _ k32 = 0.47 , + f_k & = & 0.160 gev , 0.6 < b_k < 0.9 , [ paramk.eq ] + m_k & = & 0.497 gev , m_k = ( 3.480.01)10 ^ -15 gev , m_c=(1.20.2 ) gev .the sm prediction for @xmath67 is in agreement with the experimental data , but it should be noted that there is considerable uncertainty in the lattice computation of the bag parameter @xmath72 ( see eq .( [ paramk.eq ] ) ) .the chargino and charged - higgs contributions to @xmath75 add constructively with the sm contribution .therefore , if the true value of @xmath72 is taken to be closer to the lower limit , we can allow mfv contributions to be up by a factor of 1.2 compared to the sm value ; i.e. , @xmath94 .[ epskmfv.fig ] shows the region of mfv parameter space where this is satisfied .this justifies some of the choices we make in the list shown in table [ sparam.tab ] .we start by discussing in general mixing and later specialize in succession to ( q = d ) and to ( q = s ) .the @xmath8 effective hamiltonian is given by @xcite : ^eff_b=2 = _i=1 ^ 5 c_i q_i + _i=1 ^ 3 _ i _ i , where , for @xmath95 , q_1 & = & q_2 & = & |q_r^b_l^|q_r^b_l^ , + q_3 & = & |q_r^b_l^|q_r^b_l^ , + q_4 & = & |q_r^b_l^|q_l^b_r^ , + q_5 & = & |q_r^b_l^|q_l^b_r^ .the operators @xmath69 ( i=1,2,3 ) are obtained by exchanging @xmath70 .the wilson coefficients @xmath96 are run down from the susy scale , @xmath97 , using @xcite c_r(m_b)=_i _ s ( b_i^(r , s)+c_i^(r , s))^a_i c_s(m_s ) [ bsevol.eq ] where @xmath98 and the @xmath99 , @xmath100 and @xmath101 are constants given in ref .@xcite .the matrix elements of the @xmath102 in the vacuum insertion approximation are given by @xcite .& = & m_b_q^2 f_b_q^2 b_1 ( ) , + & = & - ( ) ^2 m_b_q^2 f_b_q^2 b_2 ( ) , + & = & ( ) ^2 m_b_q^2 f_b_q^2 b_3 ( ) , + & = & ( ) ^2 m_b_q^2 f_b_q^2 b_4 ( ) , + & = & ( ) ^2 m_b_q^2 f_b_q^2 b_5 ( ) , [ bbmatel.eq ] where we take for the decay constants @xmath103 gev and the bag parameters ( at scale @xmath80 ) @xmath104 , @xmath105 , @xmath106 , @xmath107 and @xmath108 @xcite .the @xmath95 mass difference is given by m_b_q = 2 | m_12(b_q ) element for the system , and is given by m_12 & = & m_12^sm + m_12^susy , + _ 12 & & _@xmath109 to an excellent approximation is dominated by the sm tree decay modes . from refs .@xcite we have , |m_12(b_q)| = | | , [ m12bq.eq ] + _ 12^sm = ( -1 ) , [ gam12bq.eq ] where @xmath110 , @xmath111 and we take @xmath112 .the dilepton asymmetry in @xmath95 is given by @xcite a_ll^b_q = i m ( ) .we discuss next the sm and new physics contributions to the coefficients @xmath96 and @xmath113 . + * mfv contribution * : the sm @xmath74 contribution is almost identical to that shown in eq .( [ db2c1wmt.eq ] ) but for the fact that it is sufficient to keep only the top contribution ( the @xmath114 term ) and changing the ckm factor to @xmath115 .the new physics mfv charged - higgs and chargino contributions are again identical to eqs .( [ db2c1hmt.eq ] ) and ( [ db2c1chmt.eq ] ) , respectively , with the same change for the ckm factors .@xmath116 is evolved down to @xmath80 using eq .( [ bsevol.eq ] ) .* gluino contribution * : we only include the dominant gluino - right - handed - sbottom box diagrams with @xmath61 and @xmath117 mass insertions , since @xmath63 is the only relatively light down type squark in our scenario .these contributions are given by @xcite _ 1^g(m _ ) & = & i g_s^4 ^2 , + _ 2^g(m _ ) & = & i g_s^4 ^2 , + _ 3^g(m _ ) & = & -i g_s^4 ^2 , [ c1bqgl.eq ] with the box integrals @xmath118 and @xmath119 given in appendix [ apploopfcn ] .the couplings are given by _b3^rr = & , & _b3^rl = , + _d3^rr = e^-i(_32^rr+_12^rr ) & , & _s3^rr = - e^-i_32^rr , + _d3^rl = e^-i(_32^rl+_12^rl ) & , & _s3^rl = - e^-i_32^rl , [ gq3coup.eq ] obtained from the @xmath120 mixing matrix that is the product of @xmath121 and @xmath122 , with the mixing angles @xmath123 and phases @xmath124 given in appendix [ appmixang ] . in our u(2 ) model , if @xmath125 is of the same order as @xmath31 , based on the estimate in eq .( [ delnum.eq ] ) , we expect @xmath126 to receive the dominant gluino contribution from @xmath117. we will focus on this contribution in the following .we point out in appendix [ appmixang ] , eq .( [ mdsllrr.eq ] ) , that @xmath127 mixing can be generically large ( near maximal ) , in which case the gluino contributions to both and mixing can be sizable .however , if @xmath128 , this mixing can be small and the gluino contribution to mixing is negligible since it is proportional to @xmath129 , cf .( [ c1bqgl.eq ] ) and ( [ gq3coup.eq ] ) .the gluino contribution to , however , can still be sizable in either case since it is proportional to @xmath130 . the mass difference ( @xmath131 ) , and cp violation in ( @xmath132 )have been measured to be @xcite , m_d & = & 0.502 0.007 ps^-1 , + a_k_s & = & 0.7250.037 .[ s2bexp.eq ] in the sm , the usual notation is , @xmath133 .as we have already pointed out in section [ kmix.sec ] , the charged - higgs and chargino mfv contributions add constructively with the sm contribution .the sm prediction agrees quite well with the data , but given the uncertainty in @xmath134 , cf .below eq .( [ bbmatel.eq ] ) , it might be possible to accommodate an mfv contribution up to a factor of about 1.3 bigger than the sm contribution .we show in fig .[ dmbdmfv.fig ] the region in mfv parameter space that satisfies this constraint , ignoring the gluino contribution .as pointed out in the previous subsection , in general we expect in the u(2 ) model , @xmath127mixing to be near maximal , in which case the gluino contribution to can be sizable . the gluino contribution can then be important to both @xmath131 and @xmath132 . taking this into account ,we can write @xmath135 , where @xmath136 is the new phase in @xmath137 @xcite , and we have @xcite a_k_s & = & im(_k ) , + _ k - & , & , with @xmath138 and @xmath109 given in eq . ( [ m12bq.eq ] ) . (the `` @xmath139 '' sign in @xmath140 is because the final state is cp odd . ) in our case , @xmath141 , so that a_k_s , where `` arg '' denotes the argument of the complex quantity . for the casewhen @xmath142 mixing is large , we show the gluino contribution to in fig [ m12bdnmfv.fig ] .the plot on the left also shows the constraint from @xmath132 , which is not shown in the plot on the right since almost the whole region shown is allowed .the region @xmath143 is not shown since it is identical to the region @xmath144 . from the figure, we see that in the large mixing case , the constraint on @xmath117 is quite strong .however , if @xmath142 mixing is small , the constraint on @xmath117 from mixing is weak .mixing has not yet been observed and the current experimental limit is @xmath145 @ 95% c.l .the sm prediction is : @xmath146 @xcite .the sm prediction for the dilepton asymmetry @xmath147 is small , around @xmath148 , cf .references in ref .@xcite . mixing depends quite sensitively on @xmath117 , and for the region in fig .[ m12bdnmfv.fig ] allowed by mixing , we find @xmath149 and @xmath150 .this @xmath151 is a little higher than the sm prediction , although may be within the sm allowed range , given uncertainties .as we pointed out in the previous subsection , if @xmath127 mixing is small , then the mixing constraints on @xmath117 becomes weak . if such is the case , there are essentially no constraints from mixing , and we show contours of @xmath151 and @xmath147 in fig .[ bsmixrr.fig ] .we show only the range @xmath152 , since the @xmath153 range is identical to this . it can be seen that @xmath151 can increase significantly above the sm prediction .the projected run ii sensitivity for @xmath151 at the tevatron with @xmath154 is around @xmath155 @xcite , and can probe a significant region of u(2 ) parameter space .if a higher value of @xmath151 is measured than what the sm predicts , it would indicate the presence of new physics .measuring @xmath147 can also significantly constrain @xmath117 as can be seen from fig .[ bsmixrr.fig ] ( right ) .the @xmath9 effective hamiltonian at a scale @xmath18 in the operator produce expansion ( ope ) is @xcite ^eff_b=1 = -v_ts v_tb^ * ( _ i=1 ... 6,9,10 c_i ( ) o_i ( ) + c_7 ( ) o_7 ( ) + c_8 g ( ) o_8 g ( ) ) , with o_1 & = & ( |s _ c_)_v - a ( |c _ b_)_v - a , + o_2 & = & ( |s c)_v - a ( |c b)_v - a , + o_3 & = & ( |s b)_v - a_q(|qq)_v - a , + o_4 & = & ( |s _ b_)_v - a _ q ( |q _ q_)_v - a , + o_5 & = & ( |s b)_v - a_q(|qq)_v+a , + o_6 & = & ( |s _ b_)_v - a _ q ( |q _ q_)_v+a , + o_7 & = & m_b |s_^ ( 1+_5 ) b_f _ , + o_8 g & = & m_b |s_^ ( 1+_5)t^a _ b_g^a _ , + o_9 & = & ( |s b)_v - a ( |ee)_v , + o_10 & = & ( |s b)_v - a ( |ee)_a , where , the subscript @xmath156 means @xmath157 , and @xmath158 , @xmath159 are the electromagnetic and color field strengths , respectively. the wilson coefficients can be computed at the scale @xmath160 ( the @xmath74 boson mass ) , and then run down to the scale @xmath80 ( the @xmath161 quark mass ) .below , when no scale is specified for the coefficients , it is understood to be at @xmath80 , i.e. , @xmath162 .the coefficients when run down from @xmath160 to @xmath80 mix under renormalization , so that @xcite c_j & = & _ i=1 ^ 8 k_ji ^a_i , ( j=1, ... ,6 ) , + c_7 & = & ^ c_7(m_w ) + ( ^ - ^ ) c_8g(m_w ) + _ i=1 ^ 8 h_i ^a_i c_2(m_w ) , + c_8 g & = & ^ c_8g(m_w ) + _ i=1 ^ 8 |h_i ^a_i c_2(m_w ) , [ evol.eq ] where @xmath163 and @xmath164 , @xmath165 , @xmath99 and @xmath166 are given in ref . @xcite .in addition , the evolution equation for @xmath167 is given in ref .@xcite , and @xmath168 is not renormalized .separating out the new physics contribution to the renormalization group evolution , i.e. , eq .( [ evol.eq ] ) , we get c_2 & = & c_2^sm , + c_7 & = & c_7^sm + 0.67 c_7^new(m_w ) + 0.09 c_8g^new(m_w ) , + c_8 g & = & c_8g^sm + 0.70 c_8g^new(m_w ) , [ wil.eq ] in which the superscript `` sm '' indicates the contribution from the sm , and `` new '' from new physics .* sm contribution * : the sm @xmath169 contribution to @xmath170 and @xmath171 are given by @xcite c_2^sm(m_w ) & = & 1 , + c_7,8g^sm(m_w ) & = & f^ll_7,8 ( ) , + [ c78sm.eq ] where @xmath172 are given in appendix [ apploopfcn ] .using eq .( [ evol.eq ] ) we can compute @xmath173 , @xmath174 and @xmath175 . in the following , we will discuss , in order , the new physics contribution arising from the charged - higgs boson ( @xmath83 ) , charginos ( @xmath176 ) and gluinos ( @xmath177 ) . + * charged higgs ( @xmath83 ) contribution * : the charged - higgs contribution to is given by @xcite c_7,8g^h(m_w ) & = & ^2f^ll_7,8 ( ) + ^ll_7,8 ( ) , [ c78h.eq ] where @xmath172 and @xmath178 are given in appendix [ apploopfcn ] .+ * chargino ( @xmath176 ) contribution * : the chargino - stop contribution can be comparable to the sm contribution for a light stop and chargino . in the scenario that we are considering , the stop mixing angle is negligibly small and @xmath179 and @xmath180 .we therefore run the @xmath181 contribution from @xmath182 down to @xmath160 and evaluate the @xmath183 contribution at @xmath160 .the chargino - stop contribution is @xcite c_7,8g^_1(_0 ) = -_j=1 ^ 2 , + c_7,8g^_2(m_w ) = -_j=1 ^ 2 , [ c78ch.eq ] where the loop functions @xmath184 are given in appendix [ apploopfcn ] , and @xmath185 and @xmath186 contain the stop and chargino mixing matrices .explicit expressions for @xmath185 , @xmath186 and the renormalization group equations to evolve @xmath187 down to @xmath160 are given in ref .* gluino ( @xmath177 ) contribution * : in our nmfv scenario , the gluino contributions can be sizable since they couple with strong interaction strength . furthermore , because the sbottom mixing angle is negligibly small , @xmath188 and @xmath189 . keeping only the @xmath190 enhanced piece , the gluino contribution is @xcite c_7^(m_w ) = - e^-i^rl_32 , + c_8g^(m_w ) = e^-i^rl_32 , [ c78gl.eq ] where the mixing angle @xmath191 and phase @xmath192 are defined in appendix [ appmixang ] , and @xmath193 and @xmath194 are defined in appendix [ apploopfcn ] . in the above equation , we have neglected the effect of running the @xmath195 contribution from @xmath182 to @xmath160 as the @xmath196 contribution is dominant .the dominant new physics contribution is given by adding eqs .( [ c78h.eq ] ) , ( [ c78ch.eq ] ) and ( [ c78gl.eq ] ) , which yields c_7,8g^new(m_w ) = c_7,8g^h(m_w ) + c_7,8g^(m_w ) + c_7,8g^(m_w ) .[ bsgnewphy.eq ] in what follows we will discuss in detail the new physics contribution predicted by the u(2 ) model to the rare decay processes , , and . [ [ section ] ] , ~~ the dominant operators contributing to and are @xmath197 , @xmath198 and @xmath199 .the decay branching ratio b.r .( ) , at leading order , normalized to the semi - leptonic @xmath200 , is given by @xcite |_(e _> ( 1-)e_^max ) = | | ^2 |c_7|^2 , [ brbsg.eq ] where @xmath201 is a phase space function , and @xmath202 is the fractional energy cut , i.e. , only photon energy @xmath203 is accepted .the cp asymmetry in is given by @xcite a_cp^ ( ) & = & |_e _ > ( 1 - ) e_^max , + & = & \ { a_27()im[c_2c_7^ * ] + a_87()im[c_8gc_7^ * ] + a_28()im[c_2c_8g^ * ] } .[ acpbsg.eq ] for @xmath204 , which is a typical experimental cut , we use @xmath205 , @xmath206 and @xmath207 @xcite .the experimentally measured @xcite branching ratio is @xmath208 . in the sm, we have @xmath209 , @xmath210 and @xmath211 . the sm prediction for b.r .( ) , which depends on @xmath212 , cf .( [ brbsg.eq ] ) , is largely consistent with experiment , and new physics contributions to @xmath212 is constrained by this branching ratio . in the context of susythis has been analyzed , for example , in refs .@xcite .the sm cp asymmetry in is of the order of 1% , so that a larger cp asymmetry measured would imply new physics @xcite .the present limit at 95 % c.l . is @xcite @xmath213 . the b.r .( ) is obtained simply from eq .( [ brbsg.eq ] ) = c(r ) | | ^2 @xmath214 .( ) has large experimental and theoretical uncertainties and ref .@xcite suggests that the data might prefer a b.r .value of around @xmath215 .[ bsgacpmhm2.fig ] and [ bsgacplr.fig ] show the interplay between the @xmath169 , @xmath83 , @xmath176 and @xmath177 contributions to , where the sum of these contributions to the magnitude of @xmath216 is constrained by b.r .( ) . the experimental data on b.r .( ) allows ( at @xmath217 ) the region bounded by the contours shown in the figures . in the plots , the parameters are as given in table [ sparam.tab ] , and some relevant ones are varied as shown in the figures . to illustrate the dependence on the mfv parameters , we consider for example , in fig .[ bsgacpmhm2.fig ] , the dependence of @xmath218 and b.r .( ) as a function of @xmath219 and @xmath38 ( left ) and as a function of @xmath220 and @xmath221 ( right ) , for the choice of parameters shown in table [ sparam.tab ] .the experimental @xmath217 allowed contours of b.r .( ) are also shown . in fig .[ bsgacplr.fig ] ( left ) , the shaded regions show @xmath218 as a function of the magnitude and argument of @xmath222 , the dimensionless @xmath42 coefficient defined in eq ( [ delnum.eq ] ) . in fig .[ bsgacplr.fig ] ( right ) , we show contours of b.r .( ) , and a b.r . of up to about 15% can be accommodated in this model .[ [ section-1 ] ] the dominant operators contributing to ( @xmath223 ) are @xmath198 , @xmath224 and @xmath225 .it is usual to define c_9 ( ) & & _ 9 ( ) , + c_10 & & _ 10 , and .the ( differential ) partial width @xmath226 , normalized to @xmath227 , is given by @xcite : r ( ) = ||^2 , where @xmath228 and @xmath229 are phase space functions and @xmath230 is the qcd corrected @xmath231 , given in terms of @xmath231 and @xmath96 ( @xmath232=1 ... 6 ) @xcite .integrating this we get the prediction for the decay branching ratios and we show this in table [ bslldata.tab ] for the sm along with the experimental result @xcite .we choose the lower limit on the integration to correspond to a typical experimental choice , @xmath233 .since the rate of is down by the square of the electromagnetic coupling constant compared to , the experimental errors are comparatively larger . &@xmath234 & @xmath235 fig .[ bllbrcontlr.fig ] shows the contours of b.r .( ) as a function of @xmath222 .compared to b.r .( ) , cf . fig .[ bsgacplr.fig ] ( left ) , the constraint is not very stringent right now , and improved statistics at the b - factories could place tighter constraints on the parameter space .[ [ section-2 ] ] the decay ( @xmath236 at the quark level ) can be a sensitive probe of new physics since the leading order sm contribution is one - loop suppressed , and loop processes involving heavy susy particles can contribute significantly . however , the computation of b.r .( ) suffers from significant theoretical uncertainties in calculating the hadronic matrix elements .we follow the factorization approach , details of which are presented in ref .the theoretical uncertainties largely cancel in the cp asymmetry , and is therefore a good probe of new physics .the cp asymmetry in is defined by a_cp^ & & + & = & -c_k + s_k , where c_k & & , + s_k & & , + _ k & & - e^-2i(+ _ d ) , where @xmath237 represents the state that is a @xmath238 at time @xmath239 , @xmath240 is the mass difference , @xmath241 is the usual angle in the sm ckm unitarity triangle fits to the cp asymmetry in , and @xmath136 is any new physics contributions to mixing ( @xmath136 is discussed in section [ bdmix.sec ] ) .the sm predicts that the cp asymmetry in and should be the same , i.e. , @xmath242 .the amplitude and partial decay width are given by @xcite : a ( ) & = & _ p = u , c_p + ( ) & = & ( f_1^bk)^2 |a()|^2 ^ where the phase space function . ]@xmath243 , the @xmath244 decay constant @xmath245 mev , the form factor @xmath246 and @xmath247 .the sm @xmath99 s , in terms of the @xmath96 s , are given in ref .@xcite to which we add the new physics contribution given in eq .( [ bsgnewphy.eq ] ) .we do not include the power - suppressed weak annihilation operators and we refer the reader to refs .@xcite and @xcite for a more complete discussion . asexplained in section .[ susyparam.sec ] , we are only including the @xmath61 susy contribution , as this is the dominant one .the amplitude for the cp conjugate process is obtained by taking @xmath248 .the current experimental average @xcite is summarized in table .[ bphikdata.tab ] .the sm requires @xmath249 , but the experimental data has about a @xmath217 discrepancy between @xmath250 and @xmath251 . and @xmath252 is bigger , currently at about @xmath253 , where @xmath254 and @xmath252 are the averages over all measured @xmath6 ( penguin ) and @xmath255 modes , respectively . ]though not convincing yet , this could be an indication of new physics and we ask if this can be naturally explained in the theory we are considering .@xmath256 & @xmath257 & @xmath258 @xmath250 & @xmath259 & @xmath260 @xmath261 & @xmath262 & @xmath263 we showed in section [ bdmix.sec ] , that if @xmath264 mixing is small , there is no significant new phase in @xmath137 ( i.e. , @xmath265 ) .for this case , we scan the parameter space @xmath266 , @xmath267 , @xmath268 , and in fig .[ bphikcpth0.fig ] show a scatter - plot of the points that satisfy all experimental constraints including b.r .( ) and b.r .we find that it is possible to satisfy all experimental constraints including the recent data shown in table [ bphikdata.tab ] in the framework we are considering .furthermore , there are strong correlations between @xmath218 , @xmath250 and @xmath261 .as the accuracy of the experimental data improve , we can use these correlations to ( in)validate the choices that we make in our model .large @xmath264 mixing can lead to a nonzero @xmath136 which depends on @xmath269 as explained in section [ bdmix.sec ] .we therefore include this new phase and perform a scan over @xmath269 , @xmath270 , @xmath266 , @xmath267 and @xmath268 .we show the points that satisfy all experimental constraints and the resulting @xmath218 , @xmath250 and @xmath261 in fig .[ bphikcp.fig ] .we again see from fig .[ bphikcp.fig ] that @xmath218 , @xmath250 and @xmath261 are strongly correlated , although the effect of the new phase in mixing allows new regions of parameter space compared to the small mixing case shown in fig .[ bphikcpth0.fig ] . even in the case of large mixingwe find that it is possible to satisfy all experimental data including the @xmath250 and @xmath261 .one feature that we find in either large or small mixing case is that sign @xmath271 is positively correlated with sign @xmath272 .thus , further data could shed light on the validity of the choices that we make in our model .a supersymmetric u(2 ) theory has the potential to explain the gauge hierarchy and flavor problems in the sm .we assumed an effective susy mass spectrum just above the weak scale , the only relatively light scalars being the right handed stop and sbottom ( weak scale masses ) .we analyzed what such a hypothesis would imply for @xmath1 and @xmath0 meson observables by including all the dominant contributions that can interfere in a certain observable .although for definiteness we considered a u(2 ) framework , our conclusions hold for any theory with a similar susy mass spectrum and structure of the squark mass matrix .the cp violation parameter in kaon mixing , @xmath67 , can impose constraints on the mfv parameter space of our model , as we showed in fig .[ epskmfv.fig ] , while the gluino contribution to @xmath67 is negligible .there is sufficient room to accommodate the mfv contributions to @xmath67 , given the present uncertainty in the lattice computation of the bag parameter @xmath72 .we find that mixing and @xmath132 ( @xmath273 ) can impose constraints on the supersymmetric u(2 ) theory .in addition to the mfv contribution , if @xmath142 mixing is large , the gluino contributions to mixing can be significant leading to a strong constraint on the 32 entry of the rr squark mass matrix , @xmath117 , as shown in fig . [ m12bdnmfv.fig ] . furthermore , in this case , there is a new phase in the mixing amplitude coming from the susy sector .however , if @xmath142 mixing is small , the constraint on @xmath117 from mixing is weak . mixing mostsensitively depends on @xmath117 in the susy u(2 ) theory .if @xmath117 is unconstrained by mixing ( small @xmath142 mixing ) , we showed that @xmath151 can be increased to quite large values ( up to about @xmath155 ) , cf .[ bsmixrr.fig ] .the current and upcoming experiments can reach sensitivities required to see the sm prediction for @xmath151 . seeing a higher value , or not seeing a signal at all , might hint at some new physics of the type we are considering .we also presented expectations for the @xmath2 dilepton asymmetry , @xmath147 , which can constrain @xmath117 . the experimental data on b.r .( ) imposes a constraint on the susy theory . while satisfying this constraint , we showed that an enhancement in cp violation in is possible . since in the sm, @xmath218 is predicted to be less than @xmath274 , if a much larger value is measured , it would clearly point to new physics . in fig .[ bsgacplr.fig ] we presented the expectations for @xmath218 and b.r .( ) while varying the magnitude and phase of @xmath61 .we also presented expectations for b.r .( ) in fig .[ bllbrcontlr.fig ] .the present experimental data on the cp violation in has about a @xmath217 deviation from the sm prediction , and it will be very interesting to see if this would persist with more data .we showed that such a deviation can be accommodated in the framework we are considering , both for large or small @xmath142 mixing .we showed , in fig .[ bphikcp.fig ] , that @xmath218 can be enhanced significantly while satisfying all other experimental bounds including the present data on @xmath250 and @xmath261 . in figs .[ bphikcpth0.fig ] and [ bphikcp.fig ] , for small and large @xmath142 mixing respectively , we see strong correlations between @xmath218 , @xmath250 and @xmath261 .comparing these with upcoming data with improved precision could shed light on the validity of the choices that we make in our model .we conclude by remarking that the prospects are exciting for discovering susy in b - meson processes at current and upcoming colliders . here , we showed this for a susy u(2 ) model . to unambiguously establish that it is a susy u(2 ) theory , and to determine the various susy breaking parameters ,will require looking at a broad range of observables .we thank p. ko , u. nierste , k. tobe , j. wells and m. worah for many stimulating discussions , and , d. bortoletto and c. rott for discussions on the tevatron bounds .cpy thanks the hospitality of the national center for theoretical sciences in taiwan , roc , where part of this work was completed .sg acknowledges support from the high energy physics group at northwestern university where this work was completed . this work was supported in part by the nsf grant phy-0244919 .the charged su(2 ) majorana gauginos @xmath275 , @xmath276 can be combined to form the dirac spinor w^+ = , where @xmath277 .the up and down type higgsinos can be combined to form the dirac spinor h^+ = .the chargino mass terms can then be written as - ( m _ p_l + m_^ p_r ) , where _ = .we can go to the chargino mass eigen basis @xmath278 by making the rotations p_l , r = ( c_l , r ) p_l , r , with the rotation matrices @xmath279 given as _ = , [ cdiagch.eq ] where the mixing angles and phases are @xcite _ l & = & - + & = & + _ r & = & - + & = & + _ r & = & ( c_r ( m_2 c_l + m_w s_l e^i_l ) + s_r e^-i_r ( m_w c_l + s_l e^i_l ) ) + _ r & = & ( c_r ( - m_w s_l e^-i_l - c_l ) + s_r e^i_r ( -m_2 s_l e^-i_l + m_w c_l ) ) with @xmath280 so that @xmath281 , and , @xmath282 and @xmath283 .the sbottom mass terms are given as , cf .( [ msq.eq ] ) - , where the @xmath284 are the d - term contributions given as ^d = ( t_3 - q_em^2_w ) m_z^2 .this is diagonalized by the rotation & = & + & & ( c_b ) , [ cdiagsb.eq ] where the mixing angle and phase are given by & = & , + _ b & = & . [ thdiagsb.eq ]we have similar equations for stop mixing with obvious changes , in addition to the off diagonal term now being given as : ( @xmath285 ) , and the stop mixing matrix denoted as @xmath286 . in our framework , owing to the smallness of the off diagonal rl mixing term compared to @xmath287 , we have small stop and sbottom mixing .furthermore , the sbottom mixing angle is negligibly small and we neglect its mixing effects .we thus have @xmath288 and @xmath289 .the stop mixing angle , however , is not as small and so we include its effects . to compute the interaction vertices in the superkm basis, one could diagonalize the @xmath290 squark mass matrix .since the off - diagonal entries in our case are small , we perform an approximate leading order diagonalization of the mass matrices shown in eq .( [ msq.eq ] ) . focusing first on the @xmath291 mixing , - .this is diagonalized by the rotation & = & + & & ( c_b_r s_l ) . the mixing angle and phaseare given by & = & , + _ 32^rl & = & .the @xmath292 mixing is given similarly .we have the mass terms - , which is diagonalized by the rotation & = & + & & ( c_b_r s_r ) .[ rotbrsr.eq ] the mixing angle and phase are given by & = & , + _ 32^rr & = & .[ angphbrsr.eq ] the @xmath293 and @xmath127 mixing terms are - _ ll , rr .[ mdsllrr.eq ] the matrices that diagonalizes these , @xmath294 and @xmath121 are given analogous to eq .( [ rotbrsr.eq ] ) , the angle and phase ( @xmath295 , @xmath296 ) and ( @xmath297 , @xmath298 ) given analogous to eq .( [ angphbrsr.eq ] ) , and we will not write them down explicitly .the diagonal entries are split only by @xmath93 , and therefore this mixing is maximal in general .however , if @xmath128 , this mixing can be small .the loop functions are given by f^ll_7(x ) & = & ( + ) , + f^ll_8(x ) & = & ( - ) , + f^rl_7(x ) & = & ( + ) , + f^rl_8(x ) & = & ( - ) , + ^ll_7(x ) & = & ( + ) , + ^ll_8(x ) & = & ( - ) .f_4(x ) & = & + f_(x ) & = & - the sm box functions are given by s_0(x ) & = & - , + s_0(x_1,x_2 ) & = & + + & & .[ boxsms0.eq ] the charged - higgs and chargino box functions are given by y_1(r_,r_,s_i , s_j ) = + + + + , + y_2(r_,r_,s_i , s_j ) = .[ boxhy.eq ] from which various limiting cases can be obtained .a. g. cohen , d. b. kaplan and a. e. nelson , `` the more minimal supersymmetric standard model , '' phys . lett .b * 388 * , 588 ( 1996 ) [ hep - ph/9607394 ] .a. pomarol and d. tommasini , `` horizontal symmetries for the supersymmetric flavor problem , '' nucl .b * 466 * , 3 ( 1996 ) [ hep - ph/9507462 ] .r. barbieri , g. r. dvali and l. j. hall , `` predictions from a u(2 ) flavour symmetry in supersymmetric theories , '' phys .b * 377 * , 76 ( 1996 ) [ hep - ph/9512388 ] ; r. barbieri , l. j. hall and a. romanino , `` consequences of a u(2 ) flavour symmetry , '' phys. lett .b * 401 * , 47 ( 1997 ) [ hep - ph/9702315 ] .g. degrassi , p. gambino and g. f. giudice , `` b @xmath299 x / s gamma in supersymmetry : large contributions beyond the leading order , '' jhep * 0012 * , 009 ( 2000 ) [ hep - ph/0009337 ] ; m. carena , d. garcia , u. nierste and c. e. wagner , `` b @xmath299 s gamma and supersymmetry with large tan(beta ) , '' phys . lett .b * 499 * , 141 ( 2001 ) [ hep - ph/0010003 ] .s. bertolini , f. borzumati , a. masiero and g. ridolfi , `` effects of supergravity induced electroweak breaking on rare b decays and mixings , '' nucl .b * 353 * , 591 ( 1991 ) .j. s. hagelin , s. kelley and t. tanaka , `` supersymmetric flavor changing neutral currents : exact amplitudes and phenomenological analysis , '' nucl .b * 415 * , 293 ( 1994 ) .l. j. hall , v. a. kostelecky and s. raby , `` new flavor violations in supergravity models , '' nucl .b * 267 * , 415 ( 1986 ) ; f. gabbiani and a. masiero , `` fcnc in generalized supersymmetric theories , '' nucl .b * 322 * , 235 ( 1989 ) ; f. gabbiani , e. gabrielli , a. masiero and l. silvestrini , `` a complete analysis of fcnc and cp constraints in general susy extensions of the standard model , '' nucl .b * 477 * , 321 ( 1996 ) [ hep - ph/9604387 ] ; m. misiak , s. pokorski and j. rosiek , `` supersymmetry and fcnc effects , '' adv .high energy phys .* 15 * , 795 ( 1998 ) [ hep - ph/9703442 ] .g. l. kane , p. ko , h. b. wang , c. kolda , j. h. park and l. t. wang , `` b / d @xmath299 phi k(s ) and supersymmetry , '' phys . rev .d * 70 * , 035015 ( 2004 ) [ hep - ph/0212092 ] ; k. agashe and c. d. carone , `` supersymmetric flavor models and the b @xmath299 phi k(s ) anomaly , '' phys .d * 68 * , 035017 ( 2003 ) [ hep - ph/0304229 ] .t. affolder _ et al . _[ cdf collaboration ] , search for scalar top and scalar bottom quarks in p anti - p collisions at phys .lett . * 84 * , 5704 ( 2000 ) [ hep - ex/9910049 ] ; c. rott , `` searches for the supersymmetric partner of the bottom quark , '' hep - ex/0410007 . s. eidelman _ et al . _[ particle data group collaboration ] , `` review of particle physics , '' phys .b * 592 * , 1 ( 2004 ) .g. buchalla , a. j. buras and m. e. lautenbacher , `` weak decays beyond leading logarithms , '' rev .* 68 * , 1125 ( 1996 ) [ hep - ph/9512380 ] .g. c. branco , g. c. cho , y. kizukuri and n. oshimo , `` supersymmetric contributions to b0 - anti - b0 and k0 - anti - k0 mixings , '' phys .b * 337 * , 316 ( 1994 ) [ hep - ph/9408229 ] ; `` searching for signatures of supersymmetry at b factories , '' nucl .b * 449 * , 483 ( 1995 ) .d. becirevic _ et al ._ , `` b / d anti - b / d mixing and the b / d @xmath299 j / psi k(s ) asymmetry in general susy models , '' nucl .b * 634 * , 105 ( 2002 ) [ hep - ph/0112303 ] .a. j. buras and r. fleischer , `` quark mixing , cp violation and rare decays after the top quark discovery , '' adv .high energy phys .* 15 * , 65 ( 1998 ) [ hep - ph/9704376 ] .ko , j. h. park and g. kramer , `` b0 - anti - b0 mixing , b @xmath299 j / psi k(s ) and b @xmath299 x / d gamma in general mssm , '' eur .j. c * 25 * , 615 ( 2002 ) [ hep - ph/0206297 ] .l. randall and s. f. su , `` cp violating lepton asymmetries from b decays and their implication for supersymmetric flavor models , '' nucl .b * 540 * , 37 ( 1999 ) [ hep - ph/9807377 ] .we use the ( ichep 2004 ) average from cleo , babar and belle compiled by the heavy flavor averaging group .http://www.slac.stanford.edu/xorg/hfag/ y. grossman , y. nir and m. p. worah , `` a model independent construction of the unitarity triangle , '' phys .b * 407 * , 307 ( 1997 ) [ hep - ph/9704287 ] .g. c. branco , l. lavoura and j. p. silva , `` cp violation , '' clarendon press , oxford ( 1999 ) .k. anikeev _ et al ._ , `` b physics at the tevatron : run ii and beyond , '' [ hep - ph/0201071 ] .a. j. buras , `` weak hamiltonian , cp violation and rare decays , '' [ hep - ph/9806471 ] .a. j. buras and m. munz , `` effective hamiltonian for b @xmath299 x(s ) e+ e- beyond leading logarithms in the ndr and hv schemes , '' phys .d * 52 * , 186 ( 1995 ) [ hep - ph/9501281 ] .a. j. buras , m. misiak , m. munz and s. pokorski , `` theoretical uncertainties and phenomenological aspects of b @xmath299 x(s ) gamma decay , '' nucl .b * 424 * , 374 ( 1994 ) [ hep - ph/9311345 ] .k. g. chetyrkin , m. misiak and m. munz , `` weak radiative b - meson decay beyond leading logarithms , '' phys .b * 400 * , 206 ( 1997 ) [ erratum - ibid .b * 425 * , 414 ( 1998 ) ] [ hep - ph/9612313 ] .b. grinstein , r. p. springer and m. b. wise , `` effective hamiltonian for weak radiative b meson decay , '' phys .b * 202 * , 138 ( 1988 ) .d. a. demir and k. a. olive , `` b @xmath299 x / s gamma in supersymmetry with explicit cp violation , '' phys .d * 65 * , 034007 ( 2002 ) [ hep - ph/0107329 ] .r. barbieri and g. f. giudice , `` b @xmath299 s gamma decay and supersymmetry , '' phys .b * 309 * , 86 ( 1993 ) [ hep - ph/9303270 ] .a. ali , h. asatrian and c. greub , `` inclusive decay rate for b @xmath299 x / d + gamma in next - to - leading logarithmic order and cp asymmetry in the standard model , '' phys .b * 429 * , 87 ( 1998 ) [ hep - ph/9803314 ] ; a. l. kagan and m. neubert , `` qcd anatomy of b @xmath299 x / s gamma decays , '' eur . phys .j. c * 7 * , 5 ( 1999 ) [ hep - ph/9805303 ] .a. l. kagan and m. neubert , `` direct cp violation in b @xmath299 x / s gamma decays as a signature of new physics , '' phys .d * 58 * , 094012 ( 1998 ) [ hep - ph/9803368 ] .s. bertolini , f. borzumati and a. masiero , `` supersymmetric enhancement of noncharmed b decays , '' nucl .b * 294 * , 321 ( 1987 ) ; r. barbieri and g. f. giudice , `` b @xmath299 s gamma decay and supersymmetry , '' phys . lett .b * 309 * , 86 ( 1993 ) [ hep - ph/9303270 ] ; p. l. cho , m. misiak and d. wyler , `` @xmath300 and @xmath301 decay in the mssm , '' phys .d * 54 * , 3329 ( 1996 ) [ hep - ph/9601360 ] .j. l. hewett and j. d. wells , `` searching for supersymmetry in rare b decays , '' phys .d * 55 * , 5549 ( 1997 ) [ hep - ph/9610323 ] .s. nishida _ et al ._ [ belle collaboration ] , `` measurement of the cp asymmetry in b @xmath299 x / s gamma , '' phys .lett . * 93 * , 031803 ( 2004 ) [ hep - ex/0308038 ] .b. aubert _ et al . _ [ babar collaboration ] , `` measurement of the direct cp asymmetry in b @xmath299 s gamma decays , '' phys .lett . * 93 * , 021804 ( 2004 ) [ hep - ex/0403035 ] .a. l. kagan and j. rathsman , `` hints for enhanced b @xmath299 s g from charm and kaon counting , '' [ hep - ph/9701300 ] .j. kaneko _ et al ._ [ belle collaboration ] , `` measurement of the electroweak penguin process b @xmath299 x / s l+ l- , '' phys .lett . * 90 * , 021801 ( 2003 ) [ hep - ex/0208029 ] .m. beneke , g. buchalla , m. neubert and c. t. sachrajda , `` qcd factorization in b @xmath299 pi k , pi pi decays and extraction of wolfenstein parameters , '' nucl .b * 606 * , 245 ( 2001 ) [ hep - ph/0104110 ] .m. beneke and m. neubert , `` qcd factorization for b @xmath299p p and b @xmath299 p v decays , '' nucl .b * 675 * , 333 ( 2003 ) [ hep - ph/0308039 ] .b. aubert _ et al . _[ babar collaboration ] , `` measurement of sin(2beta ) in b0 @xmath299 phi k0(s ) .( ( b ) ) , '' hep - ex/0207070 .k. abe _ et al . _ [ belle collaboration ] , `` an improved measurement of mixing - induced cp violation in the neutral b meson system , '' hep - ex/0207098 .t. hahn and m. perez - victoria , `` automatized one - loop calculations in four and d dimensions , '' comput .commun . * 118 * , 153 ( 1999 ) [ hep - ph/9807565 ] .

<S> we discuss b - meson signatures of a supersymmetric u(2 ) flavor model , with relatively light ( electroweak scale masses ) third generation right - handed scalars . </S><S> we impose current @xmath0 and @xmath1 meson experimental constraints on such a theory , and obtain expectations for , , , , mixing and the dilepton asymmetry in @xmath2 . </S><S> we show that such a theory is compatible with all current data , and furthermore , could reconcile the apparent deviations from standard model predictions that have been found in some experiments . </S><S> msuhep-041008 + hep - ph/0410181 + </S>

astro-ph0510245

secular evolution of galaxies includes too many processes to review thoroughly here . instead , a selection of processes with contemporary interests will be discussed .they include bar evolution in the present - day universe and at intermediate to high z , and disk formation at high z , where extremely clumpy structure seems to be common .a brief summary of secular evolution effects over a hubble time is in section [ sect - evol ] .a recent review of secular evolution leading to central mass concentrations and pseudobulges was in kormendy & kennicutt ( 2004 ) .the disk and halo of a galaxy exert negative torques on a bar and cause the bar to get stronger , longer , and slow down over time ( weinberg 1985 ; debattista & sellwood 1998 , 2000 ; athanassoula , lambert , & dehnen 2005 ; athanassoula 2005 ) . bars also change their shape as more orbital families get populated .they can thicken and develop a peanut - shaped profile perpendicular to the disk as a result of a resonance between the corotating pattern period and the vertical oscillation period ( combes & sanders 1981 ) .they can become more squared off at the ends , or boxy , as a result of increasing populations of 4:1 orbits .orbit families are described in sellwood & wilkinson ( 1993 ) and athanassoula ( 2005 ) .gas accretion to the center and subsequent star formation can lead to a dense nucleus , which either forms an ilr or shifts the existing ilr to a larger radius .tightly wound starburst spirals or rings may appear near these inner resonances and secondary bars can form inside ( e.g. , englmaier & shlosman 2004 ) .perhaps the most startling development of the 1990 s was that the growth of the ilr radius with time depopulates bar - supporting orbits ( x1 orbits ) , which are the elongated orbits parallel to the bar out to nearly the corotation radius .this depopulation can dissolve a bar completely , forming a bulge - like concentration in the center ( hasan & norman 1990 ; friedli & benz 1993 ; hasan , pfenniger , & norman 1993 ) .norman , sellwood , & hasan ( 1996 ) ran 2d models of bar dissolution with 50,000 star particles , no gas and no active halo .they found that by the time 3 - 5% of the total disk mass shrank to the galaxy core , the bar was completely dissolved .they also confirmed this result with a 3d model using 200,000 star particles and no active halo .shen & sellwood ( 2004 ) also ran 3d star models with no active halo and found in addition that bar weakening occurs in 2 phases , an initially fast stage where the bar - supporting orbits are scattered , and then a slow stage where the whole potential changes .more importantly , they pointed out that the central mass needed to destroy a bar is much larger , by a factor of 10 or more , than the nuclear black hole mass , concluding that bar destruction is in fact unlikely in real galaxies .they showed further that a dispersed central mass , as might result from gas inflow and star formation , required an even higher fraction ( e.g. , @xmath2% of the disk mass ) than a point - like distribution , making the possibility that gas accretion can eventually destroy a bar ( i.e. , because of its central mass ) seem even more remote .athanassoula , lambert , & dehnen ( 2005 ) studied bars with active halos and found that if the halo was massive , the bars survived even with a central mass of 10% of the disk mass . bars in low - mass active halos were still destroyed , like before .the difference is that an active halo continuously removes angular momentum from the bar , strengthening it by populating x1 orbits .then the deleterious effect of the central mass is reduced .halos also promote spiral arm formation by the removal of disk angular momentum ( fuchs 2004 ) .while bars are strengthened by a loss of angular momentum to a disk or halo , thereby resisting dissolution at the ilr from a central concentration , bars also weaken by the absorption of angular momentum from gas , and thereby become more prone to dissolution ( bournaud & combes 2002 ) .gas within the bar radius shocks at the bar - leading dust lanes .the asymmetric pressures in these shocks cause the gas to lose orbital angular momentum , eventually migrating to the center .the shock is asymmetric only if there is energy dissipation ; otherwise the backward impulse the gas feels when entering the shock is compensated by a forward impulse it feels when leaving the shock .from the point of view of the stars , there is a gravitational attraction to the dust lane , which is shifted forward of the orbital apocenter . because stars in bar orbits spend a longer time at apocenter than at pericenter ( on the minor axis ) , the time - integrated gravitational force from the gas produces a net positive torque on the stars .the resulting angular momentum gain causes the elongated stellar orbits to open up into circular orbits , which have higher angular momentum for the same energy ( the same major axis length ) .circularization of orbits makes the bar ultimately dissolve .recent simulations of this effect are in bournaud , combes , & semelin ( 2005 ) .dissolution of a bar by gas torques requires a large amount of gas in the bar region , as essentially all of the angular momentum that was removed from the stars to make the bar in the first place has to be returned from the gas .if the total mass fraction of gas in the bar region is small , then the circularization of bar orbits may be minimal .block et al .( 2002 ) propose that continued accretion of gas onto the galaxy disk can cause a second or third event of bar dissolution if a new bar forms in the mean time . in this way, bars can dissolve but still have an approximately constant fraction over a large portion of the hubble time , as observed by sheth , et al .( 2003 ) , elmegreen , elmegreen , & hirst ( 2004 ) , jogee et al .( 2004 ) , and zheng , et al .( 2005 ) .what is the observational evidence for bar dissolution and what happens to old bars ?the evidence for bar circularization in more centrally concentrated galaxies ( das et al .2003 ) does not actually indicate that bars dissolve .dissolution may take a very long time and the observed circularization could be only the first step .there is good evidence for some bar dissolution , however , with the resulting structure a lens ( sandage 1961 ; freeman 1975 ; kormendy 1977 ) , or perhaps a pseudo - bulge ( kormendy & kennicutt 2004 ) .a lens is not a spheroid or part of the outer exponential disk .it is a stellar distribution in the inner region with a rather shallow brightness profile and a sharp cutoff at mid - disk radius . outside the lens ,the main disk exponential continues .kormendy ( 1979 ) suggested that lenses are dissolved bars .that is , s0 galaxies with lenses formerly had bars .of the 121 low - inclination , bright sb0-sbd type galaxies in kormendy ( 1979 ) , 54% of early bar types , sb0-sba , have a lens , 0% of late bar types sbab - sbc have a lens , and 16% of the sample have both a bar and a lens .the lens looks like it is coming from the bar in these latter cases because the lens size is the same as the bar size .the lenses in early - type non - barred galaxies are also the same size as the inner rings in barred ( non - lens ) galaxies for the same absolute magnitude galaxy ; these rings are another measure of bar size .lens colors are generally the same as bar colors too .some bars even `` look '' like they are dissolving into a lens because the bar spreads out into a circular shape at the ends ( e.g. , ngc 5101 ) .the axial ratio distribution of lenses weakly suggests they are thick ( triaxial ) .the presence of lenses in about half of the early type barred galaxies ( as mentioned above ) but in only a small fraction of early type non - barred galaxies suggests further that bars dissolve very slowly into lenses .the most advanced stages of dissolution are in the early type bars , which are denser and more advanced in total evolution than late - type bars .there are very few simulations of bar dissolution into a lens .a pure star simulation by debattista & sellwood ( 2000 ) shows something like a stellar ring forming at the end of the bar .stars develop complex orbits over time , migrating between inside and outside the bar as the lens builds up .we have found that barred galaxies in the early universe are slightly smaller than barred galaxies today , by a factor of @xmath3 in the exponential scale length ( elmegreen , elmegreen , & hirst 2004 ) .such a measurement avoids problems with cosmological surface brightness dimming .the same small sizes are observed for spiral galaxies in general , based on a sample of 269 spirals in the hubble ultra deep field ( elmegreen , et al .the presence of spirals in these high - z disk galaxies suggests that the disks are relatively massive compared to the halos at that time , unlike the inner disks of today s galaxies .this suggests , in turn , that galaxies build up from the inside out , as proposed many years ago ( e.g. , larson 1976 ) .with such a process , the dynamical time for evolution in a disk , which is proportional to the inverse square root of total density , will be shorter in the early universe than it is today . that is , the active parts of galaxy disks are gradually evolving toward lower and lower densities and longer evolutionary timescales ( a process related to `` downsizing '' e.g. tanaka , et al .2005 ) .the implication of increasing dynamical time in an aging universe is that bar formation and secular dissolution should have been much faster at high redshift , in approximate proportion to the bar size .also , the higher gas abundance in young galaxies would have led to a more rapid dissolution then too , by the bournaud & combes ( 2002 ) process .thus some of the very early stages of bulge formation inside a disk ( as opposed to disk accretion around a pre - existing bulge ) could have involved evolution from a bar .there is no direct evidence for this yet , however ..2 in cdm cosmology are plotted versus redshift , @xmath4 : the universe age and look back time ( top left ) , the physical size of a region subtending 10 pixels of the hst acs camera ( top right ) , the redshift range , @xmath5 , spanning 0.5 gy ( lower left ) , and the apparent surface brightness of a region having an intrinsic surface brightness equal to the standard freeman value ( lower right).,width=384 ] observations of the youngest bars are difficult for several reasons .figure 1 plots various observables as a function of redshift , using the wmap cosmology ( spergel et al .2003 ) and equations in carroll , press , & turner ( 1992 ) . beyond @xmath6 , galaxies are only a few gy old ( top left ) and a normal size bar has only rotated around only a few times .thus the bar will be dynamically young and likely to appear different , perhaps more irregular or round .also , the bar age is likely to be younger than the gas consumption time , which is typically @xmath7 orbit times ( e.g. , kennicutt 1998 ) , so the bars will still be gas - rich , like some late - type bars today .this makes them irregular also because of dust obscuration and star formation .second ( fig . 1 upper right ) , between @xmath8 and 4.6 , objects smaller than 2 kpc in physical size , such as a bar thickness , will be smaller than 10 pixels on the hst acs camera .this makes it difficult to resolve internal structure .bars were not easily seen in the hubble deep fields north and south ( e.g. , abraham et al . 1999 ) because of the factor of 2 lower angular resolution of the wfpc2 camera .only the largest bars could be seen in the hdf images ( sheth et al .2003 ) .third ( lower left ) , beyond @xmath9 the redshift range corresponding to the duration of bar formation exceeds @xmath10 , which means that the bar formation process gets stretched out over a wide range of z. this is good in the sense that different stages in bar formation may be seen distinctly , but it is bad in the sense that bars beyond @xmath9 will be incompletely formed and perhaps unrecognizable .fourth ( lower right ) , beyond @xmath11 disks with the `` standard '' rest - b band surface brightness of today s freeman - disk , 21.6 mag arcsec@xmath12 , are almost too faint to see .the @xmath13 noise limiting surface brightness of the hubble ultra deep field is @xmath14 mag arcsec@xmath12 at i band , which is the deepest passband ( elmegreen et al .at @xmath15 , a freeman disk will have a central surface brightness of @xmath16 mag arcsec@xmath12 , which requires averaging over areas at least as large as @xmath17 square pixels to observe at the @xmath13 limit .thus only large bright regions can be seen .fortunately the intrinsic surface brightnesses of disks are higher at high redshift because of enhanced star formation , unless the extinction is also high .whether this star formation is enough to reveal bars at this early epoch is unknown .surface brightness limitations have already diminished the number of spiral galaxies that can be seen in the udf ( elmegreen et al .2005a ) .the relatively late formation of disk galaxies compared to spheroids aids somewhat in the study of bar formation because it means that bars in disks form late too , when they can be more easily observed . nevertheless , there are still considerable problems recognizing and observing young bars. the apparent loss of bars beyond @xmath6 could be partly from these selection effects affects .most bars look different at @xmath18 than they do today .high - z bars are typically blue and often clumpy , like giant star - forming regions ( elmegreen , elmegreen , & hirst 2004 ). sometimes they are off - center .perhaps this irregularity is not surprising since young disks should still be gas - rich .however , the observation implies something new about bar formation : that it could occur in gas - rich or pure - gas disks with gas dissipation playing an important role .the bar formation models and simulations existing today are all in pure stellar or dominantly stellar systems , where orbital motions and resonances are important . in a gas rich disk ,bar formation is more like a dissipative instability .recent simulations of bar formation in a pure - gas disk are in kaufmann et al .most galaxies in the hubble deep fields and in the ultra deep field are very clumpy .even elliptical galaxies are clumpy in their cores ( elmegreen , elmegreen , & ferguson 2005 ) . a compendium of clumpy structures is shown in figure 2 ..2 in the clumps in galaxies in the udf contain up to @xmath19 m@xmath1 and are typically bluer than the rest of the disk or elliptical galaxy .thus they are probably young and still forming stars .they are distinct from spiral galaxy bulges , which are redder and always centrally placed .an examination of the colors and magnitudes of clumps in 10 highly clumpy galaxies , called `` clump clusters , '' suggests their ages are several @xmath20 years ( elmegreen & elmegreen 2005 ) .this is @xmath7 times longer than the dynamical times of the clumps , meaning they are gravitationally bound like star clusters , but it is not so young that they may be interpreted as currently star - bursting .moreover , the clumps have internal densities that are only a factor of @xmath21 above the local tidal densities in their disks , which means they should disperse relatively soon because of tidal forces .this is probably the process of disk formation ; i.e. , through the dissolution of giant clumps ( see also elmegreen , et al .2005b ) .the origin of the clumps is unknown at the present time .they could be formed in the disks of spirals and clump cluster galaxies or in the gaseous components of elliptical galaxies as a result of gravitational instabilities ( noguchi 1999 ; immeli et al .2004a , b ) .they could also come in as dwarf - like galaxies from the surrounding field .similar but isolated clumps are actually seen in these fields ( elmegreen & elmegreen 2005 ; elmegreen , et al .evolutionary effects may be divided into two types , non - secular , which result from one - time events , and secular , which result from a progressive slow change .non - secular events include interactions , major mergers , and gas stripping during motion through a galaxy cluster .strong interactions form bars ( noguchi 1988 ; gerin , combes , & athanassoula 1990 ; sundin & sundelius 1991 ; berentzen et al .even binary companions promote bar formation and may shift the galaxy to an earlier hubble type because of torqued inflow ( elmegreen , elmegreen , & bellin 1990 ) .bars are also more common in perturbed galaxies ( varela et al .major mergers form ellipticals ( toomre & toomre 1972 ; barnes & hernquist 1992 ; schweizer 1998 ) , and in the process , they form at least the youngest halo globular clusters ( ashman & zepf 1992 ) , and may produce counter - rotating cores ( kannappan & fabricant 2001 ) .minor mergers are non - secular events although they may occur more or less continuously .they thicken spiral disks ( walker et al . 1996 ; schwarzkopf & dettmar 2000ab ; bertschik & burkert 2002 ; gilmore , wyse , & norris 2002 ) and may also lead to counter - rotating cores .interactions may form dwarf galaxies in tidal tails ( zwicky 1959 ; barnes & hernquist 1992 ; elmegreen , kaufman , & thomasson 1993 ; duc , bournaud , & masset 2004 ) .secular interactions include accretion in the form of residual hierarchical build - up from outside the galaxy , continuous accretion of dwarfs ( e.g. , lewis et al .2004 ) , galaxy harassment ( e.g. , moore et al . 1996 ) , accretion of gas ( e.g. , bournaud et al . 2005 ) , globular cluster accretion from dwarfs ( e.g. , bellazzini , ibata , & ferraro 2004 ) , halo streams ( e.g. , ferguson et al . 2005 ) , and so on .secular changes are also driven by internal processes , such as bar and spiral torques , which lead to inner disk accretion and outer disk spreading ( see review in pfenniger 2000 ) .gas accretion at resonances makes star - forming rings ( see review in buta & combes 1996 ) .scattering and vertical resonances thicken disks .bars dissolve into lenses and ( maybe ) bulges , as discussed above . and of course , gas converts into stars , while gas viscosity redistributes the disk gas .secular changes among galaxies with different average densities cause early hubble types to age more quickly than late hubble types .most evolutionary processes are faster at higher density .as time goes by , the transition between early and late types moves , converting little - evolved galaxies , which still have a lot of gas and star formation , to highly evolved galaxies , which are inactive and fading .as galaxies progress from late to early types , they become more centrally concentrated , have less gas , less star formation , hotter disks , redder populations , and more metals . galaxy mass and size increase over time by hierarchical buildup at @xmath22 , and they probably still increase over time today for late - type spirals as a result of continued accretion and coalescence .still , the average mass and size of a galaxy that is actively forming stars tends to decrease with time ( `` downsizing '' ) .this is because the most massive galaxies tend also to be the densest , and they completed their star formation first . also with time , galaxies become more and more isolated , which means that flocculent spiral types and perhaps low surface brightness disks appear relatively late .schweizer , f. 1998 , in galaxies : interactions and induced star formation , saas - fee advanced course 26 , swiss society for astrophysics and astronomy , xiv , eds .r. c. kennicutt , jr . , f. schweizer , j. e. barnes , d. friedli , l. martinet , & d. pfenniger , p. 105

<S> simulations and observations of galactic bars suggest they do not commonly evolve into bulges , although it is possible that the earliest bars formed bulges long ago , when galaxies were smaller , denser , and had more gas . </S><S> the most highly evolved of today s bars may become lenses over a hubble time . </S><S> most galaxies in the early universe are extremely clumpy , with @xmath0 m@xmath1 blue clumps that resemble in color and magnitude the isolated field objects nearby . </S><S> the presence of blue and irregular bars at high redshift suggests that some bars formed primarily in the gas phase accompanied by giant starbursts , rather than in pure stellar disks like most models . </S><S> secular and non - secular processes that cause galaxies to evolve are summarized . </S><S> barred galaxies , spiral galaxies , galaxy evolution </S>

math0310121

in this section we give background information on posets , coxeter groups , bruhat order and polytopes . the poset terminology and notation here generally agree with @xcite . throughout this paper , all posets considered are finite .let @xmath3 be a poset .given @xmath12 , say @xmath0 _ covers _ @xmath1 and write `` @xmath13 '' if @xmath14 and if there is no @xmath15 with @xmath16 .given @xmath17 , define @xmath18 . if @xmath3 has a unique minimal element , it is denoted @xmath19 , and if there is a unique maximal element , it is called @xmath20 .the dual of @xmath3 is the same set of elements with the reversed partial order .an _ induced subposet _ of @xmath3 is a subset @xmath21 , together with the partial order induced on @xmath22 by @xmath23 .often this is referred to simply as a _subposet_. a _ chain _ is a totally ordered subposet of @xmath3 .a chain is _ maximal _ if it is not properly contained in any other chain .a poset is _ graded _ if every maximal chain has the same number of elements .function on a graded poset @xmath3 is the unique function such that @xmath24 for any minimal element @xmath0 , and @xmath25 if @xmath13 . given two posets @xmath3 and @xmath26 , form their _ product _ @xmath27 .the underlying set is the ordered pairs @xmath28 with @xmath29 and @xmath30 , and the partial order is @xmath31 if and only if @xmath32 and @xmath33 . the product of @xmath3 with a two - element chainis called the pyramid @xmath34 .a poset @xmath26 is an _ extension _ of @xmath3 if the two are equal as sets , and if @xmath35 implies @xmath36 .the _ join _@xmath37 of two elements @xmath0 and @xmath1 is the unique minimal element in @xmath38 , if it exists .the _ meet _ @xmath39 is the unique maximal element in @xmath40 if it exists .a poset is called a _ lattice _ every pair of elements @xmath0 and @xmath1 has a meet and a join . given posets @xmath3 and @xmath26 , a map @xmath41 is _ order - preserving _ if @xmath35 implies @xmath42 .consider the set @xmath43 of fibers of @xmath44 , and define a relation @xmath45 on @xmath46 by @xmath47 if there exist @xmath48 and @xmath49 such that @xmath50 .if @xmath45 is a partial order , @xmath46 is called the _ fiber poset _ of @xmath3 with respect to @xmath44 . in this case, there is a surjective order - preserving map @xmath51 given by @xmath52 , and an injective order - preserving map @xmath53 such that @xmath54 .call @xmath44 an _ order - projection _ if it is order - preserving and has the following property : for all @xmath55 in @xmath26 , there exist @xmath56 with @xmath57 and @xmath58 . in particular , an order - projection is surjective .[ o - pproperties ] let @xmath41 be an order - projection. then 1 .@xmath46 is a fiber poset .@xmath59 is an order - isomorphism .assertion ( i ) is just the statement that @xmath45 is a partial order .the reflexive property is trivial .let @xmath60 and @xmath61 for @xmath62 .if @xmath63 and @xmath64 , we can find @xmath65 with @xmath66 , @xmath67 , @xmath68 and @xmath69 . because @xmath44 is order - preserving , @xmath55 and @xmath70 , so @xmath71 and therefore @xmath72 .thus the relation is anti - symmetric . to show that @xmath45 is transitive , suppose @xmath63 and @xmath73 .then there exist @xmath74 , @xmath75 and @xmath76 with @xmath57 , @xmath67 and @xmath77 such that @xmath78 and @xmath79 . because @xmath44 is order - preserving , we have @xmath80 . because @xmath44 is an order - projection , one can find @xmath81 with @xmath82 and @xmath83 .so @xmath84 .since @xmath44 is surjective , @xmath59 is an order - preserving bijection .let @xmath55 in @xmath26 . then , because @xmath44 is an order - projection , there exist @xmath56 with @xmath57 and @xmath58 .so @xmath85 in @xmath46 .thus @xmath86 is order - preserving .here we give the definition of a coxeter group and the bruhat order ( or strong order ) on a coxeter group , as well as one alternate characterization of the bruhat order .further information on coxeter groups and the bruhat order can be found for example in @xcite . a _ coxeter system _ is a pair @xmath87 , where @xmath88 is a group, @xmath22 is a set of generators , and @xmath88 is given by the presentation @xmath89 for all @xmath90 , with the requirements that : 1 .@xmath91 for all @xmath92 , and 2 .@xmath93 for all @xmath94 in @xmath22 .we use the convention that @xmath95 for any @xmath0 , so that @xmath96 is a trivial relation .the coxeter system is called _ universal _ if @xmath97 for all @xmath98. we will refer to a `` coxeter group , '' @xmath88 with the understanding that a generating set @xmath22 has been chosen such that @xmath87 is a coxeter system . inwhat follows , @xmath88 or @xmath87 will always refer to a fixed coxeter system , and @xmath99 will be an element of @xmath88 .examples of finite coxeter groups include the symmetric group , other weyl groups of root systems , and symmetry groups of regular polytopes. readers not familiar with coxeter groups should concentrate on the symmetric group @xmath100 of permutations of the numbers @xmath101 .in particular , some of the figures will illustrate the case of @xmath102 .let @xmath103 be the transposition @xmath104 , let @xmath105 and @xmath106 .then @xmath107 is a coxeter system with @xmath108 and @xmath109 .call a word @xmath110 with letters in @xmath22 a _ reduced _ word for @xmath99 if @xmath7 is as small as possible . call this @xmath7the _ length _ of @xmath99 , denoted @xmath111 .we will use the symbol `` 1 '' to represent the empty word , which corresponds to the identity element of @xmath88 . given any words @xmath112 and @xmath113 and given words@xmath114 with @xmath115 and @xmath116 with @xmath117 , the words @xmath118 and @xmath119 both stand for the same element .such an equivalence is called a _braid move_. a theorem of tits says that given any two reduced words @xmath120 and @xmath121 for the same element , @xmath120 can be transformed into @xmath121 by a sequence of braid moves .there are several equivalent definitions of bruhat order .see @xcite for a discussion of the equivalent formulations .one definition is by the `` subword property . ''fix a reduced word @xmath110 . then @xmath122if and only if there is a reduced subword @xmath123 corresponding to @xmath8 such that @xmath124 we will write @xmath125 for @xmath122 when the context is clear . bruhat order is ranked by length . the element @xmath99 covers the elements which can be represented by reduced words obtained by deleting a single letter from a reduced word for @xmath99 .we will need the `` lifting property '' of bruhat order , which can be proven easily using the subword property .[ lifting ] if @xmath126 and @xmath127 have @xmath128 and @xmath129 , then the following are equivalent : 1 .@xmath130 2 .@xmath131 3 .@xmath132 a _ ( convex ) polytope _@xmath133 is the convex hull of a finite number of points , or equivalently a bounded set which is the intersection of a finite number of closed halfspaces .affine subspace _ of @xmath134 is a subset of @xmath134 which can be written as @xmath135 for some linear subspace @xmath136 and some vector @xmath137 .the _ affine span _ of a set @xmath138 is the intersection of all affine subspaces containing @xmath22 .the _ dimension _ of a polytope @xmath133 is the dimension of its affine span .a _ hyperplane _ is the set of points @xmath0 satisfying @xmath139 for some fixed @xmath120 and @xmath121 .it is called a _ supporting _ hyperplane of a polytope @xmath133 if @xmath140 for every point in @xmath133 or if @xmath141 for every point in @xmath133 .a _ face _ of @xmath133 is any intersection of @xmath133 with a supporting hyperplane . in particular@xmath142 is a face of @xmath133 , and any face of @xmath133 is itself a convex polytope . by convention, @xmath133 is also a face of @xmath133 .a _ facet _ of @xmath133 is a face whose dimension is one less than the dimension of @xmath133 .the _ face lattice _ of @xmath133 is the set of faces of @xmath133 partially ordered by inclusion , and this partial order can be shown to be a lattice .two polytopes are of the same _ combinatorial type _ if their face lattices are isomorphic as posets .we will need two geometric constructions on polytopes , the pyramid operation @xmath143 and the vertex - shaving operation @xmath144 . given a polytope @xmath3 of dimension @xmath11 ,@xmath34 is the convex hull of the union of @xmath3 with some vector @xmath8 which is not in the affine span of @xmath133 .this is unique up to combinatorial type and the face poset of @xmath145 is just the pyramid of the face poset of @xmath133 . consider a polytope @xmath3 and a chosen vertex @xmath8 .let @xmath146 be a hyperplane that separates @xmath8 from the other vertices of @xmath3 .in other words , @xmath147 and @xmath148 for all vertices @xmath149 . then the polytope @xmath150 is called the _ shaving _ of @xmath3 at @xmath8 .this is unique up to combinatorial type .every face of @xmath3 , except @xmath8 , corresponds to a face in @xmath151 and , in addition , for every face of @xmath3 strictly containing @xmath8 , there is an additional face of one lower dimension in @xmath151 . in section [ cw ]we describe how this operator can be extended to regular cw spheres , and in section [ building ] we describe the corresponding operator on posets .further information on polytopes can be found for example in @xcite .this section provides background material on finite cw complexes and pl topology which will be useful in section [ zipping ] .more details about cw complexes , particularly as they relate to posets , can be found in @xcite .additional details about pl topology can be found in @xcite .a set of @xmath152 points in @xmath134 is _ affinely independent _ if the smallest affine subspace containing them has dimension @xmath153 .simplex _ is a polytope which is the convex hull of an affinely independent set @xmath154 of points .the faces of the simplex are the convex hulls of subsets of @xmath154 .a _ geometric simplicial complex _@xmath155 is a finite collection of simplices ( called _ faces _ of the complex ) such that 1 . if @xmath156 and @xmath157 is a face of @xmath158 , then @xmath159 .2 . if @xmath160 then @xmath161 is a face of @xmath158 and of @xmath157 . the zero - dimensional faces are called _ vertices_. the _ underlying space _@xmath162 of @xmath155 is the union in @xmath134 of the faces of @xmath155 .an _ abstract simplicial complex _@xmath155 on a finite _ vertex - set _@xmath163 is a collection of subsets of @xmath163 called _ faces _ , with the following properties : 1 .every singleton is a face .any subset of a face is another face .given a geometric simplicial complex @xmath155 with vertices @xmath163 , the collection @xmath164 is an abstract simplicial complex .the process can be reversed : given an abstract simplicial complex @xmath155 , there is a construction which produces a geometric simplicial complex whose underlying abstract simplicial complex is exactly @xmath155 . this _ geometric realization _ is unique up to homeomorphism , so it makes sense to talk about the topology of an abstract simplicial complex .two geometric simplicial complexes are _ combinatorially isomorphic _ if their underlying abstract simplicial complexes are isomorphic . if two complexes are combinatorially isomorphic then their underlying spaces are homeomorphic , but the converse is not true. given simplicial complexes @xmath155 and @xmath165 , say @xmath165 is _ subdivision _ of @xmath155 if @xmath166 and if every face of @xmath165 is contained in some face of @xmath155 .a simplicial complex is a _pl @xmath11-sphere _ if it admits a simplicial subdivision which is combinatorially isomorphic to some simplicial subdivision of the boundary of a @xmath167-dimensional simplex .a simplicial complex is a _pl @xmath11-ball _ if it admits a simplicial subdivision which is combinatorially isomorphic to some simplicial subdivision of a @xmath11-dimensional simplex .we now quote some results about pl balls and spheres .some of these results appear topologically obvious but , surprisingly , not all of these statement are true with the `` pl '' deleted .this is the reason that we introduce pl balls and spheres , rather than dealing with ordinary topological balls and spheres . statement ( iii )is known as newman s theorem .* theorem 4.7.21 ) [ pl tricks ] 1 . given two pl @xmath11-ballswhose intersection is a pl @xmath168-ball lying in the boundary of each , the union of the two is a pl @xmath11-ball .given two pl @xmath11-balls whose intersection is the entire boundary of each , the union of the two is a pl @xmath11-sphere .3 . the closure of the complement of a pl @xmath11-ball embedded in a pl @xmath11-sphere is a pl @xmath11-ball. given two abstract simplicial complexes @xmath155 and @xmath165 , let @xmath169 be the _ join _ of @xmath155 and @xmath165 , a simplicial complex whose vertex set is the disjoint union of the vertices of @xmath155 and of @xmath165 , and whose faces are exactly the sets @xmath170 for all faces @xmath171 of @xmath155 and @xmath172 of @xmath165 .let @xmath173 stand for a pl @xmath11-ball , and let @xmath174 be a pl @xmath11-sphere .( * ? ? ? * proposition 2.23 ) [ joins ] @xmath175 here @xmath176 stands for pl homeomorphism , a stronger condition than homeomorphism which requires a compatibility of pl - structures as well .the point is that @xmath177 is a pl ball , etc . given a poset @xmath3 the order complex @xmath178 is the abstract simplicial complex whose vertices are the elements of @xmath3 and whose faces are the chains of @xmath3 .the order complex of an interval @xmath2 $ ] will be written @xmath179 $ ] , rather than @xmath180)$ ] , and similarly @xmath181 instead of @xmath182 .when @xmath3 is a poset with a @xmath19 and a @xmath20 , by convention , statements about the topology of @xmath3 are understood to apply to the order complex of @xmath183 .thus for example , the statement that `` @xmath3 is a pl sphere '' means that @xmath184 is a pl sphere .the following proposition follows immediately from ( * ? ? ?* theorem 4.7.21(iv ) ) : [ interval ] if @xmath3 is a pl sphere then any interval @xmath2_p$ ] is a pl sphere . an _ open cell _ is any topological space isomorphic to an open ball .cw complex _@xmath185 is a hausdorff topological space with a decomposition as a disjoint union of cells , such that for each cell @xmath186 , the homeomorphism mapping an open ball to @xmath186 is required to extend to a continuous map from the closed ball to @xmath185 . the image of this extended map is called a _ closed cell _ , specifically the _ closure _ of @xmath186 .face poset _ of @xmath185 is the set of closed cells , together with the empty set , partially ordered by containment .@xmath7-skeleton _ of @xmath185 is the union of the closed cells of dimension @xmath7 or less .a cw complex is _ regular _ if all the closed cells are homeomorphic to closed balls .call @xmath3 a cw poset if it is the face poset of a regular cw complex @xmath185 .it is well known that in this case @xmath185 is homeomorphic to @xmath187 .the following theorem is due to bjrner @xcite .[ cw thm ] a non - trivial poset @xmath3 is a cw poset if and only if 1 .@xmath3 has a minimal element @xmath188 , and 2 . for all @xmath189 , the interval @xmath190 $ ]is a sphere .given a cw poset , bjrner constructs a complex @xmath191 recursively by constructing the @xmath192-skeleton , and then attaching @xmath7-cells in a way that agrees with the order relations in @xmath3 .the polytope operations @xmath143 and @xmath144 can also be defined on regular cw spheres .both operations preserve pl sphericity by theorem [ pl tricks](ii ) .we give informal descriptions which are easily made rigorous .consider a regular cw @xmath11-sphere @xmath185 embedded as the unit sphere in @xmath193 .the new vertex in the @xmath143 operation will be the origin .each face of @xmath185 is also a face of @xmath194 and for each nonempty face @xmath171 of @xmath185 there is a new face @xmath195 of @xmath194 , described by @xmath196 the set @xmath197 is also a face of @xmath194 ( the `` base '' of the pyramid ) where @xmath198 is the point at infinity which makes @xmath199 a @xmath200-sphere .consider a regular cw sphere @xmath185 and a chosen vertex @xmath8 .adjoin a new open cell to make @xmath201 , a ball of one higher dimension .choose @xmath22 to be a small sphere @xmath202 , such that the only vertex inside the sphere is @xmath8 and the only faces which intersect @xmath22 are faces which contain @xmath8 .( assuming some nice embedding of @xmath185 in space , this can be done . )then @xmath203 is the boundary of the ball obtained by intersecting @xmath201 with the set @xmath204 . as in the polytope case , this is unique up to combinatorial type .every face of @xmath185 , except @xmath8 , corresponds to a face in @xmath203 , and for every face of @xmath185 strictly containing @xmath8 , there is an additional face of one lower dimension in @xmath203 . given a poset @xmath3 with @xmath19 and @xmath20 ,call @xmath3 a regular cw sphere if @xmath205 is the face poset of a regular cw complex which is a sphere . by theorem [ cw thm ], @xmath3 is a regular cw sphere if and only if every lower interval of @xmath3 is a sphere . in light of proposition [ interval ] , if @xmath3 is a pl sphere , then it is also a cw sphere , but not conversely .section [ building ] describes a construction on posets which corresponds to @xmath144 .in this section we give the definition of eulerian posets , flag f - vectors , flag h - vectors , and the cd - index , and quote results about the cd - indices of polytopes . the _ mbius function _@xmath206 is defined recursively by setting @xmath207 for all @xmath17 , and @xmath208 a poset @xmath3 is _ eulerian _ if @xmath209 for all intervals @xmath2 \subseteq p$ ] . for a survey of eulerian posets ,see @xcite .verma @xcite gives an inductive proof that bruhat order is eulerian , by counting elements of even and odd rank .rota @xcite proved that the face lattice of a convex polytope is an eulerian poset ( see also @xcite ) .more generally , the face poset of a cw sphere is eulerian . in @xcite, bjrner showed that bruhat intervals are cw spheres .let @xmath3 be a graded poset , rank @xmath210 , with a minimal element @xmath188 and a maximal element @xmath211 . for a chain @xmath212 in @xmath213 , define @xmath214 . let @xmath215 $ ] denote the set of integers @xmath101 .for any @xmath216 $ ] , define @xmath217 the function @xmath218}\rightarrow { \mathbb n}$ ] is called the _ flag f - vector _ , because it is a refinement of the _ f - vector _ , which counts the number of elements of each rank .define a function @xmath219}\rightarrow { \mathbb n}$ ] by @xmath220 the function @xmath221 is called the _ flag h - vector _ of @xmath3 because of its relation to the usual _ h - vector_. bayer and billera @xcite proved a set of linear relations on the flag f - vector of an eulerian poset , called the generalized dehn - sommerville relations .they also proved that the generalized dehn - sommerville relations and the relation @xmath222 are the complete set of affine relations satisfied by flag f - vectors of all eulerian posets .let @xmath223 be the vector space of _ ab - polynomials _ , that is , polynomials over non - commuting variables @xmath120 and @xmath121 with integer coefficients .subsets @xmath216 $ ] can be represented by monomials @xmath224 , where @xmath225 if @xmath226 and @xmath227 otherwise .define ab - polynomials @xmath228 and @xmath229 to encode the flag f - vector and flag h - vector respectively .@xmath230 } \alpha_p(s)u_s\\ \psi_p(a , b ) & : = & \sum_{s\subseteq [ n ] } \beta_p(s)u_s.\end{aligned}\ ] ] the polynomial @xmath229 is commonly called the _ ab - index_. there is no standard name for @xmath228 , but here we will call it the _ flag index_. it is easy to show that @xmath231 . let @xmath232 and @xmath233 in @xmath223 .the flag f - vector of a graded poset @xmath3 satisfies the generalized dehn - sommerville relations if and only if @xmath234 can be written as a polynomial in @xmath235 and @xmath11 with integer coefficients , called the _ cd - index _ of @xmath3 .this surprising fact was conjectured by j. fine and proven by bayer and klapper @xcite .the cd - index is _ monic _ , meaning that the coefficient of @xmath236 is always 1 .the existence and monicity of the cd - index constitute the complete set of affine relations on the flag f - vector of an eulerian poset . setting the degree of @xmath235 to be 1 and the degree of @xmath11 to be 2, the cd - index of a poset of rank @xmath210 is homogeneous of degree @xmath152 .the number of cd - monomials of degree @xmath153 is @xmath237 , the @xmath238 fibonacci number , with @xmath239 .thus the affine span of flag f - vectors of eulerian posets of degree @xmath152 has dimension @xmath240 .the literature is divided on notation for the cd - index , due to two valid points of view as to what the ab - index is . if one considers @xmath229 to be a polynomial function of non - commuting variables @xmath120 and @xmath121 , one may consider the cd - index to be a different polynomial function in @xmath235 and @xmath11 , and give it a different name , typically @xmath4 . on the other hand ,if @xmath229 is a vector in a space of ab - polynomials , the cd - index is the same vector , which happens to be written as a linear combination of monomials in @xmath235 and @xmath11 .thus one would call the cd - index @xmath229 .we will primarily use the notation @xmath229 , except that when we talk about inequalities on the coefficients of the cd - index , we use @xmath4 . aside from the existence and monicity of the cd - index , there are no additional affine relations on flag f - vectors of polytopes .bayer and billera @xcite and later kalai @xcite gave a basis of polytopes whose flag f - vectors span @xmath241 .much is also known about bounds on the coefficients of the cd - index of a polytope .a bound on the cd - index implies bounds on @xmath242 and @xmath243 , because @xmath242 and @xmath243 can be written as positive combinations of coefficients of the cd - index .the first consideration is the non - negativity of the coefficients .stanley @xcite conjectured that the coefficients of the cd - index are non - negative whenever @xmath3 triangulates a homology sphere ( or in other words when @xmath3 is a _gorenstein * poset _ ) .he also showed that the coefficients of @xmath4 are non - negative for a class of cw - spheres which includes convex polytopes .ehrenborg and readdy described how the cd - index is changed by the poset operations of pyramid and vertex shaving .the following is a combination of propositions 4.2 and 6.1 of @xcite .[ cd ops ] let @xmath3 be a graded poset and let @xmath120 be an atom . then @xmath244}\cdot d \cdot \psi_{[x,\hat{1 } ] } \right)\\ \psi_{{{{\mathrm{sh}}}}_a(p ) } & = & \psi_p + \frac{1}{2}\left(\psi_p\cdot c - c\cdot \psi_p + \sum_{a < x<{{\hat{1}}}}\psi_{[a , x]}\cdot d\cdot\psi_{[x,{{\hat{1 } } } ] } \right).\end{aligned}\ ] ] ehrenborg and readdy also defined a derivation on cd - indices and used it to restate the formulas in proposition [ cd ops ] .the derivation @xmath172 ( called @xmath245 in @xcite ) is defined by @xmath246 and @xmath247 .the following is a combination of theorem 5.2 and proposition 6.1 of @xcite .[ cd derivation ] let @xmath3 be a graded poset and let @xmath120 be an atom. then @xmath248}\right).\end{aligned}\ ] ] [ pyr ] let @xmath3 be a homogeneous cd - polynomial whose lexicographically first term is @xmath249 . then the lexicographically first term of @xmath34 is @xmath250 .in particular , the kernel of the pyramid operation is the zero polynomial .in this section we introduce the zipping operation and prove some of its important properties .in particular , zipping will be part of a new inductive proof that bruhat intervals are spheres and thus eulerian .zipper _ in a poset @xmath3 is a triple of distinct elements @xmath251 with the following properties : 1 .@xmath252 covers @xmath0 and @xmath1 but covers no other element .2 . @xmath253 .3 . @xmath254 .call the zipper _ proper _ if @xmath252 is not a maximal element .if @xmath255 is a zipper in @xmath3 and @xmath256 $ ] is an interval in @xmath3 with @xmath257 $ ] then @xmath255 is a zipper in @xmath256 $ ] . given @xmath3 and a zipper @xmath255 one can `` zip '' the zipper as follows : let @xmath258 stand for a single new element not in @xmath3 .define @xmath259 , with a binary relation called @xmath260 , given by : @xmath261 for convenience , @xmath256 $ ] will always mean the interval @xmath256_\le$ ] in @xmath3 and @xmath256_\preceq$ ] will mean an interval in @xmath262 . in each of the following propositions ,@xmath262 is obtained from @xmath3 by zipping the proper zipper @xmath255 , although some of the results are true even when the zipper in not proper .[ poset ] @xmath262 is a poset under the partial order @xmath260 .one sees immediately that @xmath260 is reflexive and that antisymmetry holds in @xmath263 .if @xmath264 and @xmath265 , but @xmath266 , then @xmath267 .we have @xmath268 and @xmath269 . also , either @xmath270 or @xmath271 .by antisymmetry in @xmath3 , either @xmath272 or @xmath273 .this contradiction shows that @xmath274 .transitivity follows immediately from the transitivity of @xmath3 except perhaps when @xmath275 and @xmath276 . in this case ,@xmath268 and @xmath269 .also , either @xmath277 or @xmath278 . in either case , @xmath279 and therefore @xmath280 .[ mu1 ] if @xmath275 then @xmath281 . if @xmath282 with @xmath266 , then @xmath283 .suppose @xmath256_\preceq$ ] is any non - trivial interval in @xmath262 . if @xmath284 , then @xmath256_\preceq=[a , b]_\le$ ] . if @xmath285 , then @xmath256_\preceq\cong[a , x]_\le$ ] .if @xmath286 and @xmath287 , then @xmath256_\le$ ] does not contain both @xmath0 and @xmath1 , and we obtain @xmath256_\preceq$ ] from @xmath256_\le$ ] by replacing @xmath0 or @xmath1 by @xmath258 if necessary .thus in the proofs that follow , one needs only to check two cases : the case where @xmath288 and @xmath289 and the case where @xmath274 .let @xmath280 with @xmath266 .one needs only to check the case where @xmath288 and @xmath289 .this is done by induction on the length of the longest chain from @xmath252 to @xmath121 . if @xmath290 then @xmath291 here the second line is obtained by properties ( i ) and ( iii ) .if @xmath121 does not cover @xmath252 , use the same calculation , employing induction to go from the second line to the third line .[ mu2 ] if @xmath292 , then @xmath293 . in light of proposition [ mu1 ] , one can write : @xmath294 the following two corollaries follow trivially from propositions [ mu1 ] and [ mu2 ] and the observation that if @xmath3 is ranked , then @xmath262 inherits a rank function .[ thin ] if @xmath3 is thin , then so is @xmath262 [ eul ] if @xmath3 is eulerian , then so is @xmath262 .[ cd thm ] @xmath3 has a cd - index if and only if @xmath262 has cd - index .the cd - indices are related by : @xmath295_\le}\cdot d \cdot\psi_{[z,{{\hat{1}}}]_\le}.\ ] ] we subtract from @xmath228 the chains which disappear under the zipping .first subtract the terms which came from chains through @xmath0 and @xmath252 .any such chain is a chain in @xmath190_p$ ] concatenated with a chain in @xmath296_p$ ] .so the terms subtracted off are @xmath297_p } \cdot b \cdot b \cdot \upsilon_{[z,\hat{1}]}$ ] .then subtract a similar term for chains through @xmath1 and @xmath252 .in fact , by condition ( iii ) of the definition of a zipper , the term for chains through @xmath1 and @xmath252 is identical to the term for chains through @xmath0 and @xmath252 .subtract @xmath297_p } \cdot a \cdot b \cdot \upsilon_{[z,\hat{1}]}$ ] for the chains which go through @xmath252 but skip the rank below @xmath252 .finally , @xmath0 is identified with @xmath1 , so there is a double - count which must be subtracted off . if two chains are identical except that one goes through @xmath0 and the other goes through @xmath1 , then they are counted twice in @xmath3 but only once in @xmath262 . because @xmath298 , if such a pair of chains include an element whose rank is @xmath299 , then that element is @xmath252 .but the chains through @xmath252 have already been subtracted , so we need to subtract off @xmath297_p } b\cdot a \cdot \upsilon_{[z,\hat{1}]}$ ] .we have again used condition ( iii ) here .thus : @xmath300_p}(2bb+ab+ba)\cdot\upsilon_{[z,\hat{1}]_p}.\end{aligned}\ ] ] replacing @xmath120 by @xmath301 one obtains : @xmath302_p}\cdot ( ab + ba)\cdot \psi_{[z,\hat{1}]_p}\\ & = & \psi_p - \psi_{[\hat{0},x]_p}\cdot d\cdot\psi_{[z,\hat{1}]_p}. \nonumber\end{aligned}\ ] ] [ pl ] if @xmath3 is a pl sphere , then so is @xmath262 . to avoid tedious repetition, we will omit `` pl '' throughout the proof .all spheres and balls are assumed to be pl .suppose @xmath3 is a @xmath7-sphere .let @xmath303 be the simplicial complex whose facets are maximal chains in @xmath213 passing through @xmath0 , @xmath1 or @xmath252 .our first goal is to prove that @xmath304 is a ball .let @xmath305 be the simplicial complex whose facets are maximal chains in @xmath306 through @xmath0 .similarly @xmath307 .one can think of @xmath308 as @xmath309 .thus , by proposition [ joins ] , @xmath308 is a @xmath7-ball , and similarly , @xmath307 .let @xmath310 .then @xmath165 is the complex whose facets are almost - maximal chains that can be completed to maximal chains either by adding @xmath0 or @xmath1 .these are the chains through @xmath252 which have elements at every rank except at the rank of @xmath0 .thus @xmath165 is @xmath311 , a @xmath192-ball , and @xmath165 lies in the boundary of @xmath308 , because there is exactly one way to complete a facet of @xmath165 to a facet of @xmath308 , namely by adjoining @xmath0 .similarly , @xmath165 lies in the boundary of @xmath307 .so by theorem [ pl tricks](i ) , @xmath312 is a @xmath7-ball .consider @xmath313 , which is the closure of @xmath314 . by theorem [ pl tricks](iii ) , @xmath313 is also a @xmath7-ball . also consider @xmath315 , which is isomorphic to @xmath313 .the boundary of @xmath315 is a complex whose facets are chains @xmath235 with the property that for each @xmath235 there is a unique element of @xmath316 that completes @xmath235 to a maximal chain .however , since @xmath317 is thin by corollary [ thin ] , it has the property that any chain of length @xmath318 can be completed to a maximal chain in @xmath317 in exactly two ways .therefore every facet of the boundary of @xmath315 is contained in a chain through @xmath258 .so @xmath319 is the union of a @xmath7-ball @xmath315 with the pyramid over the boundary of @xmath315 . by theorem [ pl tricks](ii ) , @xmath319 is a @xmath7-sphere . in the casewhere @xmath3 is thin , the conditions for a zipper can be simplified .[ simplezipper ] if @xmath3 is thin , then ( i ) implies ( iii ) .thus @xmath255 is a zipper if and only if it satisfies conditions ( i ) and ( ii ) .suppose condition ( i ) holds but @xmath320 .then without loss of generality @xmath0 covers some @xmath120 which @xmath1 does not cover . since @xmath252 covers no element besides @xmath0 and @xmath1 , @xmath321 $ ] is a chain of length 2 , contradicting thinness .in this section we state and prove the structural recursion for bruhat intervals .when @xmath92 , @xmath322 and @xmath323 , define a map @xmath324\times [ 1,s]\rightarrow [ u , ws]$ ] , as follows : @xmath325 to show that @xmath44 is well - defined , let @xmath326 $ ] . then @xmath327 $ ] because @xmath328 .either @xmath329 $ ] or @xmath330 . in the latter case , @xmath331 , so @xmath332 by the lifting property .[ eta ] if @xmath333 and @xmath334 , then @xmath324 \times [ 1,s ] \rightarrow [ u , ws]$ ] is an order - projection . to check that @xmath44 is order - preserving ,suppose @xmath335 in @xmath336 \times [ 1,s]$ ] .we have to break up into cases to check that @xmath337 . 1 . 1 .+ if @xmath339 as well , then @xmath340 . if @xmath341 , then @xmath342 is either @xmath343 with @xmath344 or it is @xmath345 with @xmath346 .2 . @xmath347 .+ so @xmath348 is either @xmath349 , with @xmath350 or it is @xmath351 with @xmath352 .we must also have @xmath341 , so @xmath342 is either @xmath343 with @xmath353 or it is @xmath345 with @xmath354 . if @xmath355 then @xmath356 . if @xmath357 and @xmath358 , we have @xmath359 and @xmath353 . by hypothesis ,@xmath360 , so by the lifting property @xmath361 . if @xmath357 and @xmath362 then @xmath359 and @xmath363 , so by the lifting property @xmath364. it will be useful to identify the inverse image of an element @xmath365 $ ] .the inverse image is : @xmath366 provided that these elements are actually in @xmath336 \times [ 1,s]$ ] . in the case where @xmath367 , we have @xmath368 , where the second inequality is by the lifting property .so @xmath369 is indeed an element of @xmath336 \times [ 1,s]$ ] . in the casewhere @xmath370 , we have by hypothesis @xmath371 , so by lifting , @xmath372 . also by lifting , since @xmath373 and @xmath374 , we have @xmath375. so @xmath376 \times [ 1,s]$ ] . now, suppose @xmath377 $ ] .to finish the proof that @xmath44 is an order - projection , we must find elements @xmath378\times[1,s]$ ] with @xmath379 and @xmath380 .consider 4 cases : 1 .1 . @xmath381 and @xmath382 .+ by the inverse - image argument of the previous paragraph , @xmath383 $ ] , so @xmath384 , @xmath385 and @xmath386 .2 . @xmath381 and @xmath387 .+ by the previous paragraph , @xmath388 $ ] .again we have @xmath384 , and @xmath389 . by lifting , @xmath390 , so @xmath391 .3 . @xmath392 and @xmath387 .+ we have @xmath393 $ ] , @xmath394 and @xmath389 . by lifting , @xmath395 , so @xmath396 .4 . @xmath392 and @xmath382 .+ we have @xmath397 $ ] . since @xmath398 ,@xmath399 $ ] as well .so @xmath384 , @xmath385 and @xmath386 .in light of the previous section , @xmath44 induces an isomorphism @xmath59 between @xmath400 $ ] and a poset derived from @xmath336 \times [ 1,s]$ ] , as follows : for every @xmath326 $ ] with @xmath370 , `` identify '' @xmath369 , @xmath401 and @xmath402 to make a single element . since @xmath402 covers only @xmath369 and @xmath401 we can also think of @xmath44 as deleting @xmath402 and identifying @xmath369 with @xmath401 .the map @xmath44 induces a map ( also called @xmath44 ) on the cw - spheres associated to bruhat intervals , as illustrated in figure [ fig4 ] .[ usnotininterval ] let @xmath333 , @xmath334 and @xmath403 .then @xmath331 for all @xmath404 $ ] , and @xmath44 is an isomorphism .suppose for the sake of contradiction that there is a @xmath405 $ ] with @xmath406 .since @xmath129 and @xmath407 , by lifting , @xmath408 . by transitivity , @xmath409 .this contradiction shows that that @xmath331 for all @xmath404 $ ] .now , looking back at the proof of proposition [ eta ] , we see that @xmath410 because in the @xmath411 case , the other two possible elements of @xmath412 do nt exist . thus the map @xmath413 is an order - isomorphism and therefore @xmath54 is also an order - isomorphism .the following corollary is easy .[ zeta ] if @xmath333 , @xmath334 and @xmath403 the map @xmath414\rightarrow [ us , ws]$ ] with @xmath415 is an isomorphism .we would also like to relate the interval @xmath416 $ ] to @xmath336 $ ] in the case where @xmath417 . to do this , we need an operator on posets corresponding to vertex - shaving on polytopes or cw spheres .let @xmath3 be a poset with @xmath188 and @xmath211 , and let @xmath120 be an atom of @xmath3 .the _ shaving _ of @xmath3 at @xmath120 is an induced subposet of @xmath418 $ ] given by : @xmath419 \times { { \lbrace \hat{0 } \rbrace } } \right ) \cup { { \lbrace ( \hat{0},\hat{0 } ) \rbrace}}.\ ] ] we can also describe @xmath151 as follows : let @xmath262 be obtained from @xmath420 $ ] by zipping the zipper @xmath421. denote by @xmath120 the element created by the zipping .then @xmath151 is the interval @xmath422 $ ] in @xmath262 .figures [ fig5 ] and [ fig6 ] illustrate the operation of shaving .let @xmath92 , @xmath322 , @xmath323 and @xmath417 .define a map @xmath423\rightarrow [ us , ws]$ ] as follows .starting with @xmath336 \times [ 1,s]$ ] , zip @xmath424 , call the new element @xmath425 , and identify @xmath426 $ ] with the interval @xmath427 $ ] in the zipped poset .now define : @xmath428\\ \theta(v , s ) & = & \left\{\begin{array}{ll } vs & \mbox{if } vs > v\\ v & \mbox{if } vs < v. \end{array } \right .\end{aligned}\ ] ] to check that @xmath429 is well - defined , begin by noting that @xmath430 $ ] , and if @xmath431 $ ] , then @xmath432 $ ] . if @xmath326 - { { \lbrace u , us \rbrace}}$ ] , there are two possibilities , @xmath433 or @xmath434 . in either case ,@xmath435 by lifting .so @xmath429 is well - defined .[ theta ] the map @xmath423\rightarrow [ us , ws]$ ] is an order - projection .notice that @xmath429 , restricted to @xmath436 - { { \lbrace us \rbrace}}$ ] is just @xmath44 restricted to an induced subposet .recall that in the proof of proposition [ eta ] , it was shown that for @xmath437 $ ] , if @xmath367 then @xmath438 and if @xmath374 then @xmath439 .the existence of these elements of @xmath412 was used to check that @xmath44 is an order - projection .the same argument accomplishes most of the present proof . for @xmath440, we are done , and it remains to check that for @xmath441 , there exist elements @xmath279 in @xmath436 $ ] with @xmath442 and @xmath443 .this is easily accomplished by setting @xmath444 $ ] and letting @xmath121 be an element of @xmath445 .if @xmath446 , then set @xmath447 $ ] .figures [ fig7 ] and [ fig8 ] illustrate the map @xmath429 and the corresponding map on cw spheres .since @xmath44 is an order - projection , @xmath400 $ ] is isomorphic to the fiber poset of @xmath336\times[1,s]$ ] with respect to @xmath44 .similarly , @xmath416 $ ] is isomorphic to the fiber poset of @xmath436 $ ] with respect to @xmath429 .we will show that one can pass to the fiber poset in both cases by a sequence of zippings .order the set @xmath448 linearly such that the elements of rank @xmath449 in @xmath336 $ ] precede the elements of rank @xmath450 for all @xmath449 .write this order as @xmath451 .define @xmath452\times[1,s]$ ] and inductively define @xmath453 to be the poset obtained by zipping @xmath454 in @xmath455 .we show inductively that this is indeed a proper zipping .first , notice that @xmath456 , @xmath457 and @xmath458 are indeed elements of @xmath455 .the element @xmath458 has not been deleted yet , and we have not identified @xmath458 with any element because it is at a rank higher than we have yet made identifications .the only elements ever deleted are of the form @xmath459 where @xmath460 , so @xmath456 and @xmath457 have not been deleted .the only identification one could make involving @xmath456 and @xmath457 is to identify them to each other , and that has not happened yet .we check the properties in the definition of a zipper : by corollary [ thin ] and induction , @xmath455 is thin , so by proposition [ simplezipper ] it is enough to check that properties ( i ) and ( ii ) hold . properties ( i ) and ( ii ) hold in @xmath461 and therefore in @xmath455 because we have made no identifications involving @xmath456 , @xmath457 and @xmath458 or higher - ranked elements .the zipper is proper because @xmath462 in @xmath461 and thus in @xmath455 . by definition ,@xmath436 $ ] is an interval in the poset @xmath463 defined above .specifically , @xmath463 was obtained from @xmath336\times [ 1,s]$ ] by zipping @xmath464 .let @xmath425 be the element of @xmath463 resulting from identifying @xmath465 with @xmath466 .then @xmath436 $ ] is isomorphic to the interval @xmath427 $ ] in @xmath463 .the remaining deletions and identifications in the map @xmath429 are really zippings in the @xmath453. therefore they are zippings in the @xmath453 restricted to @xmath427 $ ] .we have proven the following : [ etathetazippers ] let @xmath373 , @xmath129 and @xmath467 . if @xmath468 $ ] then @xmath400\cong[u , w]\times[1,s]$ ] and @xmath416\cong[u , w]$ ] .if @xmath469 $ ] , then @xmath400 $ ] can be obtained from @xmath336\times[1,s]$ ] by a sequence of zippings , and @xmath416 $ ] can be obtained from @xmath436 $ ] by a sequence of zippings .[ spheres ] bruhat intervals are pl spheres .one only needs to prove the corollary for lower intervals , because by proposition [ interval ] it will then hold for all intervals .intervals of rank 1 are empty spheres .it is easy to check that a lower interval under an element of rank 2 is a pl 0-sphere . given the interval @xmath470 $ ] with @xmath471 , there exists @xmath127 such that @xmath472 .then @xmath470 $ ] can be obtained from @xmath473\times [ 1,s]$ ] by a sequence of zippings . by induction @xmath473 $ ] is a pl sphere , and thus @xmath473\times [ 1,s]$ ] is as well . by repeated applications of theorem [ pl ] ,@xmath470 $ ] is a pl sphere .the following observation will be helpful in section [ recursion section ] , when theorem [ etathetazippers ] is combined with theorem [ cd thm ] .[ intervals ] for @xmath474 , @xmath475_{p_{i-1 } } \cong [ ( u,1),(v_i,1)]_{p_0}\ ] ] @xmath476_{p_{i-1 } } \cong [ ( v_i , s),(w , s)]_{p_0}\ ] ] the second statement is obvious because of the way the @xmath477 were ordered . for the first statement, there is the obvious order - preserving bijection between the two intervals .the only question is whether the right - side has any extra order relations .extra order relations will occur if for some @xmath8 with @xmath374 , there exists @xmath478 with @xmath479 but @xmath480 .this is ruled out by transitivity .theorems [ cd thm ] and [ etathetazippers ] yield theorem [ formula ] , a set of recursions for the cd - indices of bruhat intervals . in this sectionwe prove theorem [ formula ] , then apply it to determine the affine span of the cd - indices of bruhat intervals . for @xmath481 and @xmath127 ,define @xmath482 thus @xmath483 is 1 if @xmath484 is lengthened by @xmath485 on the right and @xmath486 if @xmath8 is shortened by @xmath485 on the right .[ formula ] let @xmath322 , @xmath323 and @xmath467 .if @xmath468 $ ] , then @xmath487}={{{\mathrm{pyr}}}}\psi_{[u , w]}$ ] , and @xmath488}=\psi_{[u , w]}$ ] . if @xmath469 $ ] , then @xmath489 } & = & { { { \mathrm{pyr}}}}\psi_{[u , w ] } - \sum_{v\in(u , w):vs < v } \psi_{[u , v]}\cdot d\cdot\psi_{[v , w ] } \\ & = & \frac{1}{2}\left(\psi_{[u , w]}\cdot c + c\cdot\psi_{[u , w ] } + \sum_{v\in ( u , w ) } \sigma_s(v)\psi_{[u , v]}\cdot d \cdot \psi_{[v , w ] } \right)\\ \psi_{[us , ws ] } & = & { { { \mathrm{sh}}}}_{us}\psi_{[u , w ] } - \sum_{v\in ( us , w):vs < v } \psi_{[us , v]}\cdot d\cdot\psi_{[v , w ] } \\ & = & \psi_{[u , w ] } + \frac{1}{2}\left(\psi_{[us , w]}\cdot c - c\cdot\psi_{[us , w ] } + \!\!\sum_{v\in ( us , w ) } \sigma_s(v ) \psi_{[us , v]}\cdot d \cdot \psi_{[v , w ] } \right)\end{aligned}\ ] ] the first line of each formula looks like an augmented coproduct @xcite on a bruhat interval , with an added sign . the second line of each formula is more efficient for computation , because the formulas in proposition [ cd derivation ] are more efficient than the forms quoted in proposition [ cd ops ] .the statement for @xmath468 $ ] follows immediately from propositions [ usnotininterval ] and [ zeta ] .define the @xmath453 as in section [ building ] .thus by theorem [ cd thm ] , @xmath490_{p_{i-1}}}\cdot d\cdot\psi_{[(v_i , s),(w , s)]_{p_{i-1}}}.\ ] ] since @xmath491 $ ] , sum from @xmath492 to @xmath493 to obtain @xmath494}=\psi_{p_0}-\sum_{j=1}^k \psi_{[(u,1),(v_j,1)]_{p_{j-1}}}\cdot d\cdot\psi_{[(v_j , s),(w , s)]_{p_{j-1}}}.\ ] ] by proposition [ intervals ] , @xmath495_{p_{j-1 } } \cong [ ( u,1),(v_j,1)]_{p_0}$ ] .this in turn is isomorphic to @xmath496 $ ] .similarly , by proposition [ intervals ] , @xmath497_{p_{j-1 } } \cong [ ( v_j , s),(w , s)]_{p_0}$ ] which is isomorphic to @xmath498 $ ] .thus we have established the first line of the first formula in theorem [ formula ] .the second line follows from the first by proposition [ cd ops ] .a similar proof goes through for the second formula .the isomorphisms from proposition [ intervals ] restrict to the appropriate isomorphisms for the @xmath416 $ ] case .the second line of the formula follows by proposition [ cd ops ] . in @xcite , bayer and billerashow that the affine span of the cd - indices of polytopes is the entire affine space of monic cd - polynomials . as an application of theorem [ formula ], we prove that the cd - indices of bruhat intervals have the same affine span .[ span ] the set of cd - indices of bruhat intervals spans the affine space of cd - polynomials .the space of cd - polynomials of degree @xmath153 has dimension @xmath237 , the fibonacci number , with @xmath499 and @xmath500 . for each @xmath152 we will produce a set @xmath501 consisting of @xmath237 reduced words , corresponding to group elements whoselower bruhat intervals have linearly independent cd - indices . let @xmath502 have a complete coxeter graph with each edge labeled 3 .each @xmath501 is a set of reduced words of length @xmath152 in @xmath88 , with @xmath503 , @xmath504 and @xmath505 where @xmath506 means disjoint union .given a word @xmath507 , by proposition [ usnotininterval ] , @xmath508\cong{{{\mathrm{pyr}}}}[1,w]$ ] , so @xmath509}={{{\mathrm{pyr}}}}(\psi_{[1,w]})$ ] .similarly , given a word @xmath510 , @xmath511\cong{{{\mathrm{pyr}}}}[1,w]$ ] .since @xmath512 does not commute with any other generator , and since @xmath512 is not a letter in @xmath99 , by proposition [ eta ] , @xmath513 $ ] is obtained from @xmath511 $ ] by a single zipping . in particular , by theorem [ formula ] , @xmath514}={{{\mathrm{pyr}}}}^2(\psi_{[1,w]})-d\cdot\psi_{[s_n , s_nw]}$ ] . bya `` left '' version of corollary [ zeta ] , we have @xmath515\cong[1,w]$ ] , so @xmath514}={{{\mathrm{pyr}}}}^2(\psi_{[1,w]})-d\cdot\psi_{[1,w]}$ ] .let @xmath516 be the set of cd - indices of lower intervals under words in @xmath501 . the proof is now completed via proposition [ indep ] , below , which in turn depends on lemma [ pyr ] .[ indep ] for each @xmath517 , the @xmath237 cd - polynomials in @xmath516 are linearly independent . as a base for induction , the statement is trivial for @xmath518 . for general @xmath152 , form the matrix @xmath519 whoserows are the vectors @xmath516 , written in the cd - index basis .order the columns by the lexicographic order on cd - monomials .order the rows so that the cd - indices in @xmath520 appear first .we will show that there are row operations which convert @xmath519 to an upper - unitriangular matrix .notice that for each @xmath521 , @xmath522}$ ] occurs in @xmath520 .also , @xmath522}-d\cdot\psi_{[1,w]}$ ] occurs in @xmath523 .thus by row operations one obtains a matrix @xmath524 whose rows are first @xmath525 , then @xmath526 . by induction , there are row operations which convert the matrix with rows @xmath527 to an upper - unitriangular matrix . by corollary [ pyr ] , these yield row operations which give the first @xmath528 rows of @xmath524 an upper - unitriangular form . also by induction , there are row operations which convert the matrix with rows @xmath529 to an upper - unitriangular matrix .corresponding operations applied to the rows @xmath526 of @xmath524 complete the reduction of @xmath524 to upper - unitriangular form .the proof of theorem [ span ] uses infinite coxeter groups. it would be interesting to know whether the cd - indices of bruhat intervals in finite coxeter groups also span , and whether a spanning set of intervals could be found in the finite coxeter groups of type a.in this section we discuss lower and upper bounds on the coefficients of the cd - index of a bruhat interval .the conjectured lower bound is a special case of a conjecture of stanley @xcite .[ nonneg ] for any @xmath530 in @xmath88 , the coefficients of @xmath10}$ ] are non - negative . the coefficient of @xmath236 is always 1 , and for the other coefficients the bound is sharp because the dihedral group @xmath531 has cd - index @xmath532 .computer studies have confirmed the conjecture in @xmath100 with @xmath533 .the conjectured upper bounds are attained on bruhat intervals which are isomorphic to the face lattices of convex polytopes . for convenience, we will say that such intervals `` are '' polytopes .we now use the results of section [ building ] to construct these intervals .a polytope is said to be _ dual stacked _ if it can be obtained from a simplex by a series of vertex - shavings .as the name would indicate , these polytopes are dual to the _ stacked polytopes _ @xcite which we will not define here .there are bruhat intervals which are dual stacked polytopes .let @xmath88 be a universal coxeter group with coxeter generators @xmath534 .define @xmath535 ( for `` cyclic word '' ) to be the word @xmath536 , where the subscript @xmath7 is understood to mean @xmath537 .so for example , if @xmath538 , then @xmath539 . in a universal coxeter group , every group element corresponds to a unique reduced word .thus we will use these words interchangeably with group elements .[ dual stacked ] the interval @xmath540 $ ] in @xmath88 is a dual stacked polytope of dimension @xmath11 with @xmath9 facets .the proof is by induction on @xmath7 . for @xmath541 ,the interval is @xmath542 $ ] , and because all subwords of @xmath543 are distinct group elements , this interval is a boolean algebra the face poset of a @xmath11 dimensional simplex . for @xmath544 , the interval @xmath540$ ] is obtained from the interval @xmath545 $ ] by shaving the vertex @xmath546 and then possibly by performing a sequence of zippings . by induction , @xmath545 $ ] is a dual stacked polytope of dimension @xmath11 with @xmath547 facets .the zippings correspond to elements of @xmath548 which are shortened on the right by @xmath549 .we will show that in fact there are no such elements .let @xmath550 , and let @xmath8 also stand for the unique reduced word for the element @xmath8 . since every element of @xmath88 has a unique reduced word , the fact that @xmath551 means that @xmath8 contains @xmath535 as a subword .but the only subword @xmath535 of @xmath552 is the first @xmath7 letters of @xmath552 .thus @xmath8 is a subword of @xmath552 consisting of the first @xmath7 letters and at least one other letter .now , @xmath553 , so @xmath8 ends in some generator other than @xmath554 , and therefore , @xmath8 is not shortened on the right by @xmath555 .thus there are no zippings following the shaving , and so @xmath540 $ ] is a dual stacked polytope of dimension @xmath11 with @xmath9 facets .proposition [ dual stacked ] shows that the following conjectured upper bound is sharp if indeed it holds .[ dual stacked upper ] the coefficientwise maximum of all cd - indices @xmath10}$ ] with @xmath556 and @xmath557 is attained on a bruhat interval which is isomorphic to a dual stacked polytope of dimension @xmath11 with @xmath9 facets .this conjecture is natural in light of proposition [ dual stacked ] and theorem [ formula ] .there are two issues which complicate the conjecture .first , any proof using theorem [ formula ] requires non - negativity ( conjecture [ nonneg ] ) .second , and perhaps even more serious , there is the issue of commutation of operators . given @xmath29denote the corresponding `` downstairs '' element in @xmath34 by @xmath558 and the `` upstairs '' element by @xmath559 .denote the operation of zipping a zipper @xmath255 by @xmath560 . then @xmath561 .the triple @xmath562 becomes a zipper only after @xmath560 is applied .pyramid and shaving also commute reasonably well : @xmath563 .however , zipping does not in general commute nicely with the operation of shaving off a vertex @xmath120 . given @xmath564denote the corresponding element of @xmath565 again by @xmath558 , and if in addition @xmath566 , write @xmath567 for the new element created by shaving . if @xmath568 and @xmath569 then @xmath570 . if @xmath571 then @xmath572 .however , if @xmath0 and @xmath1 are vertices then @xmath573 , where @xmath574 is the operation of shaving off the edge @xmath252 .since the pyramid operation commutes nicely with zipping , and any lower interval is obtained by pyramid and zipping operations , it is possible to obtain any lower interval by a series of pyramid operations followed by a series of zippings .thus by theorem [ formula ] : [ lower ] assuming conjecture [ nonneg ] , for any @xmath99 in an arbitrary coxeter group , @xmath575 } \le \phi_{b_{l(w)}}.\ ] ] here @xmath576 is the boolean algebra of rank @xmath152 .it is not true that the cd - index of general intervals is less than that of the boolean algebra of appropriate rank .for example , @xmath577 $ ] is the face lattice of a square , with @xmath578 } = c^2 + 2d$ ] .however , @xmath579 . equation ( [ ab formula ] ) in the proof of theorem [ cd thm ] is a formula for the change in the ab - index under zipping .thus theorem [ formula ] has a flag h - vector version , and since the flag h - vectors of bruhat intervals are known to be nonnegative , the following theorem holds .[ boolean h ] for any @xmath99 in an arbitrary coxeter group , @xmath580 } \le \psi_{b_{l(w)}}.\ ] ] here `` @xmath581 '' means is coefficientwise comparison of the ab - indices , or in other words , comparison of flag h - vectors .the author wishes to thank his adviser , vic reiner , for many helpful conversations , as well as lou billera , francesco brenti , kyle calderhead , richard ehrenborg , margaret readdy and michelle wachs for helpful comments .

<S> we study flag enumeration in intervals in the bruhat order on a coxeter group by means of a structural recursion on intervals in the bruhat order . </S><S> the recursion gives the isomorphism type of a bruhat interval in terms of smaller intervals , using basic geometric operations which preserve pl sphericity and have a simple effect on the cd - index . </S><S> this leads to a new proof that bruhat intervals are pl spheres as well a recursive formula for the cd - index of a bruhat interval . </S><S> this recursive formula is used to prove that the cd - indices of bruhat intervals span the space of cd - polynomials . </S><S> the structural recursion leads to a conjecture that bruhat spheres are `` smaller '' than polytopes . </S><S> more precisely , we conjecture that if one fixes the lengths of @xmath0 and @xmath1 , then the cd - index of a certain dual stacked polytope is a coefficientwise upper bound on the cd - indices of bruhat intervals @xmath2 $ ] . </S><S> we show that this upper bound would be tight by constructing bruhat intervals which are the face lattices of these dual stacked polytopes . as a weakening of a special case of the conjecture , </S><S> we show that the flag h - vectors of lower bruhat intervals are bounded above by the flag h - vectors of boolean algebras ( i. e. simplices ) . </S><S> a graded poset is _ eulerian _ if in every non - trivial interval , the number of elements of odd rank equals the number of elements of even rank . </S><S> face lattices of convex polytopes are in particular eulerian and the study of flag enumeration in eulerian posets has its origins in the face - enumeration problem for polytopes . </S><S> all flag - enumerative information in an eulerian poset @xmath3 can be encapsulated in a non - commutative generating function @xmath4 called the _ cd - index_. the cd - indices of polytopes have received much attention , for example in @xcite . </S><S> a _ coxeter group _ is a group generated by involutions , subject to certain relations . </S><S> important examples include finite reflection groups and weyl groups . </S><S> the _ bruhat order _ on a coxeter group is a partial order which has important connections to the combinatorics and representation theory of coxeter groups , and by extension lie algebras and groups . </S><S> intervals in bruhat order comprise the other important class of eulerian posets . </S><S> however , flag enumeration for bruhat intervals has previously received little attention . </S><S> the goal of the present work is to initiate the study of the cd - index of bruhat intervals . </S><S> the basic tool in our study is a fundamental structural recursion ( theorem [ etathetazippers ] ) on intervals in the bruhat order on coxeter groups . </S><S> this recursion , although developed independently , has some resemblance to work by du cloux @xcite and by dyer @xcite . </S><S> the recursion gives the isomorphism type of a bruhat interval in terms of smaller intervals , using some basic geometric operations , namely the operations of pyramid , vertex shaving and a `` zipping '' operation . </S><S> the result is a new inductive proof of the fact @xcite that bruhat intervals are pl spheres ( corollary [ spheres ] ) as well as recursions for the cd - index of bruhat intervals ( theorem [ formula ] ) . </S><S> the recursive formulas lead to a proof that the cd - indices of bruhat intervals span the space of cd - polynomials ( theorem [ span ] ) , and motivate a conjecture on the upper bound for the cd - indices of bruhat intervals ( conjecture [ dual stacked upper ] ) . </S><S> let @xmath5 $ ] be an interval in the bruhat order such that the rank of @xmath6 is @xmath7 and the rank of @xmath8 is @xmath9 . </S><S> we conjecture that the coefficients of @xmath10}$ ] are bounded above by the coefficients of the cd - index of a dual stacked polytope of dimension @xmath11 with @xmath9 facets . </S><S> the dual stacked polytopes are the polar duals of the stacked polytopes of @xcite . </S><S> this upper bound would be sharp because the structural recursion can be used to construct bruhat intervals which are the face lattices of duals of stacked polytopes ( proposition [ dual stacked ] ) . </S><S> stanley @xcite conjectured the non - negativity of the cd - indices of a much more general class of eulerian posets . </S><S> we show ( theorem [ lower ] ) that if the conjectured non - negativity holds for bruhat intervals , then the cd - index of any lower bruhat interval is bounded above by the cd - index of a boolean algebra . since the flag h - vectors of bruhat intervals are non - negative , we are able to prove that the flag h - vectors of lower bruhat intervals are bounded above by the flag h - vectors of boolean algebras ( theorem [ boolean h ] ) . the remainder of the paper is organized as follows : we begin with background information on the basic objects appearing in this paper , namely , posets , coxeter groups , bruhat order and polytopes in section [ pre ] , cw complexes and pl topology in section [ cw ] and the cd - index in section [ cd ] . in section </S><S> [ zipping ] , the zipping operation is introduced , and its basic properties are proven . </S><S> section [ building ] contains the proof of the structural recursion . in section [ recursion section ] </S><S> we state and prove the cd - index recursions and apply them to determine the affine span of cd - indices of bruhat intervals . </S><S> section [ bounds section ] is a discussion of conjectured bounds on the coefficients of the cd - index of a bruhat interval , including the construction of bruhat intervals which are isomorphic to the face lattices of dual stacked polytopes . </S>

1205.4372

the mechanical action of light upon neutral atoms placed in a laser standing wave is at the heart of laser cooling , trapping , and bose - einstein condensation @xcite .numerous applications of the mechanical action of light include isotope separation , atomic interferometry , atomic lithography and epitaxy , atomic - beam deflection and splitting , manipulating translational and internal atomic states , measurement of atomic positions , etc .atoms and ions in an optical lattice , formed by a laser standing wave , are perspective objects for implementation of quantum information processing and quantum computing .advances in cooling and trapping of atoms , tailoring optical potentials of a desired form and dimension , controlling the level of dissipation and noise are now enabling the direct experiments with single atoms to study fundamental principles of quantum physics , quantum chaos , decoherence , and quantum - classical correspondence .nonlinear dynamics of cold atoms in optical lattices is a fastly growing branch of atomic physics .there are a number of theoretical works and impressive experiments on quantum chaos , dynamical localization , chaos - assisted tunneling , lvy flights , etc .( for reviews see @xcite ) . to suppress spontaneous emission and provide a coherent quantum dynamics one usually works far from the optical resonance .adiabatic elimination of the excited state amplitude leads to an effective hamiltonian for the center - of - mass motion , whose 3/2 degree - of - freedom classical analogue has a mixed phase space with regular islands embedded in a chaotic sea .new possibilities are opened if one works near the optical resonance and take the internal atomic dynamics into account .a single atom in a standing - wave laser field may be semiclassically treated as a nonlinear dynamical system with coupled internal ( electronic ) and external ( mechanical ) degrees of freedom @xcite . in the semiclassical and hamiltonian limits ( when one treats atoms as point - like particles and neglects spontaneous emission and other losses of energy ) , a number of nonlinear dynamical effects have been analytically and numerically demonstrated with this system : chaotic rabi oscillations @xcite , hamiltonian chaotic atomic transport and dynamical fractals @xcite , lvy flights and anomalous diffusion @xcite .these effects are caused by local instability of the cm motion in a laser field .a set of atomic trajectories under certain conditions becomes exponentially sensitive to small variations in initial quantum internal and classical external states or / and in the control parameters , mainly , the atom - laser detuning .hamiltonian evolution is a smooth process that is well described in a semiclassical approximation by the coupled hamilton - schrdinger equations . a detailed theory of hamiltonian and dissipative chaotic transport of atoms in a laser standing wave has been developed in refs .@xcite and @xcite , respectively .additional possibilities to manipulate the atomic transport are created by applying an external force to the standing - wave optical potential .it is obvious that for cold atoms in a vertical optical lattice it is necessary to account for the earth s acceleration .it is possible as well to create horizontal accelerated optical lattices by adding a constant force whose magnitude along the optical axis can be easily varied .the problem of atomic motion in a tilted optical potential is closely related to the old problem of electron motion in a a periodic crystal with dc or ac forces applied .the analogue of well known bloch oscillations with cold atoms has been experimentally found in ref.@xcite . in the present paperwe apply the ideas and methods , elaborated in the field of nonlinear dynamics of cold atoms , to study theoretically and numerically motion of point - like atoms in a tilted optical lattice .it will be shown that varying only one parameter , the detuning between the frequencies of a working atomic transition and the laser field , one can explore a variety of regimes of atom motion , including chaotic walking , dynamical fractals and chaotic scattering .in the one - dimensional case , the hamiltonian of a two - level atom in a standing - wave laser field and an additional external field can be written in the frame rotating with the laser frequency @xmath1 as follows : @xmath2 where @xmath3 are the pauli operators for the internal atomic degrees of freedom , @xmath4 and @xmath5 are the classical atomic position and momentum , @xmath6 and @xmath7 are the atomic transition and maximal rabi frequencies , respectively .@xmath8 stands for the static force induced by external field . in the semiclassical approximation , where the transversal atomic momentum @xmath9 is supposed to be , in average , much larger than the photonone @xmath10 , atom with quantized internal dynamics is treated as a point - like particle to be described by the bloch hamilton equations of motion without relaxation terms @xmath11 where @xmath12 and @xmath13 are synchronized ( with the laser field ) and quadrature components of the atomic electric dipole moment , respectively , and @xmath14 is the atomic population inversion . equations ( [ hamsys ] ) are written in the dimensionless form with @xmath15 and @xmath16 to be classical atomic center - of - mass position and momentum , respectively .dot denotes differentiation with respect to the dimensionless time @xmath17 .the set ( [ hamsys ] ) has the three control parameters @xmath18 which are the normalized recoil frequency , @xmath19 , atom - field detuning , @xmath20 , and applied force @xmath21 .the system has two integrals of motion , namely the total energy @xmath22 and the length of the bloch vector @xmath23 .the external force is directed in the negative direction of the optical axis @xmath24 .so , if the initial atomic momentum , @xmath25 , is chosen to be in the negative direction , the force will simply accelerate the corresponding atoms .if @xmath26 , then one may expect much more complicated motion .equations ( [ hamsys ] ) constitute a nonlinear hamiltonian autonomous system with two and half degrees of freedom .owing to two integrals of motion , phase points move on a three - dimensional hypersurface with a given energy value @xmath27 . in general , motion in a three - dimensional phase space in characterized by a positive lyapunov exponent , a negative exponent equal in magnitude to the positive one , and zero exponent .the maximal lyapunov exponent characterizes the mean rate of the exponential divergence of initially close trajectories and serves as a quantitative measure of dynamical chaos in the system . because of a transient character of chaos , we have computed the finite - time lyapunov exponent @xmath28 by the method developed in refs .it has been found that at the fixed value , @xmath29 , of the recoil frequency @xmath28 is positive in the following ranges of values of the other control parameters : the detuning @xmath30 and the force @xmath31 .therefore , we expect chaotic motion of atoms with the parameter s values in those ranges . in numerical experiments throughout the paperwe suppose that two - level atoms are initially prepared in the ground states , @xmath32 , with @xmath33 and fixed values of the two control parameters , the normalized recoil frequency , @xmath29 , and the external force @xmath34 .the atom - field detuning , @xmath20 , can be changed in a wide range .it will be shown in this paper that atoms may perform chaotic walking , the new type of motion in absolutely deterministic environment where atoms can change the direction of motion alternating between flying through the standing wave and being trapped in its potential wells .we would like to stress that the local instability produces chaotic center - of - mass motion in a rigid optical lattice without any modulation of its parameters . in fig .[ fig1 ] we illustrate different regimes of the center - of - mass atomic motion along the optical axis with the initial atomic momentum chosen to be @xmath35 .it is simply the motion on the phase plane @xmath36 . a typical picture with chaotic walkingis shown in fig .[ fig1]a with the value of the detuning , @xmath37 , at which the maximal lyapunov exponent , @xmath28 , is positive .the atom starts to move in the positive @xmath24-direction , changes soon the direction of motion a few times , acquiring irregularly positive and negative values of the momentum , and suddenly begins to move in the positive @xmath24-direction for a comparatively long time .then it is decelerated , turns back and flies in the negative direction .after that it changes the direction of motion many times demonstrating what we call `` chaotic walking '' .for comparison , we show in fig .[ fig1]b the phase - plane motion with the larger detuning @xmath38 ( and at the same other conditions ) at which the maximal lyapunov exponent is not positive .the atom moves initially in the positive @xmath24-direction , accelerating and decelerating alternatively .soon it changes the direction of motion and moves permanently in the negative direction .the motion is regular with a slight modulation of the momentum .the motion at exact resonance , @xmath39 , is even more simple ( fig .[ fig1]c ) .what is the ultimate reason of chaotic walking ? for an optical lattice without an external force , it has been found in ref .@xcite that instability is caused by the specific behavior of the bloch - vector component of a moving atom , @xmath12 , whose shallow oscillations between the standing - wave nodes are interrupted by sudden jumps with different amplitudes while atom crosses each node of the wave .it looks like a random like shots happened in a fully deterministic environment .the reason of chaotic walking in a tilted potential is the same .it follows from the second equation in the set ( [ hamsys ] ) that those jumps result in sudden changes of the atomic momentum while crossing nodes .if the value of the atomic energy is close to the separatrix one , the atom after the corresponding jump - like change in @xmath9 can either overcome the potential barrier and leave a potential well or it will be trapped by the well , or it will move as before crossing the node . the evolution of all the bloch components in the regime of chaotic walking is shown in fig . [ fig2 ] .for comparison , we show in figs .[ fig3 ] and [ fig4 ] their evolution in the regular regimes , far off the resonance and at exact resonance , respectively .different types of fractal - like structures may arise in chaotic hamiltonian systems ( see reviews @xcite ) .it is known from many studies in celestial mechanics @xcite , fluid dynamics @xcite , atomic physics @xcite , cavity quantum electrodynamics @xcite , underwater acoustics @xcite and other disciplines @xcite that under certain conditions the motion inside an interaction region may have features that are typical for dynamical chaos , ( homoclinic and heteroclinic tangles , fractals , strange invariant sets , positive finite - time lyapunov exponents , etc . )although the particle s trajectories are not chaotic in a rigorous sense because chaos is defined as an irregular motion over infinite time .let us place atoms one by one at the point @xmath33 with the same value of the initial momentum @xmath40 but change slightly the value of the detuning @xmath20 .all the other initial conditions and the control parameters are supposed to be the same for all the atoms .we fix the time moment @xmath41 when each atom crosses the point @xmath42 moving in the negative direction .the exit time function @xmath43 in fig .[ fig5 ] demonstrates the complicated structure with smooth intervals alternating with wildly oscillating peaks that can not be resolved in principle , no matter how large the magnification factor . the panels ( b ) and ( c ) in fig . [ fig5 ] are successive 50 times magnifications of the detuning intervals shown in the panel ( a ) .further magnifications reveal a self - similar fractal - like structure that is typical for hamiltonian systems with chaotic scattering .the exit time @xmath41 increases with increasing the magnification factor .the same picture is observed when computing the exit time function in the position and momentum spaces .it is a clear demonstration of a fractal - like behavior of chaotically walking atoms .it is established in theory of one and half degree - of - freedom systems that transient hamiltonian chaos in the interaction region occurs due to existence of , at least , one non - attractive chaotic invariant set consisting of an infinite number of localized unstable periodic orbits and aperiodic orbits .this set possesses stable and unstable manifolds extending into the regions of regular motion .the particles with the initial positions close to the stable manifold follow the chaotic - set trajectories for a comparatively long time , then deviate from them , and leave the interaction region along the unstable manifold . in a typical hamiltonian systemthere exists an infinite number of trajectories of zero measure with infinite exit time which belong to that chaotic invariant set .our system with two and half degrees of freedom is a much more complicated one , and it is practically impossible to reveal the corresponding chaotic invariant set with its stable and unstable manifolds .however , the mechanism of chaotic scattering and fractal - like structures should be the same .the statistics of exit times @xmath41 is shown in fig .[ fig6 ] in a semilogarithmic and logarithmic scales .the probability distribution function ( pdf ) in this figure gives the probability for an atom to have a given value of @xmath41 .the bold straight line in fig .[ fig6]a implies that the pdf is exponential in its middle part , @xmath44 , with the exponent @xmath45 .however , the tail of the pdf is not exponential . to prove that we plot the function in the logarithmic scale in fig .[ fig6]b and compute the slope at the tail .it has no sense to calculate the slope at the very tail because of a small number of events with very large values of @xmath41 .the bold straight line implies that the pdf is a power - law one , @xmath46 , with the coefficient @xmath47 .it is interesting that the slope at the pdf tail around the value @xmath48 is rather typical for many chaotic hamiltonian systems @xcite .the reason of that is unclear .in hyperbolic chaotic systems the pdfs should decay exponentially because the phase space of such systems is homogeneous , and all the trajectories are unstable .it is not the case even with one and half degree - of - freedom systems with inhomogeneous phase space , where exist so - called stability islands embedded in a stochastic sea , because the borders of those islands are `` sticky '' .it means that a typical chaotic trajectory , wandering in the stochastic sea , approaches the island s borders and `` stick '' to them for a long time . by that reason ,the corresponding pdfs are not exponential but power - law ones at their tails .pdfs with power - law decay imply that the corresponding quantity , the exit time in our case , is scale invariant i.e. , there is no a single dominant scale in the process .geometrically it means that chaotic trajectories for such a process are self - similar .it is shown that point - like atoms in a tilted optical potential with a constant external force applied can move chaotically changing the direction of motion in a random - like way .the existence of chaos is confirmed by direct computation of the maximal finite - time lyapunov exponent of the equations of motion that is shown to be positive in a range of the atom - laser detuning and the applied - force strength .the ultimate reason of chaotic walking is the specific behavior of the bloch - vector component of a moving atom , @xmath12 , whose shallow oscillations between the standing - wave nodes are interrupted by sudden jumps with different amplitudes while atom crosses each node of the wave .it is demonstrated numerically that such a behavior arises exactly at those values of the detuning for which the lyapunov exponent is positive and atoms move chaotically .we illustrate different regimes of the center - of - mass motion simply varying the detuning .it is an easy way to manipulate the atomic transport in tilted optical lattices . treating motion of atoms in a tilted optical lattice as a scattering problem, we show that the scattering of atoms under conditions of chaotic walking is chaotic and typical for hamiltonian systems . fixing the time moment @xmath41 when atoms with slightly different values of the detuning , momentum or initial position cross a fixed point ( @xmath42 ), we show that the corresponding scattering functions demonstrate the complicated structure that can not be resolved in principle , no matter how large the magnification factor . owing to that the probability to have a given value of @xmath41 is not exponential but decays at its tail by a power law .this work was supported by the integration grant from the far - eastern and siberian branches of the russian academy of sciences ( 12-ii-0 - 02 - 001 ) , and by the program `` fundamental problems of nonlinear dynamics in mathematics and physics '' . .( a ) chaotic walking at @xmath37 , @xmath49 , @xmath29 .( b ) regular motion at @xmath38 and with the same other conditions . ( c )regular motion at the resonance , @xmath39 , with the same other conditions.,title="fig:",scaledwidth=30.0% ] .( a ) chaotic walking at @xmath37 , @xmath49 , @xmath29 .( b ) regular motion at @xmath38 and with the same other conditions .( c ) regular motion at the resonance , @xmath39 , with the same other conditions.,title="fig:",scaledwidth=30.0% ] .( a ) chaotic walking at @xmath37 , @xmath49 , @xmath29 .( b ) regular motion at @xmath38 and with the same other conditions . ( c )regular motion at the resonance , @xmath39 , with the same other conditions.,title="fig:",scaledwidth=30.0% ] but far from the resonance ( @xmath38).,title="fig:",scaledwidth=30.0% ] but far from the resonance ( @xmath38).,title="fig:",scaledwidth=30.0% ] but far from the resonance ( @xmath38).,title="fig:",scaledwidth=30.0% ] , with given initial position , @xmath33 and momentum @xmath35 , on the detuning .the successive magnifications are shown.,title="fig:",scaledwidth=30.0% ] , with given initial position , @xmath33 and momentum @xmath35 , on the detuning .the successive magnifications are shown.,title="fig:",scaledwidth=30.0% ] , with given initial position , @xmath33 and momentum @xmath35 , on the detuning .the successive magnifications are shown.,title="fig:",scaledwidth=30.0% ] in ( a ) semilogarithmic scale ( exponential decay in the middle part with the exponent @xmath45 ) and ( b ) logarithmic scale ( power - law decay at the tail with the coefficient @xmath47).,title="fig:",scaledwidth=50.0% ] in ( a ) semilogarithmic scale ( exponential decay in the middle part with the exponent @xmath45 ) and ( b ) logarithmic scale ( power - law decay at the tail with the coefficient @xmath47).,title="fig:",scaledwidth=50.0% ] 99 s. chu , _ rev ._ , * 73 * , 685 ( 1998 ) ; c. cohen - tannoudji , ibid , 707 ( 1998 ) ; w.d .phillips , ibid , 721 ( 1998 ) . m. g. raizen , _ adv . at . mol ._ , * 41 * , 43 ( 1999 ) .prants , _ hamiltonian chaos with a cold atom in an optical lattice . in book: hamiltonian chaos beyond the kam theory .( editors : a.c.j .luo and n. ibragimov ) _ ( springer verlag and beijing : higher education press , berlin , 2010 ) , 193 - 223. s. v. prants and l.e .konkov , _ jetp letters _ , * 73 * , 1801 ( 2001 ) [ _ pisma zhetf _ , * 73 * , 200 ( 2001 ) ] . s.v .prants and v.yu .sirotkin , _ phys .a _ , * 64 * , 033412 ( 2001 ) . s.v .prants , _ jetp letters _ , * 75 * , 651 ( 2002 ) [ _ pisma zhetf _ , * 75 * , 777 ( 2002 ) ] . v. yu . argonov and s.v. prants , _ jetp _ , * 96 * , 832 ( 2003 ) [ _ zhetf _ , * 123 * , 946 ( 2003 ) ] . s. v. prants and m. yu .uleysky , _ phys . lett .a _ , * 309 * , 357 - 362 ( 2003 ) .prants , m.yu .uleysky , and v.yu .argonov , _ phys . rev ., * 73 * , art . 023807( 2006 ) . v. yu .argonov and s. v. prants , _ physa _ , * 75 * , art .063428 ( 2007 ) . s. v. prants , m. edelman , and g. m. zaslavsky , _ phys .e _ , * 66 * , art .046222 ( 2002 ) .argonov and s. v. prants , _laser res ._ , * 27 * , 360 ( 2006 ) . v. yu .argonov and s. v. prants , _ physa _ , * 78 * , art . 043413 ( 2008 ) .v.yu . argonov and s.v .prants , _ europhys_ , * 81 * , art . 24003 ( 2008 ) . m. ben dahan , e. peik , j. reichel , y. castin , and c. salomon , _ phys ._ , * 76 * , 4508 ( 1996 ) .fischer , k.w .madision , q. niu , and m.g .* 58 * , r2648 ( 1998 ). l.e . konkov and s. v. prants , _ j. math_ , * 37 * , 1204 ( 1996 ) . l.e . konkov and s. v. prants , _ jetp letters _ , * 65 * , 833 ( 1997 ) [ _ pisma zhetf _ , * 65 * , 801 ( 1997 ) ] .p. gaspard , _ chaos , scattering and statistical mechanics _ , ( cambridge university press , cambridge , 1998 ) .j. aguirre , r.l .viana , and m.a.f .sanjuan , _ rev ._ , * 81 * , 333 ( 2009 ) . j.m . petit and m. henon , _ icarus _ , * 60 * , 536 ( 1986 ). m. budyansky , m. uleysky , and s. prants , _ physica d _ ,* 195 * , 369 ( 2004 ) .budyansky , m.yu .uleysky , and s.v .prants , _ jetp _ , * 99 * , 1018 ( 2004 ) [ _ zhetf _ , * 126 * , 1167 ( 2004 ) ] .makarov , m.yu .uleysky , and s.v .prants , _ chaos _ , * 14 * , n1 79 - 95 ( 2004 ) .m. glck , a.r kolovsky , and h.j .korsch , _ phys ._ , * 366 * , 103 ( 2002 ) .b. eckhardt , _ physica d _ , * 33 * , 89 ( 1988 ) .zaslavsky _ hamiltonian chaos and fractional dynamics _( oxford : university press , oxford , 2005 ) , 421 .prants , m.v .budyansky , m.yu .uleysky , and g.m .zaslavsky , _ chaos _ , * 16 * , art .033117 ( 2006 ) .

<S> chaotic walking of cold atoms in a tilted optical lattice , created by two counter propagating running waves with an additional external field , is demonstrated theoretically and numerically in the semiclassical and hamiltonian approximations . </S><S> the effect consists in random - like changing the direction of atomic motion in a rigid lattice under the influence of a constant force due to a specific behavior of the atomic dipole - moment component that changes abruptly in a random - like manner while atoms cross standing - wave nodes . </S><S> chaotic walking generates a fractal - like scattering of atoms that manifests itself in a self - similar structure of the scattering function in the atom - field detuning , position and momentum spaces . </S><S> the probability distribution function of the scattering time is shown to decay in a non - exponential way with a power - law tail . </S><S> [ sh ] * * [ cols="^ " , ] * s.v . </S><S> prants , v.o . </S><S> vitkovsky * _ laboratory of nonlinear dynamical systems , + pacific oceanological institute of the russian academy of sciences , + 690041 vladivostok , russia , url : dynalab.poi.dvo.ru_ @xmath0corresponding author e - mail : [email protected] + * keywords : * cold atom , tilted potential , chaos , fractal </S>

0708.0585

recent simulations suggest that the colors of massive galaxies in the local universe can only be explained if agn feedback quenches star formation in the host galaxy @xcite .the link between supermassive black hole and galaxy formation @xcite may depend critically on agn feedback , via coupling of the mechanical power ( via winds or jets ) of agn to the baryonic component of forming galaxies @xcite , and radio jets are known to both quench @xcite and enhance @xcite star formation .mergers , galaxy harassment , and other processes invoked in agn triggering , are common in clusters .studies of the agn and galaxy populations , kinematics , and substructures in clusters provide insight into how galaxies , agn and clusters of galaxies form and evolve .clusters also contain some of the most luminous galaxies in the local universe , the brightest cluster galaxies ( bcgs ) , which are more likely to be radio - loud than other cluster galaxies @xcite .feedback from the dissipation of the energy of expanding radio bubbles from such agn has been proposed as a solution ( e.g. , * ? ? ?* ) to the cooling - flow problem ( the lack of evidence for the extreme inflows of cold gas which are predicted to occur in clusters due to their short radiative cooling timescales ) .@xcite used a sample of 625 nearby groups and clusters of galaxies selected from the sdss @xcite .they showed that galaxies close to the centers of clusters have an enhanced probability of being radio - loud agn compared to field galaxies of similar mass .@xcite studied a sample of 573 x - ray luminous clusters , and also found an enhancement of the fraction of galaxies that were radio - loud , as well as tentative evidence for an increase in this fraction with the mass of the cluster . in this paper, we use a cluster sample more than 20 times larger than those used for the above studies .the maxbcg @xcite red sequence galaxy cluster finder was used by @xcite to create a catalog of 13,823 galaxy clusters from the sdss the largest galaxy cluster catalog assembled to date .the clusters have redshifts between 0.1 and 0.3 , and contain between 10 and 190 red sequence galaxies brighter than 0.4l@xmath2 within the virial radius .we correlate the bcg positions with the first @xcite radio source catalog , in order to study the bcgs which are radio - loud , as well as the spatial distribution of radio sources in the surrounding clusters . of the 13,823 clusters in the sdss sample , 13,240 are within the region covered by first .[ sec : cosmo]we assume an @xmath3 cosmology @xcite .we begin our analysis by finding the angular distance @xmath4 from the bcg position to the closest first source .a histogram of the results for all 13,240 fields is shown in the left panel of fig .[ fig : excess ] , along with a histogram of the mean number of random matches obtained by repeating this task around 4 positions offset by @xmath5 in the cardinal directions from each of the bcg positions .we also express the number of matches as a percentage of the total number of bcgs .note that with increasing @xmath4 , the bcg field counts initially fall below the background counts , due to the tendency of bcg fields to have a first source closer to the bcg position than the average comparison field ( which is as expected due to conservation of number ). the background counts peak at @xmath6 , which corresponds to the average distance from the comparison field position to the nearest first source .the first bin ( @xmath7 ) contains 2328 sources ( compared to a mean of @xmath8 in the comparison fields ) ; i.e. , 17.5% of fields contain a first source within 5 of the bcg position , most of which are likely associated with an agn hosted by the bcg .an additional 287 bcgs appear to host extended double - lobed radio sources ( [ sec : frii ] ) .therefore at least 19.7% of bcgs in our sample are associated with first sources.[sec : rlfrac ] some of these extended sources show up as additional counts in the @xmath9 bins of fig .[ fig : excess ] . in the right - hand panel ,we again count first sources as a function of distance , but this time we count all matching first sources , not just the closest ( so that there are multiple first matches per field , and sometimes more than one first source per annulus ) .it is easy to show that in this case the number of matches per 5 bin for the comparison fields is given by @xmath10 , where @xmath11 sq .arcsec@xmath12 is the surface density of first sources , @xmath13 is the radius of the center of the annulus , and @xmath14 is the number of fields . at @xmath15 ,the counts in the bcg fields begin to fall off less steeply than at smaller radii , as they cease to be dominated by the intrinsic size of the central source ( see [ sec : size ] ) , and begin to be dominated by other radio galaxies within the cluster . at @xmath16 ,the counts become indistinguishable from the background counts , suggesting that radio source activity is only enhanced ( on average ) within the central @xmath1mpc of the cluster ( at the mean bcg redshift for this sample @xmath17 ; see also fig .[ fig : realcounts ] ) .the power of this analysis is that even with no assumptions about the redshifts of sources , or attempts to associate multiple components of individual radio galaxies with the `` parent '' optical source , one can see the excess of radio sources extends far out into the cluster . by contrasting the left and right panels of fig .[ fig : excess ] , one can also see that there is a significant excess of multiple - component radio sources on scales @xmath18 , before the slope in the right panel changes at about @xmath19 , where the excess is mainly due to cluster radio sources unassociated with the bcg .since the maxbcg catalog provides accurate photometric redshifts for all the bcgs , we can use these to convert @xmath20 into real projected distances ( and radio flux densities into radio luminosities ; [ sec : frii ] ) . in this analysis ( fig .[ fig : realcounts ] ) , we assume that each radio source in a particular field is at the same redshift as the bcg . as can be seen from fig .[ fig : excess ] ( right panel ) , this is not true at large angular separations @xmath20 , where most of the first sources are unassociated `` background '' sources .we attempt to compensate for this by scaling the background counts in the same manner ( where we use the bcg redshift to scale the counts in all four associated comparison fields , and then compute the histogram of background counts for the entire sample ) . at smaller separations ,most of the sources really are associated with the bcg , i.e. , they are at the cluster redshift .it would of course be possible to match the other radio sources to galaxies from sdss and attempt to select only sources at similar photometric redshifts to the bcg .in this analysis we wish to concentrate on the `` pure '' radio population in the clusters , however , so we defer such radio optical matching to future work . in fig .[ fig : realcounts ] , we show the average number of first sources per field , divided by the area of each annulus , i.e. , the surface density of radio sources in the clusters .again , there is a prominent peak associated with the radio - detected bcgs , which has a wide shoulder extending to @xmath21kpc due to the presence of extended radio sources ( see [ sec : frii ] ) .even out to large radii , there is an excess of sources in the cluster environments compared to the comparison fields .since we have such a large cluster sample , we are also able to bin by other parameters such as radio luminosity , redshift , cluster richness , etc . , and look for changes in the distribution of radio sources .these results will be discussed in an upcoming paper . in fig .[ fig : rlfrac ] we show the fraction of bcgs which are radio - loud ( which , for consistency with @xcite , we define as having an associated first source with @xmath22whz@xmath12 ) . @xciteprovide bcg luminosities in sdss @xmath13-band , @xmath23-corrected to @xmath24 .we convert these luminosities to absolute magnitudes , @xmath25 , and assume an @xmath13-band stellar mass - to - light ratio log@xmath26 ( * ? ? ?we recover the result from @xcite that the radio - loud fraction increases with increasing stellar mass .@xcite found that bcgs in rich clusters were more likely to be radio - loud than those in poor clusters , but concluded that this was due to the tendency of the richest clusters to host the most massive bcgs . using our larger sample , we are able to show that _ for a given stellar mass _ the probability of a bcg to be radio loud is enhanced somewhat in richer clusters , suggesting that the wider environment plays a role in radio - loudness , in agreement with the findings of @xcite .by converting the measured radio fluxes for the bcgs into rest - frame 1.4ghz radio luminosities ( @xmath27 ) we can investigate the dependence of radio luminosity on the optical luminosity of the bcg .for consistency with previous literature ( e.g. , * ? ? ?* ; * ? ? ?* ) we adopt an average radio spectral index @xmath28 ( @xmath29 ) when applying the @xmath23-correction to calculate @xmath27 . at the relatively low redshifts considered in this paper, the choice of spectral index does not strongly affect our results . in order to account for extended emission, we use the technique of @xcite to find double - lobed radio sources associated with the bcgs , even where the bcg does not have a radio core detected by first . in fig .[ fig : frii ] , we plot the lobe opening angle ( @xmath30 , the angle subtended at the bcg position by the first sources identified as lobes ; see @xcite for more details ) for double - lobed radio sources in two size bins .in the subsequent discussion , we restrict our analysis to double - lobed sources with @xmath31 , where the fraction of false positives ( shown as the blue histogram in fig . [fig : frii ] ) is low . at small angles ( @xmath32 )we also see a small secondary peak in the cluster fields , due an excess ( compared to the comparison fields ) of real double sources in the cluster which are associated with galaxies other than the bcg . due to their true hosts being offset from the bcg position , these are sometimes erroneously identified by the algorithm as being small opening - angle sources associated with the bcg itself ( but again we emphasise that most sources with @xmath33 are real ) .we find 453 bcgs ( compared to a background expectation of 1.00 ) associated with double - lobed radio sources with @xmath33 and diameter @xmath34 ( 83 where the core is also detected , and 370 where it is not ) .additionally , we find 196 bcgs ( compared to 4.90 in the background case ) with double - lobed radio sources with @xmath33 and diameter @xmath35 ( 108 with cores , and 88 without ) .note that 362 of the 649 bcgs associated with doubles were also counted in the 2328 sources with a first source within 5 discussed in [ sec : rlfrac ] . in fig .[ fig : lo ] , we plot @xmath27 ( inferred from the total fluxes of the radio components associated with each bcg ) for the 649 double - lobed sources ( red squares ) as a function of @xmath25 . the black squares in fig .[ fig : lo ] are the 1966 bcgs _ not _ identified as doubles , which nevertheless have a first source within 5 .we also plot the empirical division between fr i and fr ii radio galaxies from @xcite , where we have assumed @xmath36 ( the isophotal magnitude to 24.5 mag arcsec@xmath37 from @xcite ) .in contrast to the results of @xcite for sdss quasars , we find very little excess at @xmath38 for diameters @xmath39 ( i.e. , there are very few doubles as large as this in our bcg sample ) . since @xcite studied quasars with median redshift @xmath40 and associated radio emission from first , their sample has a higher mean radio luminosity , so this result is perhaps unsurprising , even given the smaller angular diameter distance to our sources ; the physical sizes of our sources are about a factor 4 smaller than the fr iis of @xcite , peaking at 55kpc ( fig . [fig : friihist ] ) , [ sec : size ] and dropping to zero at around 200kpc . as can be seen from fig .[ fig : realcounts ] , 200kpc is where the slope of the histogram changes , showing that the excess matches at @xmath41kpc are dominated by other radio galaxies in the cluster , and at @xmath42kpc , by the presence of multiple - component radio sources associated with the bcg .[ fig : lo ] shows that most of the double - lobed sources are located near the fr i / ii division from @xcite , which adds weight to the argument that radio agn in cluster environments at low redshift do not tend to be powerful fr iis @xcite they could certainly be seen by us if present .it is clear that the upper envelope in radio luminosity in this plot shows an increase in radio luminosity with optical luminosity ( a proxy for stellar mass ) , in agreement with previous results ( e.g. , * ? ? ?we note that only a few detected sources ( 7.2% ) are fainter than @xmath43whz@xmath12 ( essentially due to the first flux limit ) .for such a radio luminosity to be due entirely to star formation would require sfr @xmath44m@xmath45yr@xmath12 @xcite , so we conclude that radio - loud agn are the dominant source of radio emission in those of our bcgs detected in first .of course , even in sources with @xmath46mjy , star formation may contribute to the radio emission . in some cases , this star formation will also manifest as optical emission lines , whereas in other cases the star formation may be optically obscured .by matching the maxbcg and first catalogs , we were able to determine that 19.7% of bcgs in our sample are detected in the radio ; the majority ( 92.8% ) of these are radio - loud agn ( @xmath47whz@xmath12 ) .the radio loud fraction has a strong dependence on stellar mass of the host , increasing from @xmath48% at @xmath49@xmath50 to @xmath51% at @xmath52@xmath50 . for a given stellar mass , bcgs in rich clustershave a higher tendency to be radio - loud than bcgs in poor clusters or groups , suggesting that the wider environment has some influence on radio - loudness .signs of double - lobed emission are seen ( at the resolution of first ) in 24.8% of the 2615 bcgs associated with first sources .the physical size of these double - lobed sources peaks at 55kpc , with a tail extending to @xmath21kpc .their radio luminosities place them close to the fr i / ii division .there is an excess of radio sources in our clusters , which are associated with other cluster members , extending out to @xmath1mpc , but the central bcgs are , overall , the dominant source of agn activity in these clusters .future investigations using this dataset could address the relationship between optical spectroscopic indicators of star formation and agn activity ( as determined from the sdss `` value added '' catalogs of @xcite and @xcite ) , radio loudness / luminosity , and environment , as well as extending the first - sdss match to other galaxies in the clusters ( as determined by their photometric redshifts ) .for those bcgs undetected in first , it is possible to `` stack '' the first maps @xcite to determine the average radio properties of sources too faint to be detected individually . in this way, we can study bcgs with radio emission two orders of magnitude below the first flux limit .these results will be discussed in a future paper .work was performed under the auspices of the u. s. department of energy , national nuclear security administration by the university of california , lawrence livermore national laboratory under contract no . w-7405-eng-48 .research made use of the vla first survey ( nrao programs ab628 , ab879 and ab950 ) and the sdss . the national radio astronomy observatory is a facility of the national science foundation operated under cooperative agreement by associated universities , inc .becker , r. h. , white , r. l. , & helfand , d. j. 1995 , , 450 , 559 best , p. n. , von der linden , a. , kauffmann , g. , heckman , t. m. , & kaiser , c. r. 2006 , astro - ph/0611197 brinchmann , j. , charlot , s. , white , s. d. m. , tremonti , c. , kauffmann , g. , heckman , t. , & brinkmann , j. 2004 , , 351 , 1151 brggen , m. , & kaiser , c. r. 2002 , , 418 , 301 condon , j. j. 1992 , , 30 , 575 croft , s. , et al .2006 , , 647 , 1040 croton , d. j. , et al .2006 , , 365 , 11 de vries , w. h. , becker , r. h. , & white , r. l. 2006 , , 131 , 666 de vries , w. h. , hodge , j. a. , becker , r. h. , white , r. l. , & helfand , d. j. 2007 , arxiv:0704.2074 , aj , august issue haehnelt , m. g. , & kauffmann , g. 2000 , , 318 , l35 hill , g. j. , & lilly , s. j. 1991 , , 367 , 1 hodge , j. , et al .2007 , , in prep .jarvis , m. j. , rawlings , s. , eales , s. , blundell , k. m. , bunker , a. j. , croft , s. , mclure , r. j. , & willott , c. j. 2001 , , 326 , 1585 kauffmann , g. , et al .2003 , , 341 , 33 kauffmann , g. , et al .2003 , , 346 , 1055 koester , b. p. , et al .2007 , , 660 , 221 koester , b. p. , et al .2007 , , 660 , 239 ledlow , m. j. , & owen , f. n. 1996 , , 112 , 9 lin , y .- t . , & mohr , j. j. 2006 , astro - ph/0612521 rawlings , s. , & jarvis , m. j. 2004 , , 355 , l9 spergel , d. n. , et al .2003 , , 148 , 175 von der linden , a. , best , p. n. , kauffmann , g. , & white , s. d. m. 2006 , astro - ph/0611196 springel , v. , di matteo , t. , & hernquist , l. 2005 , , 361 , 776 white , r. l. , helfand , d. j. , becker , r. h. , glikman , e. , & de vries , w. 2007 , , 654 , 99 york , d. g. , et al .2000 , , 120 , 1579 histogram of the total number of first sources in all 13,240 bcg fields , as a function of angular distance @xmath20 ( in arcseconds ) from the target ( bcg or offset comparison field ) position , in bins ( i.e. , annuli ) of width 5 . the black histogram shows the counts in the fields containing the bcg , and the grey histogram shows the mean counts in the comparison fields .the left panel shows just the closest first source to the target position ( i.e. , for each field , we consider only the single first source closest to the target ) , whereas the right panel shows _ all _ first sources in the target field ( which can occasionally include multiple first sources in a given annulus ) .the right - hand vertical axis in both panels shows the counts per bin as a percentage of the total number of fields ( 13,240 ) . ]histogram of the average projected surface density of first sources ( black ) , as a function of distance in kpc from the bcg position ( assuming , for each field , that all first sources are at the bcg photometric redshift ) .the grey histogram shows the comparison field counts , at the mean of four positions offset by 30 from the bcg positions . ]the percentage of bcgs that are radio - loud agn , as a function of stellar mass ( black line ) . also plotted are the same data split into two subsamples ; bcgs in poor clusters ( @xmath53 members ; blue line ) and rich clusters ( @xmath54 members ; red line ) . at a given stellar mass ,bcgs in rich clusters have a slightly higher probability of being radio - loud . ]histogram of opening angles for double and triple radio sources associated with bcgs ( black ) and expected contamination from random superpositions of sources which appear to be doubles / triples ( grey ) .the left panel shows doubles / triples with major axes between 0 and 30 , and the right panel , major axes between 30 and 60 .the bin size is 5 . ]rest - frame 1.4ghz radio luminosity as a function of sdss @xmath13-band absolute magnitude for the bcgs detected in first , @xmath23-corrected to @xmath24 .bcgs with a first source within 5 are shown as black squares , except for cases where our `` double - detector '' algorithm ( see [ sec : frii ] ) finds a double radio source associated with the bcg ( sometimes with a radio - detected core , sometimes without ) .these 649 doubles and triples are plotted as red squares , with the radio luminosity determined from the total integrated radio flux of the components .the solid line is the division between fr i and fr ii radio galaxies from @xcite ( we make no correction from their @xmath55 to our @xmath25 ) . ]

<S> we correlate the positions of 13,240 brightest cluster galaxies ( bcgs ) with @xmath0 from the maxbcg catalog with radio sources from the first survey to study the sizes and distributions of radio agn in galaxy clusters . </S><S> we find that 19.7% of our bcgs are associated with first sources , and this fraction depends on the stellar mass of the bcg , and to a lesser extent on the richness of the parent cluster ( in the sense of increasing radio loudness with increasing mass ) . </S><S> the intrinsic size of the radio emission associated with the bcgs peaks at 55kpc , with a tail extending to 200kpc . </S><S> the radio power of the extended sources places them on the divide between fr i and fr ii type sources , while sources compact in the radio tend to be somewhat less radio - luminous . </S><S> we also detect an excess of radio sources associated with the cluster , instead of with the bcg itself , extending out to @xmath1mpc . </S>

astro-ph9605024

n - body simulations give reliable predictions for the redshift dependence of the two - point correlation function of the density field , @xmath27 , as a function of @xmath28 .a convenient power - law parameterization to describe the evolving correlation function of galaxies is ( @xcite ) @xmath29 where the lengths @xmath30 are measured in physical ( proper ) co - ordinates . for this double power - law approximationthe predicted evolution of clustering in the mass field is faster for @xmath31 ( @xmath32 ) than it is for low-@xmath28 values , for instance @xmath33 for @xmath34 ( @xcite ) .therefore , measurement of the redshift evolution of clustering can be used to test the gravitational instability theory of structure formation and the relation of galaxy clustering to dark matter clustering , and provides a constraint on @xmath28 .knowledge of these quantities enables predictions of the distribution of assembly times of dark halos , which on the relatively small scales investigated here is of great interest for the mass evolution of galaxies . at present , observational measures of clustering evolution are uncertain simply due to the difficulties of assembling large samples of faint galaxies with sufficient sky coverage to give a statistically representative sample . at low redshift the form of nonlinear galaxy clusteringis accurately established ( @xcite ) , with a basic characterization of its dependencies on galaxy color , luminosity , and morphology . at higher redshiftsthe clustering is only now being directly measured ( @xcite ) , although the small fields leave concerns that field - to - field variations are not yet well controlled . the galaxy luminosity function and its color dependenceevolve substantially over the redshift 0 to 1 interval ( @xcite ) .differential luminosity evolution of blue and red galaxies , in which the blue galaxies are less correlated at low redshift , can cause the apparent correlation of a magnitude - limited sample to change faster than either of the two underlying populations are changing . in this paperwe report the clustering properties of a very deep redshift survey selected in the @xmath0 band .a near - ir selected survey has the enormous advantage that both k - corrections and the evolutionary corrections are small , allowing galaxy luminosities to be identified with total stellar mass with reasonable confidence .the hawaii @xmath0-band survey ( @xcite ) with a couple hundred galaxy redshifts , is large enough to be useful for correlation studies .furthermore this survey contains galaxies up to a redshift of 2.19 , which provides a fairly large redshift baseline over which correlation changes can be measured .the next section summarizes the sample properties .measures of the angular correlation are given in section 3 and of the real space correlation function in section 4 . in section 5the correlations of red and blue galaxies and of low- and high - luminosity galaxies are compared .section 6 discusses the redshift evolution of galaxy correlations and compares the available data to various model predictions .section 7 summarizes our results .all measurements in this paper assume @xmath35 and @xmath6 .it should be noted that correlation amplitudes at @xmath36 are reduced by about 30% for @xmath37 .the hawaii @xmath0-selected redshift survey of two fields constitutes a nearly complete sample down to @xmath38 , @xmath39 , and @xmath40 mag .the survey is described in detail elsewhere ( @xcite ) although this analysis uses 20 new redshifts that have been recently obtained to complete the @xmath20 selected subsample .the sample of 248 galaxies with redshifts in an area of about 27 square arcminutes compares favourably with the moderate redshift cnoc sample , about 200 galaxies in a single field covering 221 square arcminutes ( @xcite ) , and with the 591 galaxies covering about 71 square arcminutes ( a result of the high sampling rate in @xmath41 strips ) in the cfrs study ( @xcite ) .on the other hand , the redshifts here extend all the way from 0.08 to 2.19 , with 80% between 0.28 and 1.39 .the number in any one redshift interval is sufficient to make useful correlation measurements ( figure [ fig : nz ] ) .all magnitudes and colors used in this paper are k - corrected .the sky positions of the objects are plotted in figure [ fig : xy ] , where open squares are galaxies , crosses represent stars , and objects that were not observed or which lack secure redshift identifications are shown with triangles .the galaxy redshifts , plotted against the projected physical distance in the ra direction from the field center , are shown in figure [ fig : pie ] .the symbol area is proportional to the luminosity of the galaxy .the sample is known to be somewhat incomplete for the redder galaxies at the faintest magnitudes , which are generally expected to be galaxies beyond redshift one with low star formation rates .incomplete samples , provided that they have no spatial bias , do not pose a problem for correlation studies , provided that the unclustered background distribution is generated from a smoothed version of the observed redshift distribution .we will approximate the observed redshift distributions as being constant over the various subranges of interest , which tests indicate to be an adequate approximation for these data .the angular correlation function is estimated as @xmath42 ( @xcite ) where in a given range of angles @xmath43 is the number of data pairs , @xmath44 is the number of pairs between the data sample and a uniform random sample .this estimator is particularly useful when the clustering amplitude is significantly less than unity , as it is here .the resulting angular correlation function for the full photometric sample , @xmath45 mag , is shown in figure [ fig : aw ] .the errors are assigned using the bootstrap method ( @xcite ) with 100 resamplings .the angular galaxy correlations are diluted by the uncorrelated foreground stars , because the full photometric sample is used , which requires that the correlation amplitude be corrected upward by a factor of @xmath46 , where @xmath47 is the fraction of objects which are stars in the two combined fields .we estimate that @xmath48 , based on the stellar fraction of the spectroscopically identified sample that are stars .the uncorrected correlation function is @xmath49 , from which we find that the correlation angle of the galaxies is @xmath50 arcseconds , where the error is the internal error of the fit .several recent studies in the optical region have found correlation angles of @xmath51 at @xmath52 mag ( @xcite ) , @xmath53 at @xmath54 mag ( @xcite ) and in a much fainter sample to @xmath55 mag , @xmath56 ( @xcite ) . sinceour mean @xmath57 we note that the angular correlation that we observe is substantially larger than for a comparable @xmath58- or @xmath59-band sample .this is a consequence of the @xmath0-band sample s being fundamentally more correlated than optical samples ; we provide direct evidence for this below . in a survey done with a multi - object spectrographthere is the possibility that the angular correlation of the galaxies with redshifts is biased relative to the parent sample as a result of instrumental constraints for the selection of objects for spectroscopy .the simplest test for a selection effect of this type is to measure the ratio of @xmath43 pairs as a function of separation in the photometric and redshift sample as normalized to the total numbers in the two samples .we find that there is a 10% reduction in the number of pairs within about @xmath60 , and a 5% overselection at separations around @xmath61 , diminishing with increasing angle .this bias is generally less than the statistical errors , so we do not apply any geometric corrections in the correlation measurements .the velocity precision in this redshift survey is not adequate to allow a measurement of the redshift space correlation function at small scales .however , we can measure the projected real space correlation function , @xmath62 ( @xcite ) .the primary drawback of this estimator is that it averages over long cylindrical shells in redshift space , so it has a very broad window function ( @xcite ) .the real space correlation is derived from a power - law fit to the projected correlation function , eq .[ eq : wp ] , calculated from the galaxy sample .operationally , @xmath63 is the integral over @xmath64 of the 2d correlation function @xmath65 , where @xmath66 and @xmath64 are the proper separation of galaxy pairs in the projected and redshift directions , respectively . because our sample is not very large the errors are dominated by the small number statistics , rather than any complications of the estimator .we estimate @xmath65 as @xmath67 , in the @xmath68 co - ordinates .the random sample is 5 to 10@xmath69 points .the sum over the @xmath64 co - ordinate is cut off at a practical range of 10 in proper co - ordinates at the redshift of the object ( @xcite ) .this value was selected from a range of trial values as being approximately the optimal value to maximize the signal - to - noise .this cutoff distance is sufficiently large that the correction to the integral for the correlation beyond the cutoff is everywhere less than about 20% , and is ignored .the redshift distribution of the random sample is generated assuming that the unclustered redshift distribution is constant over the various redshift and luminosity ranges . altering the redshiftranges shows this to be an entirely adequate approximation for these data .the errors are bootstrap estimates , reduced by a factor of @xmath70 ( @xcite ) .the @xmath63 measured over the redshift ranges 0.20.4 , 0.40.8 , 0.81.2 , and 1.21.6 are shown in figure [ fig : wp ] .the galaxies have minimum luminosities of @xmath71 mag in the three higher redshift bins and @xmath72 in the lower redshift range , to increase the sample size .a rest - frame luminosity of @xmath73 mag is about half of @xmath74 .the @xmath63 are fitted to a power - law correlation function , @xmath75 ( @xcite ) .the gamma function factor is 3.68 for @xmath76 . over the redshift range @xmath77 we find @xmath78 at a mean @xmath9 . for the @xmath79 rangewe find @xmath80 at a mean @xmath11 , and for @xmath81 , @xmath82 at a mean @xmath83 .using wider projected radius bins we find that for the galaxies with redshifts 1.21.6 , @xmath84 at a mean @xmath85 .this high redshift correlation is formally an upper limit on the basis of the bootstrap error estimates , although the poisson errors do indicate a significant measurement .we will accept the result as an indicative measurement , because it remains unclear how best to estimate the errors .the galaxies in this highest redshift bin are dominated by the faint @xmath20 subsample and are not complete in the @xmath0 subsample this selection bias likely leads to an underestimate of the correlation at this redshift if the enhanced correlation of red - selected galaxies over blue - selected ones seen at lower redshift is present at this redshift .the decrease in the correlation length with increasing redshift is significant within these data although the quoted errors are purely the internal errors of the fit and do not account for field - to - field differences beyond these two patches .the fitted @xmath86 , based on our @xmath0-band sample alone .the color dependence of faint galaxy clustering is an important clue to the formation mechanisms of galaxies .higher mass galaxies are expected to be more correlated at formation and subsequently environmental effects can modify the correlations ( @xcite ) . in figure[ fig : wc ] the @xmath63 are shown for the @xmath87 galaxies with colors redder or bluer than @xmath88 mag ( k - corrected , rest - frame ratios between the flux @xmath89 at 3500and at 21000 on the ab magnitude system , see @xcite ) .the red galaxies , with @xmath90 , are about 5 times ( with about a 50% error ) more strongly clustered than the blue objects , which have @xmath91 , both at fixed @xmath76 .the red galaxies appear to have a substantially steeper correlation slope , @xmath92 , than our adopted value of 1.8 , although a survey with larger sky area is needed to assess the slope . the average @xmath93 mag for the blue sample , whereas it is @xmath94 mag for the red sample , which is quite a small luminosity difference for the large correlation difference ( @xcite ) .the blue population , at a mean @xmath20 of 23.7 mag , has strong lines , with a mean @xmath95 of 41 .therefore , these are very likely to be substantially brightened relative to their intrinsic luminosities at a more normal star formation rate .the blue galaxy population must largely disappear from the normal galaxy population at low redshift , since its correlation length would grow to only about 2 given our estimated @xmath96 . within the luminosity range explored in the low redshift apm survey ,no population is this weakly correlated ( @xcite ) although the low luminosity , @xmath97 , late type galaxies have @xmath98 . we conclude that these high redshift faint blue galaxies can not become a significant component of the low redshift galaxy population above luminosities of @xmath23 .it is possible that these objects merge with higher luminosity galaxies to drive their evolution .the clustering of lower luminosity galaxies is weaker than that of high luminosity galaxies .for @xmath99 the galaxies with @xmath100 have @xmath101 and for @xmath102 those with @xmath103 have @xmath104 . in both these redshift ranges ,the higher - luminosity galaxies are more correlated that the lower luminosity ones .the luminosity difference is about 2.5 magnitudes .it is quite unlikely that the high - redshift , high - luminosity , weakly - correlated population could have evolved into a weakly - clustered , low - luminosity , low - redshift population .first , it requires that the clustering is being reduced in physical co - ordinates with time .second , about 4 magnitudes of fading per unit redshift is required , whereas a much slower rate of luminosity density evolution of the population average is observed ( @xcite ) .third , it would require that the high - luminosity galaxies at lower redshift originate from some unobserved high redshift population an effect which @xmath0-band observations minimize through their relatively small redshift corrections . on the other hand , this discussion does emphasize that a more precise definition of comparable galaxy samples at low and high redshift would be very desirable , ideally done based on the intrinsic properties of the galaxies themselves .the cross - correlation of low - luminosity galaxies , @xmath105 mag , with high - luminosity galaxies , @xmath106 mag , is shown in figure [ fig : w2 ] for the @xmath107 range ( expanded slightly to boost the sample size ) .although the sample is smaller than is really desirable it shows quite intriguing how much more strongly the low luminosity galaxies cluster to high luminosity `` hosts '' within 100 , beyond which the cross - correlation drops to a value similar to the field cross - correlation of low - luminosity galaxies .the effect is directly visible in the redshift diagrams of figure [ fig : pie ] .the enhanced cross - correlation of close pairs is not primarily the result of an enhanced star formation which would raise the optical band luminosity between 1 and 2 magnitudes , and hence would increase the numbers above some flux limit .the lower luminosity galaxies have substantially stronger ; however the expected accompanying increase in the @xmath0 luminosity is small .consequently these data favour a genuine increase in close pairs at small separations over the power - law correlation .to follow the clustering amplitude as a function of redshift we need to compare the results of a variety of redshift surveys , having varying @xmath108 and @xmath92 .the quantity @xmath109 , is a useful measure of the clustering amplitude , which can be interpreted either as the amplitude at 1 , or , given that these quantities are normally the results of a fit to data over a range of scales , @xmath109 is a measure of the average correlation within a fixed proper volume . at low redshiftthere are several recent measurements of clustering with a range of sample definitions .the @xmath110-selected apm survey finds @xmath111 and @xmath112 giving @xmath113 at @xmath114 ( @xcite ) .the @xmath59-selected lcrs finds @xmath115 and @xmath116 giving @xmath117 at a mean @xmath118 ( @xcite ) .the iras - selected correlation function is @xmath119 and @xmath120 ( @xcite ) for @xmath121 .there is no large @xmath0-band selected redshift survey at low redshift ; however the angular correlation of bright k - selected galaxies is measured , from which we estimate @xmath122 and 13.8 at @xmath123 and @xmath124 , respectively , with @xmath12520% systematic errors ( @xcite referred to as bffs ) .a preliminary measurement in the lcrs survey finds that the red galaxies have a correlation comparable to these @xmath0-band results ( @xcite ) . at higherredshift we have the results of this paper , the @xmath30-selected cnoc field survey at a mean redshift of 0.36 ( @xcite and work in progress ) , which gives @xmath126 for the red half of the sample and 7.5 for the blue half .the cfrs results ( @xcite ) provide correlation estimates over the redshift range 0.2 - 0.9 .they find @xmath127 for @xmath128 , @xmath129 for @xmath130 , and @xmath131 for @xmath132 , where we have adjusted their correlation lengths for @xmath37 to our @xmath6 using their formula .the cfrs data indicate @xmath133 , which would be a relatively poor description of the @xmath0-selected data here . as a direct comparison with the cfrs measurements ,we define a complete subsample limited at @xmath54 mag from our data .we find that @xmath134 for @xmath135 , which is statistically identical to the cfrs measurement of @xmath136 ( adjusted to @xmath6 ) over the @xmath137 range .the correlation amplitudes are plotted against redshift in figure [ fig : ev ] . for comparison the similarly measured correlations from n - body simulations ( @xcite )are also shown , all scaled with a linear multiplicative factor roughly to fit the lcrs correlation measurement .the fitted @xmath92 values from these simulations are in the range 1.8 - 2.0 , which is compatible with the @xmath0-band sample but not the shallower slopes usually seen in optically selected samples .the scaling factors are in the range of 0.55 to 0.70 , which is a little larger than desirable ( ideally one would do specially matched simulations ) , however , these factors are small compared to the factor of 20 or so in the evolution of the correlation functions .there are two conclusions to be drawn from figure [ fig : ev ] .first , optically selected galaxies appear to always be significantly less correlated than @xmath0-selected galaxies , the difference being typically about a factor of two in the amplitude .second , for the @xmath0-band and red - selected samples , which should be least sensitive to galaxy population evolution , we see that the evolution of galaxy clustering is reasonably well described by an @xmath34 model .however , the amplitude measured in the n - body simulation is multiplied by a factor of 0.55 , which is approximately the square of the bias factor , @xmath138 .hence , either the galaxies are anti - biased with respect to the matter clustering , or , the normalization of used for the n - body simulations should have been approximately @xmath139 . the observed @xmath0-band selected evolution appears to be too slow with redshift to readily agree with the @xmath31 predictions , although the formal level of exclusion is strongly dependent on the low redshift normalization .the second conclusion to be drawn is that the amplitude of the correlations increases as the color used to select the objects becomes redder and there is weak evidence that @xmath92 is also steeper in the @xmath0 sample .the fitted correlation length of luminous @xmath0-selected galaxies over the redshift range 0.2 to 1.2 is substantially stronger than that found for optically selected samples , about a factor of two in the amplitude , @xmath109 .the galaxy correlation amplitude is measured at a mean @xmath85 as @xmath140 ( formally an upper limit , but deficient in the more strongly clustered faint red galaxies ) , @xmath141 at @xmath83 , @xmath142 at @xmath143 , and @xmath144 at @xmath9 . togetherthese give a clustering @xmath145 .the red galaxies are about a factor of 5 more correlated than the blue galaxies , which have @xmath146 .these blue galaxies have a mean equivalent width in the line of 41 . togetherthis can be taken as strong evidence that the faint blue galaxies are an intrinsically weakly correlated population ( therefore likely low mass ) with a high star formation rate that brightens them into the range of much more strongly correlated red galaxies .these galaxies are so weakly correlated that for our measured growth of correlations , @xmath145 , they would grow to a current epoch correlation length of @xmath147 .this correlation length is less than that measured in any galaxy population at low redshift ( @xcite ) , so these faint blue galaxies can not by themselves make a significant contribution to the current epoch galaxy population .overall the @xmath0 selected galaxy correlation evolution is somewhat too slow with redshift to be easily consistent with the evolution of the matter correlation function for @xmath148 .a @xmath149 and @xmath150 would describe both the amplitude and its evolution , if these galaxies are tracing the matter clustering . to further test models of correlation evolutionrequires large datasets of precision comparable to that available in current low redshift surveys with good control over population changes with redshift .we thank the referee for constructive criticism of an earlier version of this paper .rgc acknowledges the financial support of an nserc grant .this work was partly based on observations obtained with the nasa / esa hubble space telescope .research on the hawaii survey fields was supported by the state of hawaii and by nasa through grants go-5399.01 - 93a , go-5922.01 - 94a , and go-6626.01 - 95a from the space telescope science institute , which is operated by aura , inc ., under nasa contract nas5 - 26555 .baugh , c. m. , gardner , j. p. , frenk , c. s. & sharples , r. m. 1996 , , submitted , ( astro - ph/9608142 ) brainerd , t. g. , smail , i , & mould , j. r. 1995 , , 275 , 781 colin , p. , carlberg , r. g. , & couchman , h. m. p. 1996, , submitted cowie , l , l. , songaila , a. hu , e. m. , & cohen , j. g. 1996 , , 112 , 839 davis , m. & peebles , p. j. e. 1983 , , 267 , 465 efron , b. & tibshirani , r. 1986 , _ statistical science _, 1 , 54 ellis , r. s. , colless , m. , broadhurst , t. , heyl , j. , & glazebrook , k. 1996 , , 280 , 235 fisher , k. b. , davis , m. , strauss , m. a. , yahil , a. , & huchra , j. p. 1994, , 267 , 927 groth , e. j. & peebles , p. j. e. 1977 , , 217 , 385 hudon , d. & lilly , s. j. 1996 , , 469 , 519 koo , d. c. & szalay , a. s. 1984 , , 282 , 390 landy , s. d. & szalay , a. s. 1993 , , 412 , 64 lefvre , o. , hudon , d. , lilly , s. j. , crampton , d. , hammer , f. & tresse , l. 1996 , , 461 , 534 lidman , c. e. & peterson , b. a. 1996 , , 279 , 1357 lilly , s. j. , tresse , l. , hammer , f. , crampton , d. , & lefvre , o. 1995 , , 455 , 108 lilly , s. j. , lefvre , o. , hammer , f. , & crampton , d. 1996 , , 460 , 1l lin , h. 1996 , ph .d. thesis , harvard university lin , h. , yee , h. k. c. , carlberg , r. g. , & ellingson , e. 1997 , , 475 , 494 loveday , j. , maddox , s. j. , efstathiou , g. , & peterson , b. a. 1995 , , 442 , 457 mo , h. j. , jing , y. p. & brner , g. 1992 , , 392 , 452 peacock , j. a. 1996 , , in press ( astro - ph/9608151 ) peebles , p. j. e. 1980 , _ the large scale structure of the universe _ , ( princeton : princeton university press ) shepherd , c. w. , carlberg , r. g. , yee , h. k. c. & ellingson , e. 1997 , , in press tucker , d. l. 1994 , ph . d. thesis , yale university and personal communication tucker , d. l. , oemler , a. jr . , kirshner , r. p. , lin h. , shectman , s. a. , landy , s. d. , schechter , p. l. , mller , v. , gottlber , s. , and einasto , j. 1996 , , submitted .

<S> angular and spatial correlations are measured for @xmath0-band selected galaxies , 248 having redshifts , 54 with @xmath1 , in two patches of combined area @xmath2 </S><S> arcmin@xmath3 . </S><S> the angular correlation for @xmath4 mag is @xmath5 . from the redshift sample we find that the real - space correlation , calculated with @xmath6 , of @xmath7 mag galaxies ( k - corrected ) is @xmath8 at a mean @xmath9 , @xmath10 at @xmath11 , @xmath12 at @xmath13 , and @xmath14 at @xmath15 , the last being a formal upper limit for a blue - biased sample . in general , these are more correlated than optically selected samples in the same redshift ranges . over the interval @xmath16 galaxies with red rest - frame colors , @xmath17 @xmath18 mag , have @xmath19 whereas bluer galaxies , which have a mean @xmath20 of 23.7 mag and mean equivalent width @xmath21 , are very weakly correlated , with @xmath22 . for our measured growth rate of clustering , this blue population , if non - merging , can grow only into a low - redshift population less luminous than @xmath23 . </S><S> the cross - correlation of low- and high - luminosity galaxies at @xmath24 appears to have an excess in the correlation amplitude within 100 . </S><S> the slow redshift evolution is consistent with these galaxies tracing the mass clustering in low density , @xmath25 , relatively unbiased , @xmath26 , universe , but can not yet exclude other possibilities . .^-2.1 </S><S> mm .+.6 mm 1 200 </S>

0904.1062

it is generally believed that qcd exhibits a rich phase structure in an extreme environment such as at high temperature and high baryon chemical potential . for the chiral phase transition ,it is a widely accepted view that the critical point(s ) ( i.e. the end point of the first - order phase boundary ) should exist at finite temperature and density .recently , a schematic @xmath0-@xmath1 phase diagram with one critical point was extensively adopted in the literature@xcite . in the last decadethe color - superconducting ( csc ) phase has attracted a lot of theoretical interest and prompted extensive studies of dense and cold quark matter @xcite . at asymptotically high density that justifies the perturbative qcd calculations ,the color - flavor locked ( cfl ) phase @xcite has been established as the ground state of quark matter . for the intermediate density region which may exist in the core of a compact star ,the nonperturbative features of qcd play a more important role in the phase structure of qcd , and both cfl and non - cfl csc phases may appear in this region. the possible emergence of a cfl or a non - cfl csc phase around the chiral transition boundary at low temperatures should affect the chiral phase transition , and hence the interplay between the chiral condensate and the diquark condensate may result in an unexpected phase structure of qcd .such an example was first presented in @xcite in a two - flavor nambu - jona - lasinio ( njl ) model with the vector interaction included : it was found that the repulsive vector interaction can lead to a two - critical - point structure in the @xmath0-@xmath1 phase diagram of qcd , which is due to the fact that the density - density correlation induced by the vector interaction effectively enhances the competition between the chiral and diquark correlations while weakening the first - order chiral restoration@xcite .we note that the renormalization - group analysis @xcite , the chiral instanton@xmath2anti - instanton molecule model @xcite , and the truncated dyson - schwinger model of qcd @xcite all support the existence of the vector - vector four - quark interaction. the vector - vector interaction may be responsible for the vacuum properties of vector mesons in low - energy effective theories of qcd@xcite . in a quite different context, a realization of a similar two - critical - point structure of the qcd phase diagram has been recently conjectured @xcite : that is , the @xmath3-breaking vertex may induce a new critical point .this conjecture is based on a general ginzburg - landau theory constrained by qcd symmetries .there , the three - flavor anomaly term generates the cubic coupling between the chiral and diquark condensates .it has been argued that the resultant crossover of chiral restoration at small temperatures embodies the hadron - quark continuity hypothesis @xcite .further investigation is necessary to clarify whether the new critical point really exists or not with reasonable parameter . beside the above two cases , another mechanism which can induce multiple critical - point structurewas demonstrated in @xcite , where it was first disclosed that the positive electric chemical potential @xmath4 required by the charge - neutrality constraint plays a similar role as the repulsive vector interaction on the chiral phase transition in a four - quark interaction model : in a simple two - flavor njl model , the same two - critical - point structure found in @xcite was obtained .in addition , positive @xmath4 also represents the magnitude of disparity between fermi spheres of u and d quarks when taking into account the local charge - neutrality constraint . for an asymmetric homogeneous system including two different flavor quarks with the mismatched fermi spheres , the energy gap of the cooper pairing between these two - flavor quarks can increase with temperature .this unusual behavior of the energy gap is due to the smearing of the fermi surface by temperature .for the two - flavor neutral csc phase , this unconventional thermal behavior of the diquark condensate can lead to a special competition between the chiral condensate and the diquark condensate , which may be enhanced with increasing temperature . for some model parameter regions ,this abnormal competition induced by @xmath4 can result in a nontrivial three - critical - point phase structure @xcite .the competition of two order parameters with the external constraint(s ) should also suggest a general mechanism for a realization of a multiple critical - point structure , which may have some implications to study the phase transitions in condensed matter physics .note , however , that a very simple two - flavor njl model with only scalar quark - antiquark and diquark interaction was used in ref .@xcite : if the model produces a relatively large vacuum quark mass , no multiple critical points appear in such formalism even though the charge - neutral constraint is taken into account for the csc phase . for a more realistic situation , both the vector interaction and the strange quark degree of freedom should be taken into account . because in the chiral phase transition similar roles are played by the repulsive vector interaction and the electric chemical potential under the neutrality constraint, one can expect that the chiral restoration will be weakened more significantly when simultaneously taking into account both the ingredients . in this case, it is possible that the multiple critical point structures shown in @xcite will become more robust , which we will show in the present work . in addition , for the neutral quark matter system , because of the quark density discrepancy , the vector interaction should also give different contributions to the dynamical chemical potentials for u and d quarks .this may have important influence on the property of the asymmetric homogeneous csc phase which suffers from the so called chromomagnetic instability characterized by the negative meissner mass squared @xcite .this is a significant observation and will be shown to be the case in the following sections .there have been extensive studies within the njl model on the phase diagram of dense and locally neutral three - flavor quark matter with strange quarks explicitly included @xcite : in the strong coupling case , the chiral breaking phase at low temperature is bordered by the 2csc phase , while for the weak coupling case it is surrounded by both normal quark matter and a gapless phase called g2csc . however , no new critical point was found in @xcite , in contrast to the result reported in @xcite .notice , however , that the vector interaction was not included in @xcite , and hence only a strong first - order chiral restoration happened in the low temperature region .we can expect that the qualitative feature of the phase diagram , even for the three - flavor quark matter reported in @xcite , may change when the repulsive vector interaction is taken into account together with the charge - neutrality constraint .again , we shall see that this is actually the case , and there appear multiple critical points in the phase diagram of three - flavor quark matter .the paper is organized as follows . in the next section ,a nonlocal two - flavor njl model is used to investigate the @xmath5 phase structure of qcd , with both vector interaction and charge - neutrality being taken into account .the extension of the work to the two - plus - one - flavor case with the so - called kobayashi - maskawa-t hooft(kmt ) term@xciteis presented in sec.iii .the final section is devoted to a summary and concluding remarks .we will demonstrate the joint effect of the vector interaction and neutral electric - charge constraint on the qcd phase diagram within a nonlocal two - flavor njl model .njl - type models have been extensively used to investigate the csc phase transition at moderate and large densities @xcite : see the previous study without the csc @xcite .the advantage of a njl model is that it can investigate the interplay between the chiral condensate , the diquark condensate and the quark number density on the same footing .for the two - flavor case , a nonlocal njl model @xcite is adopted here , which has the lagrangian density , @xmath6-g_v \sum_{i=0}^3\left [ \left ( \bar{q}(x ) \gamma^\mu\tau_i q(x ) \right)^2 + \left ( \bar q(x ) i \gamma^\mu \gamma_5 \tau_i q(x ) \right)^2 \right ] \nonumber \\ & + & g_d \sum_{a } \left[\bar{q}(x ) \gamma_5 \tau_2\lambda_a q_c(x ) \right ] \left [ \bar{q}_c(x)\gamma_5\tau_2\lambda_a q(x ) \right ] , \ , \label{lagrangian2f}\end{aligned}\ ] ] where @xmath7 and @xmath8 stands for the dirac charge conjugation matrix .the three coupling constants , namely @xmath9 , @xmath10 , and @xmath11 , belong to the scalar mesonic channel , the vector mesonic channel , and the diquark channel , respectively .the current quark mass matrix is given by @xmath12 in two flavors , and we shall work in the isospin symmetric limit with @xmath13 .we note that @xmath14 s are the antisymmetric gell - mann matrices ( i.e. @xmath15 runs over @xmath16 only ) for the color su(3 ) group , while @xmath17 and the @xmath18 s are the unit matrix and pauli matrices in flavor space , respectively .in contrast to the scalar interaction , the vector part has @xmath19 flavor symmetry , and hence the vector terms can be decomposed into @xmath20 without the flavor mixing term like @xmath21 , although there are terms like @xmath22 .in contrast to the local njl model , lagrangian ( [ lagrangian2f ] ) is formulated with a nonlocal interaction , which is controlled by a form factor @xmath23 .this type of model can be considered as a special case of the truncated dyson - schwinger equation with a separable effective gluon propagator or an instantaneous nonlocal chiral quark model .the main purpose of adopting such a nonlocal interaction model in our study is to deal with the one - loop ultraviolet divergence , especially for the calculation of the meissner mass squared at finite temperature @xcite . for convenience, we follow the model parametrization in @xcite and use the so - called lorentzian form factor with the form @xmath24 where @xmath25 is the fourier transformation of the form factor @xmath23 . in eq.([formfactor ] ) , @xmath26 is a dimensionless parameter and @xmath27 stands for the scale parameter of the model . to check the sensitivity of the main results on the model parameter choice ,three sets of model parameters are adopted in this paper , which are listed in tabel.([tab1 ] ) by choosing @xmath28 , if it can give reasonable vacuum properties of qcd . ]the other three model parameters , namely @xmath27 , @xmath9 and @xmath29 , are determined by the vacuum physical quantities of the pion mass @xmath30 , the pion decay constant @xmath31 , and the quark condensate @xmath32 .note that the value of the current quark mass is around @xmath33 and the gell - mann - oakes - renner relation holds well for these parametrizations of the model parameters @xcite . .modelparameter sets for two - flavor nonlocal njl .[ cols="^,^,^,^,^,^",options="header " , ] in general , there exist nine possible two - quark condensates for the two - plus - one - flavor case with the lagrangian ( [ lagrangian2 ] ) : three chiral condensates @xmath34 , three diquark condensates @xmath35 , and three vector quark condensates @xmath36 , where @xmath37 and @xmath38 range from 1 to 3 , which stand for three flavors and three colors , respectively . at the mean - field level , the thermodynamic potential for the two - plus - one - flavor njl model including the charge - neutrality constraints , is @xmath39 where @xmath40 stands for the contribution from free leptons .note that , for consistency , @xmath40 should include the contributions from both electrons and muons . since @xmath41 and @xmath42 , ignoring the contribution from muons has little effect on the phase structure .therefore , we use @xmath40 given by the last term in eq .( [ omega - l ] ) .it should be stressed here that , for simplicity , the contributions of the cubic mixing terms among three different condensates , such as @xmath43 , @xmath44 , and @xmath45 , are neglected in eq .( [ omega2 ] ) .these terms arise from the kmt interaction which may or may not affect the phase structure . in particular it was argued in @xcite that the cubic mixing term between chiral and diquark condensates may play an important role in the chiral phase transition in the low temperature region . beside the direct contribution of these cubic terms to the thermodynamic potential , the flavor mixing terms arising from the kmt interaction also have influence on the dispersion relationship of quasiquarks .for example , both the diquark condensate and the quark number density contribute to the dynamical quark mass .therefore , it is a very interesting topic to investigate the possible effect of these cubic coupling terms on the phase diagram by using dynamic models of qcd . leaving the discussion of this interesting problem to our future work , herewe just simply assume that none of these mixing terms makes a qualitative difference in the phase diagram .because of the large mass disparity between the strange quark and the u ( d ) quark , the favored phase at low temperature and moderate density might be 2csc phase rather than cfl as demonstrated in the two - plus - one - flavor njl model@xcite .these studies suggest that the strange quark mass is close to or even larger than @xmath1 near the chiral boundary , which means that the strange quark density is considerably smaller than that of the u and d quarks .therefore , for the two - plus - one - flavor case , the light quarks still play the dominant role around the chiral boundary , and the density mismatch between u and d quarks under the electric - charge - neutrality constraint is still similar to the two - flavor case . since the main purpose of our study is to investigate the influence of the neutral csc phase on the chiral phase transition by taking into account the vector interaction , we only consider the 2csc phase in the following .the inverse quark propagator in the 2csc phase in the two - plus - one - flavor case still takes the same form as @xmath46 , with the extended matrixes @xmath47 and @xmath48 in three - flavor space .the constituent quark mass is given by @xmath49 and the effective quark chemical potentials take the form @xmath50 the quantity @xmath51 ( @xmath52 ) still has the same form as eq.([average ] ) [ eq.([mismatch ] ) ] with @xmath53 .ignoring the mass difference between the u quark and the d quark ( the mass difference is very small @xcite ) , the last term in eq .( [ omega2 ] ) has an analytical form which greatly simplifies the numerical calculation . adopting the variational method, we get the eight nonlinear coupling equations @xmath54 since @xmath55 is near zero around the chiral transition region @xcite , taking it to zero makes little difference in the calculational result and the nonlinear equations can be reduced to 7 .similar to the two - flavor case , the vector interaction @xmath56 is also taken as a free parameter in the following , and the diquark coupling @xmath57 is fixed to the standard value . because of the contribution from the kmt interaction , the ratio @xmath57 from fierz transformation should be 0.95 rather than 0.75 when only considering the four - quark interaction @xcite .note that in this case the effective four - quark interaction which determines the quark constituent mass in vacuum is @xmath58 , and the standard value of @xmath59 should be 0.75 .the phase diagrams with a multiple critical - point structure for different vector interactions are shown in fig.[fig : pd-3flavor ] .one can see that these phase diagrams are very similar to fig.[fig : pdset2 ] , the two - flavor case with parameter set 2 .this is reasonable since these two models almost reproduce the same constituent quark ( u and d ) masses and have similar scale parameters .figure[fig : pd-3flavor ] indicates that the kmt interaction does not change the possible multiple critical - point structures for the chiral phase transition . for simplicity ,the unstable regions with chromomagnetic instability are not plotted in fig.[fig : pd-3flavor ] .the calculation of the meissner mass squared in the 2csc phase for the two - plus - one - flavor case is straightforward but complicated .including the s quark should have little effect on the value of the meissner masses calculated according to the formula for the two - flavor case @xcite since the s quark does not take part in cooper pairing .we can expect that the critical points e , f , and g should still be free from the chromomagnetic instability , as in figs.[fig : pdset1 ] and [ fig : pdset2 ] , since the large strange quark may give a positive contribution rather than a negative one to the meissner masses squared .as for critical point h , it may be located in the unstable region and could be safe from the instability because the relatively large @xmath56 may suppress the unstable region to lower @xmath0 and higher @xmath1 . usually , it is argued that the instantons should be screened at large chemical potential and temperature .therefore , compared to it s vacuum value , the coupling constant @xmath60 is expected to be reduced around the chiral boundary . for smaller @xmath60 ,the flavor mixing effect is suppressed and the mass mismatch between the s quark and the u(d ) quark becomes larger .accordingly , the influence of the s quark on the chiral restoration is weakened , and the situation approaches the two - flavor case . on the other hand , with decreasing @xmath60 , the u(d ) quark mass also decreases since the contribution from the s quark massis reduced .this means the first - order chiral restoration will be weakened when decreasing @xmath60 .correspondingly , the coe region should be more easily formed with the influence of the vector interaction and the neutral csc phase , which favors the multiple critical - point structures or crossover for chiral restoration at low temperature .of course , the produced u(d ) quark vacuum constituent masses with different model parameters of the two - plus - one - flavor njl model may range from 300 - 400 mev , which are all phenomenologically acceptable just as the two - flavor case .then , one can expect that all the critical - point structures found in sec .[ sec : two flavor ] should also appear in the two - plus - one - flavor case , even considering the axial anomaly interaction term .in this paper , we have explored the effect of the repulsive vector - vector interaction combined with electric - charge neutrality in @xmath61 equilibrium on the chiral phase and csc phase transitions within both two - flavor and two - plus - one - flavor njl models .for the two - flavor case with the presence of the neutral csc phase , we demonstrated that , with the help of the repulsive vector interaction , there always exists a parameter window in the njl models which favors the appearance of a multiple chiral critical - point structure for a wide range of the vacuum quark mass , i.e. , from 300 mev to 400 mev .besides the two- and three - critical - points structures found in @xcite and @xcite , respectively , we have shown for the first time that a four - critical - points structure of the qcd phase diagram can be realized ; such a multiple critical - point structure is caused by the joint effect of positive @xmath4 and @xmath10 . because the dynamical strange quark mass is still relatively large near the boundary of the chiral transition , the multiple critical - point structures present in the two - flavor case also appear in the two - plus - one - flavor case . for the intermediate diquark coupling case , the number of critical points changes as @xmath62 with an increasing vector coupling in the two - plus - one - flavor njl model . in general, one can expect that different model parameters may possibly give other order of the number of the critical points as the vector coupling is increased .although our analysis is based on a low - energy effective model which inherently has , more or less , a parameter dependence , we have seen that the physical mechanism to realize the multiple critical - point structure is solely dependent on the basic ingredients of the effective quark dynamics and thermodynamics .therefore , we believe that the results obtained in the present work should be taken seriously and examined in other effective models of qcd , or hopefully lattice qcd simulations .our result also has a meaningful implication for the study of phase transitions in condensed matter physics .that means some external constraints enforced on the system can lead to the formation or expansion of the coexisting phase , and the competition between two order parameters can give rise to multiple critical points .last but not least , we emphasize that we have shown for the first time that the repulsive vector interaction which should generically exist between the quarks does suppress the chromomagnetic instability related to the asymmetric homogeneous 2csc phase . with increasing vector interaction , the unstable region associated with chromomagnetic instability shrinks towards lower temperatures and higher chemical potentials .this means that the vector interaction can partially or even totally resolve the chromomagnetic instability problem .note that to cure the chromomagnetic instability , inhomogeneous asymmetric color superconductivity phases such as the loff phase and the gluonic phase were proposed in the literature . for the inhomogeneous phase , beside the condensate @xmath63 , there is no reason to rule out the appearance of another new condensate , @xmath64 , when considering the vector interaction .the effect of both the timelike vector condensate and the spacelike condensate on the asymmetric inhomogeneous csc phase will be reported in our future work@xcite .we acknowledge kenji fukushima for useful discussion .one of the authors ( z. z. ) is grateful for the support from the grants - in - aid provided by japan society for the promotion of science ( jsps ) .this work was partially supported by a grant - in - aid for scientific research by the ministry of education , culture , sports , science and technology ( mext ) of japan ( no .20540265 and no .19@xmath6507797 ) , by yukawa international program for quark - hadron sciences , and by the grant - in - aid for the global coe program `` the next generation of physics , spun from universality and emergence '' from mext . as a review , see , m. a. stephanov , prog .* 153 * , 139 ( 2004 ) [ int .j. mod .a * 20 * , 4387 ( 2005 ) ] ; pos * lat2006 * , 024 ( 2006 ) [ arxiv : hep - lat/0701002 ] . k. rajagopal and f. wilczek , arxiv : hep - ph/0011333 .m. g. alford , k. rajagopal and f. wilczek , nucl .b * 537 * , 443 ( 1999 ) [ arxiv : hep - ph/9804403 ] .m. kitazawa , t. koide , t. kunihiro and y. nemoto , prog .* 108 * , 929 ( 2002 ) [ arxiv : hep - ph/0207255 ] .s. klimt , m. lutz and w. weise , phys .b * 249 * , 386 ( 1990 ) . z. zhang , k. fukushima and t. kunihiro , phys .d * 79 * , 014004 ( 2009 ) [ arxiv:0808.3371 [ hep - ph ] ] .m. huang and i. a. shovkovy , phys .d * 70 * , 051501(r ) ( 2004 ) [ arxiv : hep - ph/0407049 ] ; _ ibid . _d * 70 * , 094030 ( 2004 ) [ arxiv : hep - ph/0408268 ] .s. b. ruester , v. werth , m. buballa , i. a. shovkovy and d. h. rischke , phys .d * 72 * , 034004 ( 2005 ) [ arxiv : hep - ph/0503184 ] .h. abuki and t. kunihiro , nucl .a * 768 * , 118 ( 2006 ) [ arxiv : hep - ph/0509172 ] .m. kobayashi and t. maskawa , prog .* 44 * , 1422 ( 1970 ) ; m. kobayashi , h. kondo and t. maskawa , prog .* 45 * , 1955 ( 1971 ) ; + g. t hooft , phys .* d14 * , 3432 ( 1976 ) ; phys . rep . * 142 * , 357 ( 1986 ) .m. g. alford , k. rajagopal and f. wilczek , phys .b * 422 * , 247 ( 1998 ) [ arxiv : hep - ph/9711395 ] .s. m. schmidt , d. blaschke and yu .l. kalinovsky , phys .c * 50 * , 435 ( 1994 ) .r. d. bowler and m. c. birse , nucl .a * 582 * , 655 ( 1995 ) [ arxiv : hep - ph/9407336 ] .d. blaschke , g. burau , yu .l. kalinovsky , p. maris and p. c. tandy , int .j. mod .a * 16 * , 2267 ( 2001 ) [ arxiv : nucl - th/0002024 ] .d. gomez dumm , d. b. blaschke , a. g. grunfeld and n. n. scoccola , phys .d * 73 * , 114019 ( 2006 ) [ arxiv : hep - ph/0512218 ] .d. n. aguilera , d. blaschke , h. grigorian and n. n. scoccola , phys .d * 74 * , 114005 ( 2006 ) [ arxiv : hep - ph/0604196 ] .h. grigorian , phys . part .lett . * 4 * , 223 ( 2007 ) [ arxiv : hep - ph/0602238 ] .e. v. gorbar , m. hashimoto and v. a. miransky , phys . lett .b * 632 * , 305 ( 2006 ) [ arxiv : hep - ph/0507303 ] ; phys .lett . * 96 * , 022005 ( 2006 ) [ arxiv : hep - ph/0509334 ] ; phys . rev .d * 75 * , 085012 ( 2007 ) [ arxiv : hep - ph/0701211 ] .

<S> we investigate the combined effect of the repulsive vector interaction and the positive electric chemical potential on the chiral phase transition by considering neutral color superconductivity . the chiral condensate , diquark condensate and quark number densities are solved in both two - flavor and two - plus - one - flavor nambu - jona - lasinio models with the so called kobayashi - maskawa-t hooft term under the charge - neutrality constraint . we demonstrate that multiple chiral critical - point structures always exist in the nambu - jona - lasinio model within the self - consistent mean - field approximation , and that the number of chiral critical points can vary from zero to four , which is dependent on the magnitudes of vector interaction and the diquark coupling . </S><S> the difference between the dynamical chemical potentials induced by vector interaction for u and d quarks can effectively reduce the fermi sphere disparity between the two flavors of diquark pairing . thus the vector interaction works to significantly suppress the unstable region associated with chromomagnetic instability in the phase of neutral asymmetric homogeneous color superconductivity . </S>

1002.4376

in order to learn more about the inner essence of nature , particle physicists smash together small particles , such as protons , at large energies , and use huge detectors to detect whatever comes out .the higher the energy , the smaller distances can be examined , which also means that nature at smaller scales can be studied and reveal its secrets .the story is in fact much more complicated .the protons that clash together are composite particles made up by quarks and gluons that can not be isolated and studied in their own , but are always confined inside the proton .the collision between two protons is therefore better described as a collision of two approaching bunches of particles , and for the outcome to tell us about nature we need to know what takes part in the collision .therefore , so called parton distribution functions ( pdfs ) describe how the momentum of the proton is shared between partons , i.e. quarks and gluons .the pdfs make up one of the ingredients of computer simulations , which combine theories and models in order to predict the outcome , and make it possible to test theory against experiment . in this thesiswe have included the very latest parton distribution functions in the monte carlo event generator pythia8 @xcite .we compare them and examine how the difference in pdfs affect results of simulations .we also compare the results to real data collected by experiments at tevatron @xcite and examine how the differences change when increasing the energy up to the level of a fully operational lhc ( 14 tev ) @xcite .the structure of the thesis is as follows . in section 2we introduce some of the concepts of the theory of strong interactions , quantum chromo dynamics ( qcd ) . in section 3we move on to describe monte carlo generators ( such as pythia8 ) and section 4 tells the story of parton distribution functions .section 5 describes how the pdfs were included in pythia8 and we compare the different pdfs in section 6 .subsequently we study results from simulations of minimum bias events and hard qcd events in section 7 and 8 respectively .finally we conclude with summary and conclusions in section 9 .the standard model of particle physics is the joint theory of three of the four fundamental forces of nature : electromagnetism , weak and strong force .the part describing the strong interactions , quantum chromo dynamics , is responsible for holding together protons and neutrons in the nucleus as well as quarks inside the nucleons .the strong force is mediated by gluons , which interact with particles that carry color charge , and it grows stronger with distance .colored particles are therefore confined inside colorless hadrons such as the proton and can never be observed and studied as free particles .most calculations in qcd rely on a perturbative expansion in the coupling of the theory .perturbative qcd uses the feynman diagrams and rules to calculate matrix elements and thereby obtain cross sections for different processes .this technique relies on the coupling of the strong force , @xmath0 , to be small enough to make the expansion converge and allow the first terms to be a good approximation . at high energiesthe strong force is weak , the quarks experience asymptotic freedom and the approximation of truncating the expansion at low orders is good .as one moves down to lower energy , increasingly higher orders would be required , but the perturbation expansion in @xmath0 quickly becomes very complicated and at small enough scales it finally breaks down completely .there are many possible feynman graphs for each process and all of them will contribute to the cross section . at next - to - leading orderthe diagrams start to involve loops which introduce integrals over the phase space of the internal lines .these loop integrals diverge , but at each order the real ( extra particle in the final state ) and virtual ( extra particle internal ) divergences combine , in a far from trivial way , to cancel out all infinities and leave finite results .this results in very complicated calculations and therefore many processes have only been calculated to leading order .the matrix element calculation gives rise to divergences in two cases , when two partons are collinear and when the energy of a parton is small , soft divergence .how accurate the approximation of the low order expansion is depends on the size of the coupling @xmath0 .the running of the coupling , @xmath1 , is necessary in order to absorb infinities in the theory .this is called renormalization and the running is determined by the renormalization group equation @xcite @xmath2 where @xmath3 the two constants depend on the number of flavors , @xmath4 , that have their threshold below the energy scale , @xmath5 and @xmath6 . solving this equation and introducing the mass scale @xmath7 yields @xmath8 the first termis the first order expression and the dots are terms that decrease in relative importance for large @xmath9 . from thisit can be seen that @xmath0 decreases at large @xmath9 as @xmath10 and therefore become very small , but also that the coupling increases towards infinity as @xmath9 approaches @xmath11 .the cross section describes how likely it is that an interaction will take place . in the regions where the matrix elements can be calculated these give the cross sections for hard sub - processes , such as @xmath12 . to describe the cross section for the sub - process we introduce the mandelstam variables , @xmath13 where @xmath14-@xmath15 are the four - momenta of the particles .the differential cross section is given by @xmath16 where @xmath17 and @xmath18 are energies of the two outgoing partons and the @xmath19-function ensures conservation of energy and momentum .@xmath20 is the sum ( average ) over initial- ( final- ) state spins and colors .@xmath21 is the matrix element and for the @xmath22 process , assuming massless partons , @xmath23 to leading order .the sub - process cross section , @xmath24 , is a function of the momenta of the partons , the value of the strong coupling at a relevant energy scale @xmath25 of the process and the factorization scale @xmath26 .@xmath26 can be seen as the scale which separates long- and short - distance physics @xcite , i.e. partons with transverse momentum less than @xmath26 are considered as part of the proton and are absorbed into the pdf .the standard choice is to set @xmath27 @xcite , @xcite .assuming no @xmath28 of the interacting partons , the center - of - mass energy squared for the entire collision , @xmath29 , and for the colliding sub - system , @xmath30 , are related by @xmath31 hence the product @xmath32 determines the energy fraction available in the sub - process , while the ratio gives the rapidity @xmath33 and thereby determines the direction of motion of the colliding sub - system .in large rapidity events the interacting partons therefore have very different momentum fractions .the average transverse momentum is often good as an energy scale of the process and an upper limit on the transverse energy , @xmath28 , is set by the energy available in the subprocess @xmath34 with equality only for back to back scattering of two partons perpendicular to the beam axis . in minimum bias events a typical @xmath35 gev , setting a lower limit on the product , @xmath36 at the lhc , for such an event .the cross section from the matrix element calculations divergence as @xmath37 when @xmath38 .this must be regularized for low transverse momentum which is accomplished by introducing a parameter , @xmath39 , ( to be discussed more in section [ sec : minbias ] ) such that @xmath40 to obtain the total cross section for the process when two protons collide the probability to find the partons with momentum fraction @xmath41 inside the proton have to be taken into account , see fig .[ fig : feynman ] .the total cross section is then obtained by summing over the different possible partons and integrating over the allowed momentum fractions @xmath41 , and @xmath42 .one parton from each respective proton , with momentum fractions @xmath43 and @xmath44 , take part in a hard scattering sub - process producing final - state particles.,scaledwidth=60.0% ] @xmath45 where @xmath46 and @xmath42 are the momenta of the two incoming protons .@xmath47 is the parton distribution function of parton i , which will be described in more detail in section [ chap : pdf ] .experiments such as proton collisions at high energy are too complicated to make predictions based solely on calculations from first principles of the standard model and therefore one has to rely on monte carlo generators .the generators start with a central hard collision and then combine matrix elements , parton showers , multiple parton - parton interactions and hadronization models to predict the outcome.@xcite as described in the previous chapter , matrix - element calculations do not describe qcd accurately over the entire phase space . in regions where higher order terms are necessary to give good approximations the mc generators use the technique of parton showers , which approximates the higher order terms . at each step in the showerthe emission probability is calculated and an emission is generated , building up an event with many outgoing partons .this method is a good approximation when the shower can be strongly ordered , i.e. the ordering variable such as virtuality decreases with each emission .that is the case for the largest part of the cross section but not for the hard interactions , where outgoing partons escape at large angles @xcite .the parton showers can be subdivided into two types , initial - state radiation ( isr ) and final - state radiation ( fsr ) .as the names suggest , isr describes how the incoming proton branches into many quarks and gluons before the collision while fsr takes the products of the collision and describes how they branch into more particles .fsr is evolved in strictly ordered virtuality where at each step the virtuality is smaller than in the previous .isr is more complicated .it is ordered in increasing virtuality but since the simulation starts with the hard interaction the isr has to be evolved backwards @xcite .this backward evolution of isr is one of the most challenging parts of a generator @xcite .the showers and matrix elements are good approximations in two entirely different regions .matrix elements are good for hard collisions , producing particles at large transverse momentum while parton showers are good at low energies and for collinear partons . in order to make use of their respective advantages they need to be matched at some energy scale to cover the entire phase space andin a way that assures there is no double counting @xcite , @xcite .pythia8 uses showers ordered in @xmath28 which has an advantage over mass ordered showers ( which were used in the previous version pythia6.4 @xcite ) because it automatically takes care of some effects due to coherence between emissions .these , however , are not the only possible choices and for example herwig @xcite , the other large general monte carlo generator , uses energy weighted angular ordering .this has advantages related to coherence effects but can not be used over the full phase space and hence leaves some regions to be filled in by high order matrix elements @xcite .after the fsr the hadronization starts , and the partons produced in the previous stages split up and combine into colorless hadrons than can survive and possibly reach a detector .it is also at this level that experiments collect their data .the hadronization has not yet been possible to determine from the theory and therefore models are required .pythia8 uses the lund string model @xcite where partons tie color connections to each other which break by forming quark antiquark pairs , while herwig uses cluster fragmentation @xcite .there is no reason why , in a proton collision , there should only be one parton from each proton that interacts , and in fact it is not even the case most of the time .therefore the so called multiple interaction framework simulates the additional interactions . as the collision energy increases , partons at smaller @xmath41become involved and more particles can be produced in the isr .multiple interactions are therefore increasing and the models for multiparticle interactions are put to the test as the energy at lhc rises .pdfs are used in monte carlo generators at several stages .first in calculating the cross section for the hard collision to take place , which we have seen in eqn .[ eqn : cross ] .secondly , isr use the pdfs since it is the partons inside the proton that will shower , so the pdfs are needed to describe what we have to start with and is used in the backwards evolution of the isr .finally , pdfs also enter through cross section calculations in the multiple interactions framework .the static properties of the proton are dominated by the valence quarks , i.e. two up quarks and one down quark . the dynamic picture of the proton is less simple .quarks interact by exchanging gluons , which can split up into quark antiquark pairs and send out additional gluons .the proton thus consists of three valence quarks , gluons and sea quarks which all go under the common name of partons .parton distribution functions ( pdfs ) describe how the momentum of the proton is distributed among partons and at leading order they can be interpreted as the probability to find a parton inside the proton with a certain momentum fraction . at higher order this simple picture fails since , for example , the nlo pdfs can be negative .there are six quarks with antiquarks and adding the gluon , the proton has , potentially , 13 pdfs . however , the heavy top quark / antiquark is not included taking the number down to eleven . when a gluon splits it is always into a quark antiquark pair , and hence the distributions of quarks and antiquarks should be the same ( once the valence quarks are excluded ) .therefore strange , charm and bottom quarks used to be considered to have exactly the same distribution as their respective antiquark , setting the total number of distributions to 8 .however this is not necessarily true and this symmetry can be broken , which is also the case ( at least ) for the strange quark / antiquark .the proton can be seen as accompanied by a kaon cloud , since the proton can split into a @xmath48 and a @xmath49 , where the @xmath29 quark sits in the first and the @xmath50 in the second .there is no reason to believe that the @xmath29 quark in the @xmath7 should have the same momentum as the @xmath50 in the @xmath51 @xcite .therefore some of the newer pdfs have a small difference between the two distributions .the pdfs obey a number of relations such as the the momentum sum rule @xmath52 where the sum is over all partons in the pdf , which states that the total momentum of all the partons must equal the momentum of the proton .the pdfs are functions of the fraction of the proton s momentum @xmath41 carried by the parton and of @xmath9 , which can be interpreted as a measure of resolving power .once the parton distributions are known for one specific @xmath9 , they can be evolved to higher @xmath9 by use of the dglap equation @xcite @xmath53 where @xmath54 is the parton distribution function for parton @xmath55 and @xmath56 is the splitting function .the outcome of experiments depend on how the momentum is distributed but there is no way to directly calculate what the distribution should be .therefore several groups work on parameterizing and fitting pdfs to available data @xcite-@xcite .the large @xmath41 behavior of the pdfs are constrained by fixed target experiments . at smaller @xmath41the constraints come from deep inelastic scattering ( dis ) at electron - proton colliders such as hera .the dis data have high statistics and therefore dominate the pdf fits .this gives good knowledge of the quark distributions but the gluon distribution is harder to obtain , since gluons do not directly interact with electrons . in hadroncolliders the picture is much more messy , since there is now two composite particles and the gluon distribution is therefore the least constrained part of the pdfs . for more details on the data sets of pdf fitssee @xcite and @xcite .there are pdfs of several different orders , lo , nlo and some nnlo .the general - purpose generators are all leading order and therefore one would like to combine them with leading - order pdfs .however , we know for example that higher orders generally give positive contributions to cross sections and in recent years some modified lo pdfs have been released , specifically tailored for leading - order monte carlo generators .these try to simulate some of the effects of nlo calculations by compensating for known shortcomings of the leading - order . among other thingsthese pdfs allow for a non - conservation of momentum by relaxing the momentum sum rule , i.e. the partons inside the proton are allowed to have a total momentum larger than the momentum of the proton .this permits the pdfs to grow large in some regions without decreasing in others and thereby simulate the effects of some of the next - to - leading - order corrections , in particular allowing a large value of the gluon distribution at small @xmath41 without compromising the quark distributions at large @xmath41 .the mc - adapted pdfs released so far are lo * and lo * * @xcite from the mrst group and mc1 , mc2 and mcs @xcite from the cteq group .all of them except mcs have relaxed the momentum sum rule .mc1 use a leading - order running of @xmath0 while lo * , lo * * and mc2 use next to leading - order running .lo * * also has a change in argument , to @xmath57 rather than @xmath9 , for @xmath0 for high - x evolution .mcs has more freedom in the parameterization and allows for change with scale , to simulate nlo cross section calculations ( a feature we do not make use of in pythia8 ) .mc1/2/s are all fitted to a combination of real data and nlo pseudo data in an attempt to obtain the ideal pdf for leading - order mc generators .pythia8 @xcite has so far been distributed with the option to choose between two pdfs , grv94l @xcite and cteq5l @xcite , which are both fairly old .many new and improved pdfs have been released and made available to pythia8 simulations only through lhapdf @xcite .the lhapdf package has grown quite large and in that process also a bit slow , also the code is written in fortran while the community is changing to c++ .it is desirable to include some pdfs directly into pythia8 because it can speed up simulations , make pythia8 more complete and make it easier to switch between different frequently used pdfs .furthermore some of the latest pdfs have not yet been included in lhapdf .therefore we incorporate ten new pdfs from the mrst @xcite-@xcite , mstw 2008 @xcite , cteq6 @xcite and cteq mc @xcite distributions into pythia8 .two of them are nlo which are not intended for mc use , but included for comparison .the main danger with them is for low-@xmath28 processes .inclusion of the pdfs was done in cooperation with the mstw and cteq collaborations , @xcite-@xcite , and the pdfs are listed in tab .[ pdftab ] . + .pdfs that are now included in pythia8 .we included all but two , which were already available in pythia8 . [ cols= " < , < , < " , ] in this sectionall simulations are done at the cm - energy of 1960 gev , if not explicitly stated otherwise .so as not to cram the pictures , not all sets are shown all the time .we have tried to choose a selection of pdfs to show in each plot what represents both the extremes and the middle way .the rapidity distributions of the outgoing particles at the parton level when only the @xmath58 sub - process is considered are presented in fig .[ fig : rapid - strip ] .mstw lo has a broader distribution than the rest of the pdfs with more particles at larger rapidities as an effect of the large gluon distribution at the small @xmath41 .we can also see that lo * * closely resembles mc2 , as does lo * and mc1 while the two nlo distributions are lower in the central rapidity region .the remaining leading - order distributions all show similarities to the mc - adapted ones . turning on the rest of the pythia8 machinery and looking at the distribution of charged particles after hadronization , shown in fig .[ fig : rapid ] , the distribution changes its shape .there are now more particles at larger rapidities as a result of fragmenting color field strings stretched out to the beam remnants and most of the differences between the pdfs get blurred .some differences still remain and mstw lo is still smaller for central rapidities , and as a remnant of the wider distribution lack the inward dents that all other pdfs have at rapidities around @xmath59 .the peaks of lo * * and mc2 are a bit sharper than for lo * and mc1 but the trace of the lower value of the cteq66 at central rapidity is gone .minimum bias events are dominated by @xmath60 interactions but if we go back to the partons in fig .[ fig : rapid - strip ] and select only the outgoing quarks , i.e. mainly from @xmath61 , large underlying differences are revealed , see fig .[ fig : rapid - quarks ] .the ct09 and cteq66 distributions show smooth shapes leading up to a peak at zero rapidity , while lo*/ * * have a very flat distribution at central rapidities .the outlier is once again the mstw lo distribution which shows peaks at large rapidities .this is once again caused by the large gluon distribution at small @xmath41 and we believe a reason , that it is so much more prominent for the quarks , is that mstw lo has a relatively small gluon distribution at large @xmath41 which in asymmetric @xmath62 interactions somewhat compensates , while the up - quark distribution is similar to the other pdfs . the normal lo distributions from cteq show similar distributions as the mc - adapted ones from mstw . actually cteq5l show similar peaks as mstw lo , and looking at the gluon distribution at low energies we can see that before it freezes as @xmath63 it is larger than in all other pdfs .there is also a clear difference in the amount of quarks produced where cteq66 has a much larger amount and thereby produces much more quark initiated jets than the rest of the pdfs .the multiplicity distributions are similar for most pdfs .the two nlo distributions stand out as two extremes in different directions where mstw nlo has the highest peak and the shortest tail as shown in fig .[ fig : multdist ] . all mc - adapted pdfs except mcshave a peak slightly shifted to larger multiplicities but are different in height where the two from mrst have a larger peak value .the three normal leading - order distributions are all similar and we only show the cteq6l .[ fig : ptdist ] show the @xmath28 distribution which is almost completely independent of pdf and the overlap of the three distributions in each figure makes them impossible to distinguish . repeating the simulations butnow with @xmath0 set differently for the different pdfs , i.e. the value at @xmath64 and the order of the running is determined by the pdfs , does not significantly change anything , once the multiplicity has been retuned . as an illustration of this point , the results with the two different @xmath0 for the lo ** pdf are shown in fig [ fig : comp ] .ct09mcs uses a varying running of @xmath0 which is not implemented in pythia8 and the simulations with this pdf use a nlo running @xmath0 instead .this should not change the results much since the varying @xmath0 is fine tuning @xcite . increasing the energy to the level of a fully operational lhcenhances the differences seen at tevatron energy , especially for mstw lo and the two nlo pdfs .the multiplicity of these three evolve with energy in a different way than for the other pdfs .the rapidity distribution , shown in fig .[ fig : rapid - lhc ] , naturally extends to larger rapidities and the total charged particle multiplicity increases since the energy available is larger .mstw lo here gives a much broader distribution and also has a much higher total charged particle multiplicity .this is because as the energy increases even lower values of @xmath41 come into play , so that the effect of the gluon distribution in this region has larger impact on the results .the two nlo pdfs have a flatter peak than the mc - adapted pdfs and are similar in shape to mstw lo but have much smaller multiplicity .the rest of the pdfs evolve in a fashion similar to the mc - adapted pdfs shown in the figure but with some more variation .mc1 and lo * are a little bit larger at central rapidities than mc2 and lo**. the multiplicity distributions in fig .[ fig : mult - lhc ] also show increased differences except for the two nlo pdfs which converge at this energy . not only is there a larger mean multiplicity and peak at a higher value , but the same distributions that stand out from the rest with the rapidity also do so with their charge particle multiplicity distribution .this is the case also for the @xmath28 distributions but these are still very similar , excluding the mstw lo with its large multiplicity , fig .[ fig : pt - lhc ] .the analyses in rivet ensure that the comparisons with data have the same cuts and corrections as the original experiment .therefore only the central pseudorapidity region is used and also cuts in transverse momentum @xcite .@xmath28 spectra of charged particles in fig .[ fig : sigmapt - riv ] show the same relative shape for all pdfs , which gives too large values at the low @xmath28 end , then decreases compared to data and gives too small differential cross sections at the high end .the slope shows some differences depending on the choice of pdf .mc - adapted pdfs and the cteq6l give results that are the closest to data , while mstw lo and nlo are further away than the rest .+ the @xmath65 spectrum of particles , neutral particles included , shows larger dependence on the pdfs but it is the same distributions that result in the values closest to data .since we have not done a complete tune , this is to indicate the importance of the pdfs and results far away from data are not necessarily the fault of the pdfs .however the @xmath65 distribution is less dependent on details of the mc and therefore easier for pdf developers to consider in tunes .the mc - adapted pdfs from cteq as well as the cteq6l reproduce data well , while mstw lo goes down to less than half the cross section of data at the larger energy end .mstw nlo peaks at higher energies than the rest .all pdfs give a too large value at the peak , but then decrease too fast and differ the most from data at the high energy end .+ comparing the evolution of the average transverse momentum with charge multiplicity from our simulations to data ( which used to be one of the distribution the mc generators struggled the most with ) , show that all pdfs reproduce the data fairly well and also behave in a similar fashion to one another .they all give too low @xmath66 at low multiplicity and then increase relative to data so that they get closer as the multiplicity increases , fig .[ fig : ptcmult - riv ] .the only pdf that is slightly different than the rest is mstw lo .quarks and gluons produced in collisions fragment because of the color confinement and produce jets of color neutral hadrons . the definition of what is a jet is far from trivial and identifying jets from data is even more difficult .this task is performed by different jet algorithms relying on closeness in either direction or momentum space .we compare the inclusive jet cross sections from our mc simulations with data , collected by the cdf experiment at tevatron run 2 @xcite , over five pseudorapidity intervals ranging up to @xmath67 . in the experimental analysisthe jets are identified with the midpoint cone algorithm and also compared to results with the @xmath68 algorithm @xcite .we are interested mainly in the low @xmath28 region , in order to examine whether simulations can be improved by introducing a correction ( @xmath69- ) factor , as done , for example , in @xmath70 production .a @xmath69-factor is a factor that multiplies the cross section in order to indirectly include known corrections from next - to - leading - order calculations , and by that increase the total production rate .we also examine the rapidity , multiplicity and transverse momentum distributions for the individual hadrons .the simulations were done with a @xmath28 cut at @xmath71 gev .the inclusive jet cross section in fig . [ fig : jetcrossdata ] drops rapidly with increasing @xmath28 and spans over several orders of magnitude .the cteq5l distribution , shown in fig .[ fig : jetcrossdata ] , yields results which are lower than data , with a ratio between @xmath72 and @xmath73 .the ratio remains fairly constant with @xmath28 but it is closer to unity at medium pseudorapidities .since the differences between experiments and simulations are hard to see in the main window of fig.[fig : jetcrossdata ] we only show the @xmath74 ratio in the following figures . the results with mrst lo * * , mstw lo , cteq6l1 , cteq66 , ct09mc1 and ct09mc2 are shown in fig .[ fig : jetcross2]-[fig : jetcross5 ] . mrst lo * * starts with a much too large cross section and the ratio decreases when @xmath28 rises .this behavior is the strongest at low pseudorapidity and as we move to larger @xmath75 the ratio gets smaller and flatter . all mc - adapted pdfs , except mcs ,show this type of behavior .mc2 and mc1 give results with very similar shapes but the mc2 cross section is larger .mrst lo * is related to lo * * much in the same fashion as mc1 to mc2 .mstw lo and cteq6l give cross sections which have similar behavior as with cteq5l , i.e. the ratio is less dependent on @xmath28 than with the mc - adapted pdfs .mstw lo results are less depending on the pseudorapidity than cteq5l , fig .[ fig : jetcross2 ] .cteq6l1 gives a ratio which starts to decrease with @xmath28 at larger rapidities .ct09mcs gives a too low cross section , is once again different from the other mc - adapted pdfs and gives results which behave in a way more similar to those of the normal leading - order distributions .actually , the results with the two nlo pdfs are the closest to data at central pseudorapidities but the ratios decrease towards @xmath76 at larger @xmath75 .+ + + + the main feature is the surprising decrease in the cross section ratio from low to high @xmath28 with the mc - adapted pdfs , which we do not see with neither lo nor nlo pdfs . in order to examine the origin of this difference we look at the cross section at the parton level , after the hard collision only . since we can not compare to data we choose to compare to cteq5l , which gives a more constant @xmath28 evolution of the cross section .the ratios , @xmath77 , in fig .[ fig : rjet ] show that the relative decrease with lo * and lo * * is clearly visible also at parton level .this is also the case for the three mc1/2/s pdfs in fig [ fig : rjet ] , but to less extent , while the mstw lo results have an almost completely flat ratio . to investigate thisfurther we integrate an approximation of the cross section @xmath78 for only the @xmath60 interactions and a clear pattern very similar to the one at parton level become visible , see fig .[ fig : expdfcross - quark ] . fig .[ fig : expdfcross - all ] shows that this is no longer the case if we include the other possible interactions , i.e. @xmath61 and @xmath79 but at the low @xmath28 end the gluon interactions dominate .turning our attention to the rapidity distribution the results show less variety , see fig.[fig : jetcrap ] .however , excluding mcs , all the mc - adapted pdfs give a narrower distribution .one should not be fooled by the steep slope which hides the differences .mstw lo and nlo are smaller at central rapidities and actually have a lower multiplicity while the rest of the cteq distributions follow cteq5l .even though the pdfs have equal charged particle multiplicity for minbias events this no longer holds true for the hard qcd events .all mc - adapted pdfs except mcs have multiplicity distributions shifted towards lower multiplicity , which is also the case for mstw nlo and to some extent for mstw lo , see fig [ fig : jetcmult ] .including the very latest parton distribution functions into pythia8 both caused some troubles and gave some surprises , especially while venturing outside the grid of the pdfs . at small @xmath41 , where the gluon distribution dominates , there are large differences between the pdfs , especially at low @xmath9 .the mrst / mstw distributions are much larger than the ones from cteq , and mstw lo goes sky high compared to the rest of the pdfs . at larger @xmath9 the differences are smaller between the two collaborations , except for the distributions which freeze their values at the end of the grid , and for mstw lo which is still much larger than the rest .the quarks , and in particular the up distributions , at large @xmath41 show smaller differences , but mstw lo is once again larger at small @xmath41 .some of the new mc - adapted pdfs carry a fraction of the protons momentum larger than unity , in an attempt to compensate for known shortcomings of leading - order calculations . therefore these give a larger activity in the collisions and in order to be able to compare results from the different pdfs in simulations , we first tuned them to obtain equal charged particle multiplicity .the different behavior for mstw lo was also reflected in rapidity distributions both at the level of the @xmath58 sub - process and after hadronization . in general the differences in the pdfs are blurred down to parton level and then blurred even more at hadron level , but some large differences still remain . by comparison to minimum bias data from cdf we saw that the mc adaptedpdfs gave similar results and were in general closer to data than the lo and nlo pdfs , with the exception of cteq6l .differences between the results with different pdfs were enhanced when the energy was increased up to the lhc level .we also discovered that the multiplicity evolved differently with mstw lo and with the two next to leading - order pdfs than with the rest . in the simulations of hard qcd events we saw that the rapidity distributions with the mc - adapted pdfs were narrower , and that their multiplicity distributions were shifted to lower multiplicity .the evolution of the differential jet cross section with @xmath28 was found to be strongly dependent on the choice of pdf , both when it came to its size and shape . in the simulations of hard events we saw that the rapidity distribution was narrower for the mc - adapted pdfs and that their multiplicity distributions shifted to lower multiplicity .the evolution of the cross section with @xmath28 was strongly dependent on the pdf . comparing the @xmath28 spectrum ratio ( simulation over data )the mc - adapted pdfs showed a rather surprising change in ratio with @xmath28 which we examined further .the same behavior could be seen at the level of the @xmath58 sub - process as well and this was traced to the @xmath60 interactions , that dominate in this region .the changes in behavior for the inclusive jet cross section with different pdfs can also be caused by the change of dominating process in the @xmath28 , from @xmath60 to @xmath61 .simulations of prompt photon production also showed such decrease in the cross section for @xmath62 but not for @xmath80 nor @xmath81 interactions .there is a need for better understanding of parton distribution functions at small @xmath41 where the pdfs are now very different from each other . in minbias events@xmath41 values of the order of @xmath82 are the most common .mc simulations need the pdfs to range down to @xmath83 which so far only the three brand new ct09 pdfs do , and we would therefore like to encourage mstw to extend their grid for the mc - adapted pdfs in their next release . at several occasions we were reminded that it can be risky to use nlo pdfs in lo mc generators .our implementation of lo * and lo * * with the new grid causes these two distributions to have a smaller gluon distribution at small @xmath41 , but would otherwise give results more similar mstw lo .in addition we found that there is a need for improved numerical stability at large @xmath41 in order to keep the leading - order pdfs from going negative .this could possibly be solved by using less intricate interpolation routines in this area .possibly one could choose an @xmath84 , different for sea - quarks , gluons and valence quarks , above which one uses the form @xmath85 where @xmath86 and @xmath87 are functions of the virtuality , which would ensure positivity .the large differences in the pdfs get blurred when looking at simulation results , but nonetheless do sometimes cause large variations .changing from the default @xmath0 behavior in pythia8 to @xmath0 value and running determined by the pdfs did not change the results , once the multiplicity had been retuned .interesting to note is that the ct09mcs seems to have some of the features of the other mc - adapted pdfs but in some contexts gives results more similar to ordinary leading - order pdfs .the difference in quark rapidity between mc1/2/s and the rest could be due to their fitting to the nlo pseudo data .the only two other distributions with similar shape are the two nlo pdfs .we could see that a @xmath69-factor for the leading - order pdfs could improve the fit to the inclusive jet data , but for the mc - adapted pdfs the change of the ratio makes it more complicated .finally , the differences in the pdfs have a larger impact when the cm - energy of the collisions increases , and this can cause large uncertainties in simulations at lhc energies . at this point no final answer as to which pdf gives the best results . in order to answer this questionone has to look at a much broader spectrum of observables and also make complete tunes for the different pdfs . in the last yearsthere has been a renewed interest in lo tunes with focus on the applicability in mc generators .the mc - adapted pdfs resulted in some very interesting differences compared to leading - order pdf but there is still room for further improvements . with these new pdfswe have gained a broader spectrum of tools in pythia8 and in examining the origin of differences and similarities between simulations and experiments .first and foremost i want to thank my supervisor torbjrn sjstrand for his help and ability to explain difficult things in an understandable way .i also owe many thanks to leif lnnblad , hendrik hoeth and all the other people at the department of theoretical physics , as well as to marianne ds , sanne kasemets and julia kryszewska .i would also like to mention johan bijnens , joakim cederkll , bo sderberg and hans - uno bengtsson who all played an important role in my previous education .

<S> a selection of the latest and most frequently used pdfs is incorporated in pythia8 , including the mc - adapted pdfs from the mstw and cteq collaborations . </S><S> this thesis examines the differences in pdfs as well as the effect they have on results of simulations . </S><S> the results are also compared to data collected by the cdf experiment . </S><S> lu tp 10 - 03 + mcnet/10/03 + february 15 , 2010 * inclusion of + parton distribution functions in pythia8 * + tomas kasemets + master thesis in theoretical high energy physics + department of astronomy and theoretical physics , lund university + slvegatan 14a , se 223 62 lund , sweden + thesis advisor : torbjrn sjstrand </S>

1302.2161

to describe the energy @xmath8 of quasi - particles near any specific point on the fermi surface it is convenient to change coordinate systems to three mutually orthogonal @xmath68-vector components : one perpendicular to the fermi surface , @xmath69 , and two parallel to it @xmath70 , @xmath71 .we then have @xmath72 where @xmath10 is the fermi velocity , @xmath73 when @xmath74 ( the fermi vector ) , and @xmath75 is a function determined by the superconducting gap function , which may be either singlet , triplet or a combination ( for non - centrosummetric systems ) .the gap energy spectra shown in fig . 1 ( main text ) are described by an equation for the gap energy @xmath75 which has the form @xmath76 near to a gap nodal point , where @xmath77 is a positive tensor of rank 2 which is determined by the symmetry of the gap function around that nodal point . in general @xmath77will have two principal axes , which we can choose as @xmath70 and @xmath71 .if the two eigenvalues corresponding to these axes , @xmath78 and @xmath79 are both positive and non - zero we have an ordinary point node with linear spectrum illustrated in fig . 1 ( a ) ( main text )( with @xmath80 ) .if one eigenvalue is positive and non - zero ( @xmath81 ) and the other is zero we have an ordinary line node . in the case of two zero eigenvaluesit is necessary to continue the expansion to quartic or higher order , giving the various shallow nodes or crossings illustrated in fig . 1 ( main text ) .in principle even higher powers may be possible , but we have not found any realistic examples of gap models where this could occur .in contrast , all the node types presented in fig .1 ( main text ) can be realised in the specific gap models presented in this paper .computation of the density of states integral @xmath82 for each of these gap energy spectra gives a unique expression .the specific heat capacity in the superconducting state is in turn given by @xcite @xmath83 e_{\bf k}\operatorname{sech}^{2}\left(\frac{\beta e_{\bf k}}{2}\right ) \label{eq : cv_from_leggett}\ ] ] where @xmath84 .the low @xmath12 specific heat capacity can therefore be found analytically from @xmath85 by evaluating the integral @xmath86 giving the power law dependence of a low @xmath12 specific heat capacity measurement .we find the terms listed in fig . 1 ( main text ) for point and line nodes with linear and non - linear excitation spectra , where @xmath87 is the length of the line node from end to end .for cases where line nodes cross we expect the terms listed in fig .1(main text ) , where @xmath87 is now the length of the line node on the fermi surface measured from the centre to the corner of the crossing whose contribution is being calculated . the cases in figs .1 e ) and f ) ( main text ) are not power laws , however they approximate very closely to the non - integer power laws indicated , with very slight variations depending on the constant @xmath87 . in the case of line reconnection transitions such as fig .4 , plots 3 , 8 , and 14 ( main text ) , where the fermi surface includes regions where the gap energy spectrum expands as @xmath88 , the contribution of the line crossing to the density of states is @xmath89 .the low-@xmath12 specific heat capacity can also be calculated numerically using eq .( [ eq : cv_from_leggett ] ) by summing over a sufficient number of integration points corresponding to different values of the gap in k - space near the fermi surface for selected temperature ranges below t@xmath90 .the exponent of temperature , @xmath91 , can then be obtained using the general formula @xmath92 . when this quantity approaches a constant we have @xmath35 .this yields the curves in figs . 2 and 4 ( c ) ( main text ) , where the temperature dependence of the gap is taken to be given by the expression from ginzburg - landau theory , @xmath93 .this approximation , made for convenience , introduces a weak additional temperature - dependence for @xmath94 compared to the bcs approximation and therefore provides an upper bound on the temperature dependence introduced by the gap .we find that the exponents are not affected by this : even in the case where the temperature dependence of the gap is assumed to be constant , the same power laws are obtained with flat temperature dependence .each contour in fig .4 ( a)(main text ) corresponds to a line where the gap energy @xmath75 derived from the gap function of eqs .( 2,3)(main text ) vanishes .this gap energy is given by @xmath95 , where @xmath96 .the phase diagram in fig .4 ( b ) ( main text ) shows the values of the parameters @xmath51 and @xmath52 where the nodal topology changes for this form of the gap energy .this form of @xmath75 has also been used to compute the specific heat exponent in fig .4 ( c ) ( main text ) .here we give the details of the derivation of the nodal topologies and phase diagram presented in fig . 4 ( b ) of the main text .we use a single band mean - field model with antisymmetric spin - orbit coupling to obtain the resulting gap node topologies on the fermi surface , which we take to be spherical for simplicity .the bogoliubov - de gennes hamiltionian is the noncentrosymmetric ( ncs ) gap function which mixes singlet and triplet as a result of the parity violation in the crystal lattice structure is given in eq .( 1 ) of the main text , where @xmath98 denotes the pauli matrices and @xmath46 and @xmath99 denote the singlet and triplet components of the gap respectively . the normal state hamiltonian @xmath100 given in @xcite for non - interacting electrons in a crystal without inversion center is @xmath101 where @xmath102 is the non - relativsitic metallic dispersion energy .the second term represents the antisymmetric soc where the coupling constant @xmath103. we will assume the @xmath104-vector @xmath99 to be parallel to @xmath105 .this ensures that soc is not destructive to the triplet component of the ncs gap function @xcite . in a general ginzburg - landau theory the possible symmetries of the superconducting instabilitydepend on the symmetry of the normal state . in this casethe symmetry of the normal state is given by @xmath106 where @xmath55 denotes the ncs cubic space group , @xmath12 the time reversal symmetry and @xmath107 the gauge symmetry operation .in any superconducting transition the @xmath107 symmetry is broken however other symmetries of the normal state may be broken as well .for the purposes of our analysis we will , following @xcite , look at the case where only the gauge symmetry and no further symmetries of the crystal lattice are broken i.e. the superconducting instability has the symmetry of the @xmath108 irrep of the @xmath55 group .this means that the gap function , given in general by equation ( 1 ) of the main text , takes the specific form of eqs .( 2,3 ) of the main text . with @xmath99 and@xmath105 explicitly defined the bogoliubov de - gennes ( bdg ) hamiltonian can be diagonalized .[ sec : a_1 ] the single band bdg hamiltonian with soc defined parallel to the @xmath109 component of the gap @xcite has the property that splitting of the fermi surface occurs as a result of soc whereas zero energy excitations occur as a result of the combination of singlet and triplet components of the gap function . in this casethe diagonalized bdg hamiltonian has eigenvalues of the form @xmath110 where the @xmath111 signs inside the square root always take the same value while the sign in the front varie independently .the soc term @xmath105 in this expression has the effect of splitting fermi sheets whereas the second term @xmath112 determines the node topology of the superconducting gap on either of the rashba - split fermi sheets .the zeros of the term @xmath113 are the locii of zero energy excitations on the fermi surface . in order to ensure that the soc parameter does not significantly warp the spherical fermi surface and therefore change the node topology we assume that @xmath114 .this hierarchy of energy scales is reasonable to assume as @xmath115 is of the order of ev , @xmath39 is of the order of mev , and observed soc - induced fermi surface splittings are small relativistic corrections to @xmath115 .the conditions for zeros in the gap energy spectrum for each rashba - split fermi sheet are @xmath116 and @xmath117 .thus , one of the fermi surface sheets is always fully - gapped , while the other one may be nodal .the gap function in this latter case is best normalized with respect to the singlet gap energy : @xmath118 the term @xmath119 has here been replaced with the term @xmath120 whose coefficients are scaled by a factor of @xmath121 .this expression can be fully expanded and used to derive the topological phase boundaries by substituting in the vector @xmath122 $ ] . a coordinate transformation to spherical polar coordinates using @xmath123 , @xmath124 and @xmath125 allows us to select specific points on the fermi surface and find solutions in a and b. the phase boundaries of fig .4 ( b ) of the main text separate regions of constant gap topology which were obtained for critical values of the polar and azimuthal angles corresponding to the points where nodes first appear on the gap or where line nodes reconnect at a topological transition .phases 1 - 5 in fig .4 ( a ) of the main text correspond to those considered by schnyder et al .the coordinates @xmath126 correspond to one topologically distinct set of point nodes on the spherical fermi surface shown in fig .4 ( a ) , plot 5 of the main article .substituting these into the equation @xmath127 gives the relationship between coefficients ` a ' and ` b ' : @xmath128 .the line reconnection shown in fig . 4 ( a ), plot 3 of the main article corresponds to the coordinates @xmath129 and produces solutions @xmath130 .the final phase boundary in fig . 4 ( b ) corresponds to point nodes at the coordinates @xmath131 shown in fig . 4 ( a ) , plot 1 , which produces the solutions @xmath132 ._ , _ phys .b _ * 53 * , 15147 ( 1996 ) ; yu.s . barash and a.a .svidzinsky , _ phys .b _ * 58 * , 6476 - 6492 ( 1998 ) ; v.p .mineev k.v .samokhin , `` introduction to unconventional superconductivity '' ( gordon and breach science publishers , amsterdam , 1999 ) ; h. shakeripour , c. petrovic , louis taillefer , _ new j. phys . _ * 11 * , 055065 ( 2009 ) .

<S> unconventional superconductors are most frequently identified by the observation of power - law behaviour on low - temperature thermodynamic or transport properties , such as specific heat . here </S><S> we show that , in addition to the usual point and line nodes , a much wider class of different nodal types can occur . </S><S> these new types of nodes typically occur when there are transitions between different types of gap node topology , for example when point or line nodes first appear as a function of some physical parameter . </S><S> we identify anomalous , non - integer thermodynamic power laws associated with these new nodal types , and give physical examples of superconductors in which they might be observed experimentally , including the noncentrosymmetric superconductor li@xmath0pd@xmath1pt@xmath2b . </S><S> a defining feature of many unconventional fermi superfluids and superconductors is the existence of lines or points on the fermi surface where the energy gap vanishes and so - called `` nodal quasi - particles '' can exist at arbitrarily low energies . in fermi systems with such nodal quasi - particles </S><S> the low temperature specific heat will show particular power law behaviours as a function of temperature . </S><S> the expected point and line node power law dependencies were first derived in relation to the proposed low temperature superfluid phases of liquid @xmath3he @xcite . </S><S> these were subsequently clarified further @xcite and are now widely used to identify pairing states in unconventional superconductors @xcite . here </S><S> we show that gap nodes in superconductors can occur in a number of more general types than simply the usual line or point zeros and that each of these has a corresponding thermodynamic signature , typically in the form of non - integer power laws in low temperature specific heat . </S><S> we predict that these anomalous power laws generically occur at points in the phase diagram where there is a topological change in the line or point nodal structure on the fermi surface , which we illustrate with a specific example : the noncentrosymmetric superconductors li@xmath0pd@xmath1pt@xmath2b . in this case </S><S> the gap node topological changes are also associated with changes of bulk topological quantum numbers for the quasi - particles . </S><S> the future experimental observation of such non - integer power laws could therefore be used to identify not only superconductors with highly unconventional pairing symmetries , but also topological superconducting and superfluid systems @xcite . </S><S> similar physics are realised in the high - temperature cobalt doped pnictide materials ba(fe@xmath4t@xmath2)@xmath0as@xmath0 ( t = co , ni , pd ) @xcite . for the usual types of point and line </S><S> nodes the quasi - particle energy spectrum is linear in the vicinity of the gap node . </S><S> a familiar example is the case of the @xmath3he a - phase with triplet pairing order parameter @xmath5 , which has point nodes at the points @xmath6 on the fermi sphere , of radius @xmath7 . near to these points the quasi - particle energy spectrum , </S><S> @xmath8 , obeys @xmath9 , where @xmath10 is the fermi velocity , giving a linear dispersion relation as shown in fig . </S><S> [ fig : points_lines_nodal_crossings ] a ) . </S><S> for this type of gap node we expect a specific heat , @xmath11 at temperatures @xmath12 much lower than the critical temperature @xmath13 . </S><S> the case of line nodes in the superconducting gap is also well known , as for example found in d - wave superconductors such as yba@xmath0cu@xmath14o@xmath15 . assuming in this case that the fermi surface is an open cylinder of radius @xmath7 , and the energy gap is of the form @xmath16 , then there are four line nodes spaced around the fermi surface at @xmath17 . </S><S> again the energy spectrum is linear near these gap nodes , as shown in fig . [ </S><S> fig : points_lines_nodal_crossings ] c ) , so in this case we expect @xmath18 @xcite . [ cols="^,^ " , ] while the generic point and line nodes of this type are expected in the majority of unconventional superconductors , other types of nodal energy spectrum are possible . for example </S><S> it is possible that the system has a point node , but that the energy spectrum is quadratic not linear near the nodal point see fig . </S><S> [ fig : points_lines_nodal_crossings ] b ) . </S><S> the heat capacity at this critical point is a power law , with @xmath19 rather than the usual @xmath11 for point nodes @xcite . </S><S> this aspect of non - linear excitation spectra was identified by leggett in his review of the superfluid phases of @xmath20he @xcite , though at the time no examples were known . </S><S> later a ground state with e@xmath21 symmetry having this feature was proposed for the heavy - fermion superconductor upt@xmath14 @xcite and possible thermal transport signatures were discussed @xcite . here </S><S> we show that this different type of point node behaviour is not restricted to certain special symmetries but occurs generically when the gap structure is changing from having a point node to having no nodes , as a function of some parameter defining the pairing potential on the fermi surface . </S><S> we extend this notion to a much wider class of nodal types and node topology transitions . </S><S> the fact that it is possible to continuously remove a point node of the gap is surprising , since the usual linear spectrum point node behaviour is usually thought to be topologically stable against perturbations @xcite . </S><S> the shallow node behaviour is not topologically protected , since it only occurs for specific parameter values in which accidental cancellations occur between different order parameter components . nevertheless , as we shall see its influence extends throughout the finite - temperature phase diagram . </S><S> if we generalize the shallow point node behaviour to line nodes we obtain a spectrum near to the node shown in fig . </S><S> [ fig : points_lines_nodal_crossings ] d ) , which we will refer to as a _ shallow line node_. in this case we find an anomalous non - integer power law for the low temperature specific heat , of the form @xmath22 @xmath23 @xmath24 . similarly to the shallow point node </S><S> , this type of shallow line node only occurs at a boundary between a nodeless superconductor and one with ordinary line nodes , as discussed recenly for the pnictide superconductors @xcite . </S><S> in particular , both nodal and nodeless compounds of the system ba(fe@xmath4co@xmath25)@xmath26as@xmath26 have been reported @xcite , suggesting that the @xmath24 power law could be observed experimentally . </S><S> this would provide compelling evidence for the @xmath27 gap model @xcite . </S><S> another type of nodal behaviour occurs when there is a crossing or topological reconnection of line nodes . </S><S> these again lead to distinctive low temperature power laws , with non - integer power law exponents . </S><S> a crossing of linear line nodes is illustrated in fig . </S><S> [ fig : points_lines_nodal_crossings ] ( e ) . </S><S> this type of crossing can occur in a d - wave superconductor , for example @xmath28 , in the case where the fermi surface is spherical rather than cylindrical . </S><S> although the node crossing point occurs for a small portion of the fermi surface , it turns out that this point dominates the low temperature heat capacity , and such crossings can therefore be detected experimentally . </S><S> this is because the density of states resulting from the node crossing has a logarithmic term . </S><S> this logarithmic term approximates a power law at low temperatures , and leads to a specific heat capacity contribution approximately of the form @xmath29 , which at low enough temperatures dominates over the usual line node contribution @xmath30 arising from the line nodes away from the crossing . unlike the shallow point node </S><S> , the crossing of line nodes is expected to be topologically protected against small perturbations . the final distinct type of gap node which we shall discuss is a crossing of shallow line nodes , as shown in fig . </S><S> [ fig : points_lines_nodal_crossings ] ( f ) . </S><S> again this leads to a logarithmic term in the density of states , which dominates the low temperature behaviour compared to the shallow line node case in the absence of crossings . </S><S> the corresponding specific heat capacity is approximately @xmath31 for the shallow line node crossing , compared to @xmath32 for the case without crossings . </S><S> interestingly , @xmath33 is not so far from the value obtained for the ungapped fermi surface of a normal ( non - superconducting ) fermi liquid , @xmath34 . </S><S> fig . </S><S> [ fig : exponent_plot ] shows the predicted exponents of the power law temperature dependences of specific heat for each of the different nodal types shown in fig . </S><S> [ fig : points_lines_nodal_crossings ] . in all cases </S><S> the predicted power law behaviour @xmath35 is accurate at very low temperatures , below @xmath36 , and in many cases the low temperature power laws apply over a range of temperatures up to @xmath37 . </S><S> the details of our calculations of specific heat leading onto the results in figs . </S><S> [ fig : points_lines_nodal_crossings ] and [ fig : exponent_plot ] are given in appendix [ sec : methods ] . </S><S> non - integer power laws , especially powers below @xmath38 , are a highly distinctive signature of the low energy excitation spectrum , and their experimental observation would be a relatively clear and direct form of evidence for the existence of the corresponding type of gap node in a given material . in systems where the gap topology is changing as a function of some experimental parameter ( e.g. doping or spin - orbit coupling ) these anomalous powers will only be observed to the lowest temperatures if the material is exactly at the critical parameter value for the shallow node to occur . however , </S><S> if the material is merely near to , not exactly at , the critical parameter value , then the shallow node power laws will hold over a wide temperature range , reverting to ordinary line , point , or nodeless behaviour at the very lowest temperatures . again </S><S> experimental observation of such a cross - over between power laws would be a clear signature of the unusual gap nodal structure - we provide an explicit example of this below . in practice , it is unlikely that the power laws shown in fig . </S><S> [ fig : exponent_plot ] can be distinguished from one - another in a single measurement as a function of temperature , however a transition of the type shown in fig . [ </S><S> fig : topology_plot_c ] ( inset ) , discussed below , is more feasible as a function of doping at constant temperature . </S><S> in addition to the _ ad hoc _ mechanisms mentioned above , we expect shallow line or point nodes to occur generically in superconductors with multi - component order parameters . </S><S> consider the simple model shown in fig . [ fig : multi_param ] . </S><S> this shows a hypothetical system , such as an ordinary s - wave bcs superconductor with a significant level of anisotropy in the gap . </S><S> here @xmath39 is the average gap and @xmath40 is a measure of the gap variation with @xmath41 parallel to the fermi surface . in the case </S><S> where @xmath42 the gap is nodeless everywhere on the fermi surface . on the other hand , if @xmath43 then the gap changes sign at some points on the fermi surface , leading to pairs of line nodes . at the critical parameter value @xmath44 the line node pairs coalesce , leading to a single , shallow , line node . in this example </S><S> no symmetry change occurs in the gap function , </S><S> however `` accidental '' line nodes occur within a single symmetry of the cooper pair order parameter , because of the large degree of gap anisotropy . </S><S> an explicit example of this mechanism by which anomalous power laws may be observed is provided by non - centrosymmetric superconductors where the superconductor order parameter @xmath45 has singlet and triplet components @xmath46 and @xmath47 respectively , such as the compounds li@xmath26pd@xmath1pt@xmath25b @xcite . </S><S> platinum doping in this system is thought to increase the spin - orbit coupling , and hence increase the triplet component of the order parameter relative to the singlet component @xcite . in doing </S><S> so the spin - orbit coupling introduces nodes in the gap . within a standard symmetry classification of the possible order parameter structures </S><S> the simplest one takes the form @xcite </S><S> @xmath48 - \right . </S><S> \label{lipdptb - gap-2 } \\ & & b \left . </S><S> \left [ x\left(y^2+z^2\right ) , y\left(x^2+z^2\right ) , z\left(y^2+x^2\right ) \right ] \right\ } \nonumber \end{aligned}\ ] ] where @xmath49 are functions with the same symmetry properties as the components of the electron wave vector @xmath50 and @xmath51 and @xmath52 are material - specific parameters that depend on doping @xmath53 . eqs . </S><S> ( [ lipdptb - gap-1 ] ) and ( [ lipdptb - gap-2 ] ) correspond to the @xmath54 irreducible representation of the cubic @xmath55 group , which is the only one where only gauge symmetry is broken . </S><S> other pairing states break additional symmetries and therefore are not compatible with a conventional ( phonon - mediated ) pairing mechanism . in spite of this , </S><S> a large number of different gap nodal structures are obtained , as shown in fig . </S><S> [ fig : topology_plot_a_b ] ( top panel ) . as a function of the parameters </S><S> @xmath56 we obtain the rich phase diagram of fig . </S><S> [ fig : topology_plot_a_b ] ( bottom panel ) . </S><S> the corresponding specific heat exponents are illustrated in fig . </S><S> [ fig : topology_plot_c ] . </S><S> upon doping , the material explores the phase diagram along some unspecified path from small values of @xmath51 and @xmath52 ( @xmath57 ) to larger values ( @xmath58 ) @xcite . </S><S> small values of @xmath51 and @xmath52 correspond to a fully gapped , singlet - dominated superconducting order parameter as believed to be realised in li@xmath26pd@xmath59b . </S><S> order parameters with a large triplet component and line nodes on the fermi surface are thought to occur in li@xmath26pt@xmath59b and correspond to the case when either @xmath51 or @xmath52 ( or both ) are larger . </S><S> ( color online ) the transition between a nodal and non - nodal gap structure in any superconductor as a result of continuously admixed order parameters will be characterized by a non - linear gap energy spectrum close to the node . </S><S> this is because in any wave - like order parameter the deep minima have no linear order term in the expansion . ] </S><S> inspection of the phase diagram shows that at some values of @xmath51 and @xmath52 we expect point nodes to appear , labelled by ` 1 ' in fig . </S><S> [ fig : topology_plot_a_b ] , before line nodes appear on the gap , labelled by ` 2 ' in the same figure . </S><S> these point nodes are shallow point nodes , and so instead of the usual @xmath60 power law dependence of the heat capacity for point nodes , we instead expect a @xmath61 power law dependence . upon further change of the @xmath51 and @xmath52 parameters </S><S> these shallow point nodes expand into rings of ordinary line nodes , as shown in fig . </S><S> [ fig : topology_plot_a_b ] , plot labelled ` 2 ' . increasing the parameters yet again there is a level which causes a topological line reconnection , labelled ` 3 ' , where the different line nodes intersect before becoming rings again , now centred at different points on the fermi surface , ` 4 ' . </S><S> note that the number of line nodes on the fermi surface is always a multiple of two as predicted by the topological theory of gapless phases in time - reversal - invariant superconductors @xcite . </S><S> that number jumps at the line reconnection transition , indicating its topologically non - trivial nature . </S><S> the line reconnection transition has a distinct thermal signature in the low @xmath12 heat capacity originating from the crossing of line nodes associated with such a line reconnection . as is evident from the figure , </S><S> there are many other paths along the phase diagram with similar phase transitions . in our analysis of the cubic non - centrosymmetric superconductors we also found a single instance of crossing of shallow line nodes , labelled ` 11 ' in fig . </S><S> [ fig : topology_plot_a_b ] . </S><S> these are realised when line nodes are spontaneously introduced on a nodeless section fermi surface without first appearing as points . </S><S> the specific heat exponent of @xmath62 for this case is clearly seen in fig . </S><S> [ fig : topology_plot_c ] . </S><S> schnyder _ et al . </S><S> _ @xcite have shown that a number of these phases of cubic noncentrosymmetric superconductors possess non - trivial topological quantum numbers @xcite , and have corresponding surface edge states of topological nature . in principle , these surface states can be probed experimentally by electron tunnelling through an interface with a normal metal @xcite . however , we propose that a more direct , bulk , measurement of this non - trivial gap topology can be obtained by measurements of the low temperature heat capacity . if the anomalous power laws we have predicted can be observed , then it should be possible to detect the phase boundaries between different types of topological gap structures , and hence determine in a bulk measurement the topological quantum numbers associated with different doping regimes . </S><S> phases ` 1'-`5 ' in fig . </S><S> [ fig : topology_plot_a_b ] correspond to those considered by schnyder _ </S><S> et al . </S><S> _ @xcite . </S><S> ( color online ) ( top ) fifteen gap topologies corresponding to different triplet admixtures in a cubic non - centrsymmetric superconductor . within this set the only point nodes are shallow point nodes , occuring in plots 1,5 10 and 15 . </S><S> topologies 3,6,8 and 14 are all line reconnection transitions between distinct topologies and include crossings of linear line nodes . </S><S> topology 11 is a crossing of shallow line nodes . </S><S> ( bottom ) phase diagram of gap node topology corresponding to triplet admixture parameters @xmath51 and @xmath52 in eq . </S><S> ( [ lipdptb - gap-2 ] ) on a spherical fermi surface . </S><S> each line represents a phase boundary where gap nodes appear or disappear on the fermi surface . </S><S> , title="fig : " ] + ( color online ) ( top ) fifteen gap topologies corresponding to different triplet admixtures in a cubic non - centrsymmetric superconductor . within this set the only point nodes are shallow point nodes , occuring in plots 1,5 10 and 15 . </S><S> topologies 3,6,8 and 14 are all line reconnection transitions between distinct topologies and include crossings of linear line nodes . </S><S> topology 11 is a crossing of shallow line nodes . </S><S> ( bottom ) phase diagram of gap node topology corresponding to triplet admixture parameters @xmath51 and @xmath52 in eq . </S><S> ( [ lipdptb - gap-2 ] ) on a spherical fermi surface . </S><S> each line represents a phase boundary where gap nodes appear or disappear on the fermi surface . </S><S> , title="fig : " ] ( color online ) ( top panel ) specific heat exponent as a function of temperature for different values of the @xmath51 and @xmath52 parameters , as indicated . </S><S> the inset shows the value of the specific heat as a function of @xmath52 for fixed @xmath63 and constant temperature , @xmath64 . </S><S> ( bottom panel ) specific heat exponent as a function of temperature and parameter @xmath52 for fixed @xmath63 . </S><S> details of the calculation of the quasiparticle spectrum and phase diagram are given in appendix [ sec : methods].,title="fig : " ] ( color online ) ( top panel ) specific heat exponent as a function of temperature for different values of the @xmath51 and @xmath52 parameters , as indicated . </S><S> the inset shows the value of the specific heat as a function of @xmath52 for fixed @xmath63 and constant temperature , @xmath64 . </S><S> ( bottom panel ) specific heat exponent as a function of temperature and parameter @xmath52 for fixed @xmath63 . </S><S> details of the calculation of the quasiparticle spectrum and phase diagram are given in appendix [ sec : methods].,title="fig : " ] the anomalous power - laws due to changes of nodal topology are realised exactly at each transition . on the other hand , by being close to , but not exactly at one of those transitions one expects a crossover from conventional ( fully gapped or normal power - law ) behaviour to anomalous behaviour as the temperature is increased through a characteristic temperature scale @xmath65 . </S><S> a demonstration of this is provided for the case of li@xmath26pd@xmath1pt@xmath25b see fig . </S><S> [ fig : topology_plot_c ] , specifically the curves corresponding to @xmath66 ( top panel ) and the density plot ( bottom panel ) showing how the temperature scale @xmath65 converges to zero at the transition . </S><S> this may be used to detect topological transitions in superconductors and in fact shows that the influence of the topological transition extends over a wide region in the phase diagram , not just at a single point . </S><S> the detection of the phase boundary itself is facilitated by its higher specific heat , at fixed low temperature , due to the enhanced phase space available to quasi - particles in that state . </S><S> this is illustrated in the inset to fig . </S><S> [ fig : topology_plot_c ] . </S><S> we emphasize , however , that it is in the anomalous power - laws that we find the evidence of the special nature of quasi - particles at the phase boundary . in recent years </S><S> it has become widely recognised that point or line nodes in the bulk spectrum of topological matter may evolve into topolgically - protected surface or vortex core states with zero dispersion @xcite . </S><S> in contrast , the shallow dispersions discussed here exist in the _ bulk _ excitation spectrum near topological transitions . [ </S><S> sec : conc ] to conclude , we have shown that a more general classification of nodal characters of superconductors yields anomalous , non - integer power laws for low - temperature thermodynamic quantities . </S><S> some such power laws can be used to detect line node crossings ; others are expected at phase transitions where the nodal topology reconfigures itself . </S><S> an example is provided by the topological transitions expected to occur as a function of doping in the non - centrosymmetric cubic superconductor li@xmath26pd@xmath1pt@xmath25b . </S><S> it is important to stress that the observation of anomalous low - temperature power laws , just as the ordinary , integer ones , requires going to temperatures well below @xmath67 . on the other hand , </S><S> it is not necessary to sit exactly at the transition , as there is a crossover temperature separating conventional from anomalous power - law behaviour that converges to zero there . in that sense </S><S> the influence of the topological transition extends throughout the bulk phase diagram in a manner similar to that of a quantum critical endpoint @xcite . </S><S> this works was supported by epsrc and stfc ( u.k . ) . </S><S> j.q . gratefully acknowledges funding from hefce and stfc through the south - east physics network ( sepnet ) . </S>

nucl-ex9809005

in central heavy ion collisions the number of mesons produced increases with bombarding energy . in the energy range between @xmath10 and @xmath11 agev , 8% to 22% of the available energyis dissipated into this channel which predominantly consists of pions @xcite .the amount of dissipated energy also depends on system mass : in central au + au reactions at @xmath10 agev it is 8% , but 15% in central ni + ni reactions at the same bombarding energy .the process responsible for meson production is believed to be predominantly the excitation of baryon resonances during the early compression phase of the collision . in the later expansion phase these resonances decay .the primordial decay yields a number of different mesons , mainly pions .mesons with masses higher than the pion mass , e.g. @xmath12 , may subsequently decay also into several pions . in total , the average mass of the excited baryon resonances and the number of pions produced by their decay chains increase with bombarding energy .this mechanism is the basic process of pion production which is used in nuclear transport models to describe the dynamics of relativistic heavy ion collisions @xcite . in a recent publication@xcite it was shown that all resonances up to the @xmath13 resonance have to be included to successfully reproduce the experimental pion data at energies between 1 and 2 agev .the experimental confirmation that baryon resonances are excited in relativistic heavy ion collisions is rather scarce .the earliest indication came from the comparison by brockmann et al .@xcite of the pion and proton energy spectra observed in @xmath14 agev ar @xmath15 kcl reactions .this comparison was later extended with the attempt to determine the @xmath4 resonance mass and temperature @xcite .similar analyses of the pion energy spectra measured in @xmath10 agev au + au collisions were published in @xcite . at much higher energies , i.e. @xmath16 agev , the fraction of nucleons excited to the @xmath4 resonance was determined for si @xmath15 al , pb reactions @xcite . in this study also the invariant mass distribution of the @xmath4 resonance was reconstructed from correlated @xmath7 pairs .to identify structures in the invariant mass distribution of correlated proton and pion pairs provides the direct proof that nucleons are excited to high - lying resonances .the major obstacle in reconstructing the invariant mass is the large background of non - correlated @xmath17 pairs , which increases with the number of protons and is therefore particularly large for central reactions of heavy - mass systems . in peripheral reactions with very light projectiles ,e.g. p @xcite or @xmath18he @xcite induced reactions at around @xmath11 gev bombarding energy , the @xmath7 correlations were successfully analyzed and the mass distribution of the @xmath4 resonance was determined . the resonance mass was found to be shifted by @xmath19 mev / c@xmath3 to lower masses in reactions on various targets , compared to those on protons .very recently the @xmath4 was reconstructed from @xmath7 pairs detected at @xmath20 agev bombarding energy in @xmath21ni + cu collisions @xcite . in this casemuch larger mass shifts of @xmath22 mev / c@xmath3 were observed in central collisions , the shift became smaller with increasing impact parameter . that the @xmath1 pairs can also be successfully studied in central collisions of very heavy systems at @xmath10 agev bombarding energy was shown in @xcite .this report contained a preliminary analysis of the data used in the present publication .although at @xmath10 agev energy the excitation of the @xmath4 resonance is the dominating channel there is direct experimental evidence from the identification of the @xmath23 meson @xcite , that at this energy also higher - lying resonances , in particular the @xmath24 resonance , must become excited .if this is the case then the resonance decay into the @xmath25 channel also has to be present , and the analysis of @xmath1 correlations will not yield a complete picture of the resonance distribution . on the other hand , the @xmath25 decay channel can be observed in the pion transverse momentum spectra @xcite .the subsequent sections are organized as follows : the section 2 contains a short description of the experimental procedures. a more detailed description can be found in @xcite .the methods to extract the invariant mass distributions of the resonances involved are presented in section 3 .these methods include the analyses of the measured @xmath26 and @xmath27 transverse momentum spectra , and the analyses of the @xmath6 and @xmath7 correlations . in both cases the residual coulomb interaction between the charged pions and the positively charged baryon distribution causes a perturbation @xcite , which is of opposite sign for @xmath26 and @xmath27 and which we compensate by analyzing the average of the @xmath26 and @xmath27 spectra , respectively by taking the average invariant mass distribution from @xmath6 and @xmath7 . the justification to apply this procedure is given in the appendix 6.1 .the experimental results are presented in section 4 , section 5 contains their discussion and the summary .the reactions @xmath28au + @xmath28au and @xmath21ni + @xmath21ni were studied at nominal beam energies of @xmath29 , respectively @xmath30 and @xmath31 agev .the @xmath28au and @xmath21ni beams were accelerated by the @xmath32 accelerator combination of the @xmath33 /darmstadt .the duty cycle was 75% with a spill length of @xmath34 s. the average beam intensities varied between @xmath35 and @xmath36 particles per spill .the targets consisted of self supporting foils with a thickness of @xmath37m(au ) , respectively @xmath38m(ni ) .this corresponds to an interaction probability of 0.5% .the energy loss of the beams in the targets is of around @xmath39 agev and was neglected .the particles produced by the au + au and ni + ni reactions were detected by the @xmath40 detector which is a modular detection system with almost @xmath41 coverage .of particular importance for the present investigation is the central drift chamber @xmath42 which is mounted inside the superconducting magnet with a solenoidal field of @xmath43 t strength .the @xmath42 has complete cylindrical symmetry , it covers the polar angles @xmath44 from @xmath45 to @xmath46 in the laboratory frame and allows the particles masses and charges to be simultaneously determined . for charged pionsthe geometrical boundaries of the @xmath42 guarantees a detection efficiency of better than 65% , in particular the midrapidity region @xmath47 is fully covered above a transverse momentum limit @xmath48 ( for the definition of the variables used here , see below ) .notice , however , that to distinguish the @xmath27 from protons an additional momentum limit @xmath49 gev / c was introduced @xcite .this limit which corresponds to @xmath50 at midrapidity , has no consequences for the situation shown in fig.1 , and it is not required for the @xmath26 . for protons the geometrical @xmath42 boundaries@xmath51 reduce their detection efficiency to approximately 45% .the situation for particles produced in symmetric reactions at @xmath31 agev bombarding energy is depicted in fig.1 .the fully drawn curves display the lower and upper geometrical limits @xmath52 and @xmath53 in the laboratory frame , the dotted curves are drawn for pions respectively protons with @xmath54 kinetic energies of @xmath55 mev .the shaded area displays the momentum distribution of identified @xmath4 resonances , i.e. the decay proton and pion were measured by the @xmath42 .the resonances are monte carlo generated using a fixed mass @xmath56 mev / c@xmath3 and an isotropic boltzmann distribution with temperature @xmath57 mev , which corresponds to a mean kinetic energy of @xmath55 mev . because of the decay into a correlated @xmath17 pair the @xmath4 resonance can be observed in regions of the momentum space which are prohibited for single pions or protons by the @xmath52 cut . at midrapiditythe @xmath4 resonance is detected for transverse momenta @xmath58 .notice that the symmetry law @xmath59 which holds for all symmetric reactions guarantees that the region @xmath60 can be reconstructed by reflection around midrapidity from the region @xmath61 . in this way the properties of around 50% afall @xmath4 resonances are known .the remaining ones which can not be reconstructed decay into a @xmath17 pair , where either one of the two particles or both are not detected within the geometrical boundaries of the @xmath42 . under the assumptions made ,i.e. isotropic resonance distribution and isotropic resonance decay , the efficiency of the @xmath42 for detecting resonances can be calculated as a function of the mass and the kinetic energy of the resonance .= 8.5 cm the efficiencies to detect resonances with masses @xmath62 and kinetic energies @xmath63 are plotted in the right panel of fig.1 , without employing symmetrization .the detection efficiencies decrease rapidly with mass for small kinetic energies . without the proper efficiency corrections , which in this case are solely due to the incomplete @xmath41 acceptance of the detector ,the contributions from small mass values @xmath62 would be reduced .notice that from the results published in @xcite for the au + au reaction at @xmath10 agev bombarding energy , the expected mean kinetic energy of the @xmath4 resonance with mass @xmath64 mev / c@xmath3 is @xmath65 mev . to obtain the efficiency correction for a given mass the data shown in fig.1 have to be weighted by the energy distribution of that mass and integrated .in addition to the @xmath42 , the information from the forward plastic scintillation wall @xmath66 mainly served trigger purposes . the @xmath66 is a high - granularity component of the @xmath40 detector and covers the polar angles @xmath44 from @xmath67 to @xmath68 with full cylindrical symmetry .it separates particles only according to their charges . in the present analysisthe particle multiplicities @xmath69 measured with the @xmath66 were required to be larger than 80% of the maximum multiplicity which is defined by the lower boundary of the @xmath70 multiplicity bin , see @xcite .this multiplicity criterion selects near - central collisions in the impact parameter range from @xmath71 to @xmath72 , where @xmath73 is the maximum impact parameter of a given reaction . on the averagethe fireball then contains around @xmath74 of the maximum number @xmath75 of nucleons . in the present study we shall use the identical conventions which were used in @xcite .all quantities which refer to the laboratory(@xmath76 ) or target frames will be labeled by small letters .the center of mass(@xmath54 ) or fireball frames are generally characterized by capital letters .exceptions are those quantities , like the transverse momentum , which remain unchanged under the transformation of frames . for the transverse momentum we sometimes use the quantity @xmath77 where the index @xmath78 indicates a normalization by the factor @xmath79 and where @xmath80 is the particle s mass number .similarly the rapidity y in the @xmath54 frame is normalized by the factor @xmath81 and this quantity is labeled @xmath82 . in the present experiments the normalization factors for the pion transverse momenta have the values @xmath83 ( gev / c)@xmath84 for @xmath85 agev bombarding energy , to obtain the equivalent values for protons these numbersshould be multiplied with a factor @xmath86 .for the rapidities the corresponding normalization factors are @xmath87 .the data from the drift chamber @xmath42 were analyzed by using three different tracking algorithms as described in @xcite .differences in tracking are mainly observed in case of the au + au reaction , they consist in a slightly different ( 10% ) efficiency with which negative pions are identified .equal tracking efficiencies can be obtained by adjusting the limits in which the tracking parameters are allowed to vary @xcite .we have verified that our results do not depend on these limits , and agree for the different versions of the tracking algorithm .our final data , i.e. transverse momentum and invariant mass spectra , are the average of three independent analyses with these different trackers , and this data sample has been used before in our study of the pion production in au + au @xcite and ni + ni @xcite reactions .the accuracy with which the particle momenta can be measured with the @xmath42 was determined in @xcite .for the transverse momentum the relative standard deviation @xmath88 is @xmath89 for @xmath90 gev / c , and then increases to @xmath91 for transverse momenta @xmath92 gev / c .the accuracy in determining the azimuthal angle @xmath93 is better than @xmath94 , the accuracy to determine the polar angle @xmath44 is better for protons ( @xmath95 ) than for pions ( @xmath96 ) .these values may be used to estimate the accuracy with which the mass @xmath62 of a baryon resonance b can be determined . without going into the details which will be morefully presented in the next section , the accuracy estimate is @xmath97 where @xmath98 is the sum of the proton and pion masses and @xmath99 is the angle between @xmath100 and @xmath101 .taking representative values , i.e. + @xmath102 one obtains for , e.g. the mass of the @xmath4 resonance an error of @xmath103 mev / c@xmath3 , with decreasing accuracy for larger masses @xmath62 .this estimate is certainly sufficient to determine the resonance mass distribution , it is however not representative , since a second source of systematic errors comes from the reconstruction of the combinatorial background of uncorrelated @xmath1 pairs , as will be discussed in the next section . in the present analysisthe number of directly measured @xmath1 pairs is larger than @xmath104 for a given reaction , to reconstruct the background for this reaction we have used approximately @xmath105 times more pairs .the statistical errors are therefore not the decisive factor in our analysis .we have employed two independent procedures to determine the mass distribution of baryon resonances excited in near - central collisions of identical nuclei . in the first methodthe pion @xmath0 spectrum is defolded to yield the mass distribution , in the second method the invariant mass distribution of the correlated @xmath1 pairs is deduced from the data .both methods rely on the assumption that the resonances involved in the production of pions decay via the emission of only one pion .this assumption is not necessarily fulfilled as was pointed out in the introduction .if the two - pion decay is present it effects the results of the two methods in different ways . in the first casethe presence of the two - pion channel will cause a distortion of the mass spectrum , in the second case it will mainly contribute to the background .it is therefore not expected that both methods should yield identical results , however these results should be reasonably close since the two - pion channel is expected to be weak @xcite .we therefore also take a reasonable agreement as an indication that systematic errors are sufficiently small .the sizes of the mass shifts @xmath106 and the widths @xmath107 deduced from our monte carlo studies are listed in table 4 .it is evident that the results from the correlation analyses are less affected by the coulomb distortion than those from the defolding analysis . nevertheless adding the @xmath26 and @xmath27 data one comes closest to the primordial conditions , which is the procedure employed in the present analysis .the discussion in section 2 is primarily concerned with the efficiency of reconstructing the baryon resonances within the geometrical boundaries of the @xmath42 .another efficiency reduction , which requires a careful investigation , is imposed by the limited capabilities of the available tracking algorithms to separate and identify individual particle tracks in the @xmath42 .these inefficiencies become largest when the mass of the decaying resonance is close to the threshold mass @xmath98 , i.e. when the relative momentum between the proton and the pion is small and the two particle tracks are close to each other . for the tracking algorithm , which analyzes particle tracks in the @xmath108 plane perpendicular to the @xmath109 direction of the magnetic field , the closeness of tracks is given by the relative angle @xmath110 and by the magnitude of @xmath111 . the smaller @xmath112 and the larger @xmath111the more difficult it becomes to separate pion tracks from proton tracks .the size of the efficiency reduction as function of @xmath112 and @xmath111 can be obtained from a comparison of the invariant mass distributions @xmath113 and @xmath114 as functions of @xmath112 and @xmath111 since in the latter all pairs are identified with identical efficiencies .we show in fig.12 the ratio @xmath115 for @xmath6 , @xmath7 , and monte carlo generated @xmath17 tracks as functions of @xmath112 , where the monte carlo simulation illustrates the ideal case of complete tracking efficiency . in the simulationthe conditions are similar to the experimental ones , in particular the fraction of @xmath1 pairs due to resonance decay is in the order of 1% . the comparison with the monte carlo simulation proves , that the probability to miss correlated @xmath1 pairs never exceeds 20% for a given value of @xmath112 , and that it is certainly more appropriate to correct for this inefficiency than to remove all data with @xmath116 , as was done in @xcite .that the efficiency reduction is different for @xmath6 and @xmath7 pairs was to be expected since the @xmath117 and @xmath27 have the same curvature signs , whereas these are different in case of @xmath117 and @xmath26 .furthermore the fact that in case of @xmath6 pairs the efficiency is reduced only for negative @xmath112 angles offers the possibility to check the consistency of the efficiency correction .the invariant mass spectra @xmath118 have to be the same for @xmath119 and @xmath120 .this check was done in our analyses and was shown to yield identical @xmath118 distributions within the given uncertainties .all data shown in this paper were therefore efficiency corrected according to the results of fig.12 and the similar dependences on @xmath111 .the sizes of the efficiency corrections were directly obtained from the data , their dependence on @xmath112 and @xmath111 was , however , smoothed by the spline algorithm .* acknowledgement * we would like to thank w. weinhold for providing us with his @xmath121 function . this work was supported in part by the bundesministerium fr forschung und technologie under contract 06 hd 525 i(3 ) and by the gesellschaft fr schwerionenforschung under contract hd pel k and by the korea research foundation under contract 1997 - 001-d00117 .references collaboration : d. pelte et al ., z. phys .a. * 357 * , 215(1997 ) collaboration : d. pelte et al ., z. phys .a. * 359 * , 55(1997 ) g.f .bertsch , s. das gupta , phys . rep . * 160 * , 189(1988 ) w. cassing , k. niita , s.j .wang , z. phys .a * 331 * , 439(1988 ) w. cassing , v. metag , u. mosel , k. niita , phys* 188 * , 363(1990 ) j. aichelin , phys .rep . * 202 * , 233(1991 ) s.a .bass gsi-93 - 13 report(1993 ) s.a .bass , c. hartnack , h. stcker , w. greiner , phys .c * 50 * , 2167(1994 ) s.a .bass , c. hartnack , h. stcker , w. greiner , phys .c * 51 * , 3343(1995 ) p. danielewicz , phys .rev c * 51 * , 716(1995 ) s. teis , w. cassing , m. effenberger , a. hombach , u. mosel , g. wolf , z. phys .a. * 356 * , 421(1997 ) r. brockmann , j.w .harris , a. sandoval , r. stock , h.stroebele , g. odyniec , h.g .pugh , l.s .schroeder , r.e .renfordt , d. schall , d. bangert , w. rauch , k.l .wolf , phys .lett . * 53 * , 2012(1984 ) a. sandoval , r. brockmann , r. stock , h.stroebele , d. bangert , w. rauch , r.e .renfordt , d. schall , j.w .harris , g. odyniec , h.g .pugh , k.l .wolf , gsi-85 - 10 report(1985 ) p. senger , in : multiparticle correlations and nuclear reactions , ed . by j. aichelin and d. ardouin ( world scientific publ .co. , 1994 ) , page 285 collaboration : j. barrette et al .b * 351 * , 93(1995 ) collaboration : m. trzaska et al ., z. phys .a * 340 * , 325(1991 ) j. chiba , t. kobayashi , t. nadae , i. arai , n. kato , h. katayama , a. manabe , m. tanaka , k. tomizawa , d. beatty , g. edwards , c. glashauser , g.j .kumartzki , r.d .ransome , t.t .baker , phys .* 67 * , 1982(1991 ) t. hennino , b. ramstein , d. bachalier .bohlen , j.l .boyard , c. ellegaard , c. gaarde , j. gosset , j.c .jourdain , j.s .larsen , m.c .lemaire , d.l .hote , h.p .morsch , m. sterlund , j. poitou , p. radvanyi , m. roy - stephan , t. sams , k. sneppen , o. valette , p. zupranski , phys .b * 283 * , 42(1992 ) collaboration : e.l .hjort et al . ,* 79 * , 4345(1997 ) m. trzaska , in : multiparticle correlations and nuclear reactions , ed . by j. aichelin and d. ardouin ( world scientific publ .co. , 1994 ) , page 95 collaboration : f .- d .berg et al . , phys . rev72 * , 977(1994 ) collaboration : m.a .lisa et al . , phys .lett . * 75 * , 2662(1995 ) collaboration : c. mntz et al . , z. phys .a * 357 * , 1399(1997 ) p.j .siemens , j.o .rasmussen , phys .lett . * 42 * , 880(1979 ) j. sollfrank , p. koch , u. heinz , z. phys .c * 52 * , 593(1991 ) a.p .dempster , n.m .laird , d.b .rubin , j. roy .b * 39 * , 1(1977 ) h. durst , staatsexamensarbeit ( faculty of mathematic , university of heidelberg ( 1995 ) ) , unpublished collaboration : n. herrmann et al . , nucl .phys . * a610 * , 49c(1996 ) ; b. hong , et al .b * 407 * , 115(1997 ) ; b. hong et al . ,c * 57 * , 244(1998 ) d. lhte , nucl .a337 * , 544(1994 ) collaboration : m. korolija et al ., to be published j. cugnon , m.c .lemaire , nucl .a489 * , 781(1988 ) c.m .li , j. phys .g * 22 * , 1673(1996 ) collaboration : r. barth et al .78 * , 4007(1997 ) collaboration : j. ritman et al ., z. phys .a * 352 * , 355(1995 ) w. weinhold , b.l .friman , w. nrenberg , acta phys . pol .b * 27 * , 3249(1996 ) collaboration : r. kotte et al ., z. phys . a * 359 * , 47(1997 ) collaboration : w. reisdorf et al . ,nucl . phys . *a612 * , 493(1997 ) r. averbeck et al . , nucl - ex/9803001 , to be published rev . of part .b * 239 * ( 1990 ) s. teis , w. cassing , m. effenberger , a. hombach , u. mosel , g. wolf , z. phys .a * 359 * , 297(1997 ) v.s .uma maheswari , c. fuchs , a. faessler , l. sehn , d.s .kosov , z. wang , nucl - th/9706004 , to be published in nucl .phys . * a * g.f .chapline , m.h .johnson , e. teller , m.s.weiss , phys .d * 8 * , 4302(1973 )

<S> the mass distributions of baryon resonances populated in near - central collisions of au on au and ni on ni are deduced by defolding the @xmath0 spectra of charged pions by a method which does not depend on a specific resonance shape . </S><S> in addition the mass distributions of resonances are obtained from the invariant masses of @xmath1 pairs . with both methods the deduced mass distributions </S><S> are shifted by an average value of @xmath2 mev / c@xmath3 relative to the mass distribution of the free @xmath4 resonance , the distributions descent almost exponentially towards mass values of @xmath5 mev / c@xmath3 . </S><S> the observed differences between @xmath6 and @xmath7 pairs indicate a contribution of isospin @xmath8 resonances . </S><S> the attempt to consistently describe the deduced mass distributions and the reconstructed kinetic energy spectra of the resonances leads to new insights about the freeze out conditions , i.e. to rather low temperatures and large expansion velocities . </S><S> @xmath9 0.5 cm 0.4 cm 0.5ex plus0.2ex minus0.1ex 0ex plus0.2ex national institute for physics and nuclear engineering , bucharest , romania central research institute for physics , budapest , hungary laboratoire de physique corpusculaire , in2p3/cnrs , and universit blaise pascal , clermont - ferrand , france gesellschaft fr schwerionenforschung , darmstadt , germany forschungszentrum rossendorf , dresden , germany physikalisches </S><S> institut der universitt heidelberg , heidelberg , germany institute for theoretical and experimental physics , moscow , russia kurchatov institute , moscow , russia korea university , seoul , korea institut de recherches subatomiques , in2p3/cnrs , and universit louis pasteur , strasbourg , france institute of experimental physics , warsaw university , poland rudjer boskovic institute , zagreb , croatia </S>

astro-ph0606516

when a scientist tries to analyze a given observed data set , it is customary to begin by examining the data in various ways , such as plotting the data and looking for patterns . in a second step ,the scientist proposes a number of physically plausible models that can reproduce the observed data set .a model fitting procedure is then applied to the data set so that the parameters that characterize each model are inferred .after all the proposed models are fitted to the data , the next step is to compare them and infer which of the fitted models is the most suitable .methods for such a task have been developed .for instance , for models that represent a hierarchical structure , we can use akaike s information criterion @xcite or mallows @xmath0 @xcite .these methods can be applied even in the case that the set of proposed models does not contain the perfect model . in this case , the aim is to select the most optimal one . if the models are of completely different type and do not belong to a hierarchical structure of models , cross - validation type methods can be applied .however , they can be computationally demanding . the standard procedure to select the optimum model is to make use of the occam s razor principle ( also known as the principle of parsimony ) .this principle is commonly applied in science and is usually thought to be an heuristic approach for eliminating unnecessary complex hypothesis .the principle states that the selected model has to provide an equilibrium between the model complexity and its fidelity to the data . nevertheless , determining the simplest model is often very complicated .it is usually argued that the number of parameters that parameterizes the model should be smaller than the number of degrees of freedom of the data .however , it is a difficult matter to estimate the number of degrees of freedom of the data . an alternative and successful procedure is to consider the problem in terms of a communication process .assume that a sender s is interested in sending a given observation to a receiver r. several techniques are available for such a task .the most trivial one is to send the whole amount of data through a given channel .if the data set is very large , the length of the message will be consequently very large . if s is able to model the observation using a given model , it is wiser and shorter to transmit the model followed by the points in the observation that are not correctly reproduced by the model .if the model is good and simple enough , the length of the message between s and r will be much shorter than sending the complete observed data set .this compression will be degraded when the model needed for explaining the observations is made unnecessarily complex .@xcite was the first in suggesting that the code length could be used for model comparison .this principle is nowadays known as the minimum description length ( mdl ) principle , which states that we should _ choose the model that gives the shortest description of the data_. in this framework , there is an interplay between models that are concise and easy to describe and models that produce a good fit to the data set and captures the important features evident in the data .of course , this is neither the only nor the best strategy for such purposes .however , its application is desirable in comparison with ad - hoc or trial - and - error ways of performing model selection .the reason is that mdl principle has strong theoretical roots that lie on the kolmogorov complexity theory @xcite .as we will show , the practical application of the mdl principle is very easy to implement and it constitutes an ideal approach for model selection .this paper will be focused on one of the version of the mdl principle , the so - called two - part mdl .this strategy was developed by rissanen in a series of papers published during the seventies and eighties ( e.g. , * ? ? ?* ; * ? ? ?* ; * ? ? ?* ) and summarized in @xcite . we give a brief summary of the main results in section [ sec : theory ] . as shown below , the main idea is to write the description length of a given model applied to a data set as the sum of the length of the code for describing the model and the length of the code for describing the data set fitted by the model . the interpretation of spectropolarimetric data in solar and stellar observations allows to infer information about the properties of the magnetic field .almost always , the recovery of the magnetic field vector is based on the assumption of a model .these models depend on a given amount of variables that can be estimated by fitting the observed stokes profiles . among these models, we can find the milne - eddington approximation , the lte approximation , the misma hypothesis @xcite , etc .sometimes , it is clear that a model is the most appropriate when observational clues are available .however , it is customary that these clues are not present and one relies on one model based on completely or partially subjective reasons .some of these models are based on an extraordinarily large number of parameters . until now , there is not any critical investigation toward analyzing whether enough information is available in the observations to constraint such a large amount of parameters .this work is a first step in this direction .we propose the application of the mdl criterion to quantitatively differentiate between possible models taking into account the information available in the observations .@xcite related the problem of finite dimensional parameter estimation to the problem of designing an optimal encoding scheme .we consider the problem of transmitting a set of data from a sender s to a receiver r. the sender must first communicate the type of model that can be used for describing the data .consider , for instance , a set of points that we want to fit with a polynomial model . in this case , neither the order of the polynomial @xmath1 , nor the value of the parameters of the polynomial @xmath2 are known , so that this constitutes an extremely ill - posed problem .however , the mdl principle can be used to find the `` most plausible '' model that explains the observed points . in the first step, the sender has to communicate the order of the polynomial @xmath1 .when the sender and the receiver agree on the type of model to use , the sender has to communicate the model itself , sending through the channel the @xmath1 parameters @xmath3 . if noise is present or if the proposed model is incomplete , the sender has to additionally communicate the deviations of the model from the data. the question of simplicity can consequently be tackled by taking advantage of the work of @xcite .each model is reduced to bits , the most fundamental information unit .this allowed to transform the occam s razor principle into a completely functional principle .the measure of simplicity is just the number of bits required to correctly and univoquely transmit a set of observations by using a model .the sender s takes a set of observations as input , encodes and sends a message that contains all the information about the model and the data to the receiver r. finally , the receiver decodes the message and produces an output . in information - preserving encoding schemes ,the output obtained by r has to be identical to the original observation performed by s. let @xmath4 be a set of observations ( dataset ) and @xmath5 a model that is used to describe them .the quantity @xmath6 represents the length of the code in bits necessary to encode the model @xmath5 .as well , @xmath7 represents the length of the data encoded using the model @xmath5 ( this term can be alternatively seen as the residual between the data @xmath4 and the model @xmath5 ) .the total length of the message is : @xmath8 the mdl principle tries to minimize @xmath9 and the model associated with this minimum length is selected as the most plausible model .the previous analysis can be alternatively viewed from the bayesian perspective ( see , e.g. , * ? ? ?the aim is to infer a model @xmath5 from a set of observations @xmath4 .the solution lies in choosing the model that maximizes the posterior probability @xmath10 , which can be expressed as follows with the aid of the bayes theorem : @xmath11 the term @xmath12 can be considered as a normalizing factor and represents the probability that the data set @xmath4 occurs .the term @xmath13 is the a priori probability of the model , that is , the probability that the model @xmath5 is true before any data set has been observed .finally , @xmath14 is the likelihood of the data given the model @xmath5 .the relationship between the mdl formalism given by eq .( [ eq : length ] ) and the bayes formalism given by eq .( [ eq : bayes ] ) is obtained by making use of shannon s optimal coding theorem @xcite .this theorem states that the length of the ideal code for a value @xmath15 of a variable @xmath16 which follows from a known probability distribution @xmath17 is given by : @xmath18 taking the negative logarithm of eq . ([ eq : bayes ] ) , we obtain : @xmath19 the most plausible model is the one that minimizes @xmath20 , that is , the one that minimizes @xmath21 .note that we have ignored the influence of @xmath12 since it is a constant that is shared for all the models and it is only associated with the data set . according to shannon s theorem , the minimization of eq .( [ eq : bayes_log ] ) is equivalent to the minimization of eq .( [ eq : length ] ) .shannon s theorem states the length of the optimal code needed for transmitting a given number @xmath15 whose probability distribution is known .although extremely important , the theorem can be useless because this probability distribution is usually not known and only approximate values of the length of the message can be obtained .an example is when one needs to transmit a set of integer or real numbers for which no probability distribution is known .it has been demonstrated that knowing exactly the encoding scheme is of accesory importance @xcite .what is fundamental to know is which model gives the minimal length of the message given an arbitrary encoding scheme . in the following , we assume that the probability distribution is not known and we present existing estimations for the length of the message needed for communicating integer and real numbers .these formulae represent the first estimations that were obtained for the lengths of the message for communicating models @xcite .since they are not based on any probability distribution for the integer or real numbers , they are known as universal priors @xcite .assume that we want to encode an integer number @xmath22 in its binary representation .the binary representation of @xmath22 has a length that can be estimated to be of the order of @xmath23 bits .if a set of integers are to be encoded , confusion arises if we pack all these binary digits together because the receiver does not know where the representation of the first digit ends and the second digit starts . to resolve this, one can also encode the length of the binary representation of @xmath22 in binary representation , whose length can also be estimated to be @xmath24 .the same problem arises again since the receiver does not know the length of this preamble representation .this problem can be solved by giving the length of the preamble as @xmath25 as another preamble .this procedure can be iterated until @xmath26 is as close to zero ( and positive ) as desired .the integer number can be encoded with a scheme that includes these preambles , so that the total length of the code is given by : @xmath27 where the constant @xmath28 is included for consistency ( see , e.g. , * ? ? ?the symbol @xmath29 is chosen to represent the previous sum . for large numbers , the dominant factor is @xmath23 , so that we can approximately assume that the length of the message for encoding an integer number is of the order of @xmath23 for sufficiently large @xmath22 @xcite . for encoding integers with sign, we can add a single bit for setting the sign , so that the length of the message is @xmath30 .the dominant factor is again @xmath23 .encoding real numbers is more complicated than integer numbers because we would need an infinite number of bits to encode the real number to infinite precision . to solve this in practice, we have to encode the real number @xmath15 assuming a precision @xmath31 , so that the encoded number @xmath32 fulfills @xmath33 .once the precision is fixed , we can encode the integer part @xmath34 and the fractional part of @xmath32 separately .it can be shown @xcite that the length of the message is @xmath35 . as stated before , if @xmath15 is large and taking into account that @xmath36 , we can estimate the length of the message for encoding a real number @xmath15 with @xmath37 note that the length tends to zero when the number to encode approaches the precision . in the limit situation that @xmath15 is smaller than the precision , the length takes a negative value that does not have meaning . with the previous considerations in mind, the model selection problem can be established .let @xmath38 be the model that generates the data or the one that is more plausible and that it belongs to a class of @xmath39 models @xmath5 .consider that each model is parameterized by @xmath40 parameters @xmath41 , with @xmath42 and @xmath43 .given a set of observations , our aim is to choose the most plausible model @xmath38 from the set @xmath5 and to estimate the parameters @xmath44 .the sender splits up the message in two parts .the first one contains the model itself and the second one contains the departures between the model and the data .we transmit the information about the model by first encoding the number of parameters that characterizes it and then transmitting the parameters themselves with a given precision @xmath45 associated with each one : @xmath46.\ ] ] since the @xmath29 function behaves as @xmath23 for sufficiently large @xmath22 @xcite , we can transform the previous formula to : @xmath47.\ ] ] the previous equation gives the length of the model for encoding the parameters with arbitrary precision @xmath45 .however , it makes no sense to increase the precision of the parameters unnecessarily because there may be no information in the data for such a task .it has been demonstrated by @xcite that if the optimum parameters are computed from a large set of @xmath22 observed data points , the precision of the parameters can be effectively encoded with only @xmath48 bits .the reason for this is that , if the parameters are being estimated from the data , it makes no sense to encode them with a precision larger than the standard error of the estimation . for a typical estimation of the @xmath41 parameter ,the standard error decreases with the number of data points as @xmath49 .therefore , the code length for encoding a real number with precision @xmath49 is @xmath50 .if all the points are used for estimating all the parameters , we can rewrite eq . ( [ eq : length_n_params2 ] ) as : @xmath51 additionally , we have to encode the observed data given the model to obtain the length @xmath7 . in the simple case in which we do not have any information about the probability distribution of the data given the model, we can apply the same encoding scheme we have applied for describing the model .assume that we have @xmath22 observed data points @xmath52 and that the model @xmath53 produces a fit to the observations such that the residual can be written as : @xmath54 the encoding is performed by saving the number of data points and the value of each data point by using eq .( [ eq : length_real ] ) .the length @xmath7 is : @xmath55,\ ] ] which can be simplified if the number of data points @xmath22 is large enough to give : @xmath56.\ ] ] putting together eqs .( [ eq : length_n_params ] ) and ( [ eq : length_data ] ) , the total 2-part length of the message is : @xmath57 \nonumber \\ & + & \log^ * n + \sum_{j=1}^{n } \big [ \log^ * r_j + \log^ * ( 1/\delta_j ) \big].\end{aligned}\ ] ] if the number of data points @xmath22 is large enough , we can use eqs .( [ eq : length_n_params2 ] ) and ( [ eq : length_data2 ] ) to obtain : @xmath58 \nonumber \\ & + & \log_2 n + \sum_{j=1}^{n } \big [ \log_2 r_j - \log_2 \delta_j \big].\end{aligned}\ ] ] the previous equations are considered for an optimum encoding scheme . however , a simpler and more computer - oriented estimation can also be obtained based on the previous results with good results @xcite .we assume that integer numbers are saved in a computer using @xmath59 bits , while real numbers are saved in the memory of the computer using @xmath60 bits .usually , integer numbers are saved using @xmath61 bits , while real numbers can be stored using either @xmath62 or @xmath63 bits depending on the desired precision .therefore , we can estimate the length of the message that describes the model ( the number of bits required for its storage in the memory of the computer ) by : @xmath64 where the number of bits required for transmitting the number of parameters given by @xmath65 can be usually neglected with respect to the length of the message used for transmitting the parameters themselves . concerning the storage of the data , we consider that the model output is correct when the residual is smaller than a given precision threshold @xmath31 . if the model output is incorrect , we store the difference as a real number with @xmath60 bits , so that the length is given by : @xmath66 the length of the parameters associated with the model increase linearly with the number of parameters , while the length of the data set decreases , with a slope that depends on how well the model fits the data .the previous encoding schemes have been obtained using universal priors , thus assuming no knowledge at all about the distribution of the values of the data and/or model . however ,if some information about the probability distribution is known , it can be incorporated in the description of the encoding length through eq .( [ eq : shannon ] ) .a typical case is when the observed data is contaminated with noise described by a gaussian distribution with zero mean and a given standard deviation @xmath67 ( that may even be unknown ) . for the set of @xmath22 observations @xmath68 ,the probability distribution of the residuals is given by : @xmath69,\ ] ] where @xmath70 is the vector of @xmath22 residual @xmath71 and @xmath72 is the variance for each data point .assuming the same variance @xmath73 for all the data points and using shannon s theorem , we obtain : @xmath74 where @xmath75 is the residual sum of squares .the previous equation constitute the estimation of the length for communicating the data set when we have an estimation for the value of @xmath67 . if the variance is not known @xcite , we can use the maximum likelihood estimation @xmath76 and obtain :@xmath77 for demonstrating the previous machinery in a simple practical problem ,let us assume that we have a noisy linear combination of sinusoidal signal : @xmath78 the variable @xmath15 is always inside the interval @xmath79 $ ] and @xmath80 is a noise term with zero mean and standard deviation @xmath67 .let us consider that we have sampled the @xmath15 axis in 512 points and that @xmath81 .the frequencies of the signals can be obtained by performing the fast fourier transform of the signal . however , for applying the mdl principle , we consider models of the type @xmath82 the aim is to obtain the most plausible value of @xmath83 that produces the smallest encoding length of the model and data given the model .firstly , we apply the code length described in [ sec : computer_oriented ] using a threshold equal to the standard deviation of the data .the results are shown in the left panel of fig .[ fig : fourier ] .note that the number of bits necessary to encode the model increases linearly .the number of bits to describe the residual between the model and the observed data decreases rapidly when @xmath84 .the sum of both terms has a minimum around @xmath85 , which coincides with the maximum value of @xmath83 in the original signal .the right panel shows the results obtained when we take into account that the residuals between the model and the data are well characterized by a gaussian distribution and that the number of points is sufficiently large . in this case , the length of the model is given by eq .( [ eq : length_n_params3 ] ) , while the length of the data is given by eq .( [ eq : length_gaussian ] ) .the minimum of the total length gives the most plausible value of @xmath85 , compatible with the original data and with the value given by the previous estimation of the total encoding length .in this section we apply the mdl principle to two selected problems in the field of solar spectropolarimetry in order to show the potential of this technique for model selection .our aim is to apply the mdl criterion systematically to other similar problems in the future . as a first application, we consider the case of principal component analysis ( pca ) de - noising of spectropolarimetric observations .the dataset consists on full - stokes observations of the two at 15648 and 15652 in an extremely quiet region of the solar surface .this dataset has been used by @xcite to investigate the magnetic properties of the quiet sun .the signal - to - noise ratio of the observations have been improved by using the pca de - noising technique .pca is a statistical technique that , given a data set , produces a set of orthogonal vectors and eigenvalues that can be used for decomposing the original data .these eigenvectors point in the directions of maximum covariance .the eigenvector associated with the largest eigenvalue points along the direction with the largest covariance in the data and so on .pca provides an `` optimal '' basis set for decomposing ( and reconstructing ) the data set .alternatively , it has been used for compressing data ( by saving only the eigenvalues and eigenvectors ) or for efficiently inverting stokes profiles @xcite . for computing the pca decomposition ,the covariance matrix of the observations has to be diagonalized .once the eigenvectors @xmath86 ( vectors whose dimension @xmath87 equal the number of wavelenghts in the dataset ) and eigenvalues @xmath88 of the covariance matrix associated with the stokes parameter @xmath89 are obtained , any of the stokes profiles can be decomposed as : @xmath90 where the subindex@xmath91 refers to the wavelength position .the previous equation can be alternatively seen as a technique for reconstructing the original signal from the eigenvalues and eigenvectors obtained after the pca decomposition .the precision of the reconstruction can be modified by changing the value of the number of eigenvectors included ( @xmath92 ) .the pca reconstruction assures that the error of the reconstruction decreases when more eigenvectors are included .the eigenvectors with the largest eigenvalues represent the features that are more statistically representative of the dataset , while the noise and particular features are accounted for by the eigenvectors with the smallest eigenvalues . as a consequence , it is possible to get rid of the majority of the noise in the observation by stopping the summation of eq .( [ eq : pca_summation ] ) in a suitable @xmath93 . nevertheless , it is not an easy task to find a criterion for selecting this @xmath94 because instrumental effects plus data reduction can introduce spurious signals that have some kind of correlation . as a consequence , they contribute to the eigenvectors that carry the relevant polarimetric information .we propose to use the mdl criterion to select this optimal @xmath94 .the experiment is carried out with the stokes v profiles of the lines at 15648 and 15652 observed in a very quiet internetwork region of the sun described by @xcite .the original data presents a signal - to - noise ratio ( snr ) of @xmath955 for the 15648 line and @xmath952 for the 15652 line .the summation of eq .( [ eq : pca_summation ] ) is calculated for increasing values of @xmath92 and the mdl length is calculated using the technique described in [ sec : computer_oriented ] .the length of the model @xmath6 is obtained by calculating the number of bits to represent the eigenvectors @xmath96 with @xmath97 plus the coefficients @xmath98 for expanding all the profiles in the field of view .the length of the data set given the model @xmath7 is obtained by calculating the number of bits needed for encoding the reconstructed profiles that differ from the original profiles by more than a given threshold .figure [ fig : pca ] shows these lengths versus @xmath92 for different values of this threshold .the length of the model is plotted in dot - dashed lines , the length of the data in dashed line and the total length in solid line .we have also marked the value of @xmath92 at which we obtain the minimum of the total length .this is the mdl optimum value @xmath94 for the number of eigenvectors .note that this minimum increases when the allowed threshold decreases , a consequence of putting more restrictions to the model .of course , this threshold should be chosen consistent with the expected noise in the observations .the diagnostic of magnetic fields via the interpretation of spectropolarimetric observations is often based on the assumption of a model .sometimes , the observations themselves do not carry enough information for discriminating among several models .a similar problem arises when a model can have an arbitrarily large number of parameters and it is not an easy task to select an optimum value of such parameters .a commonly used technique to minimize over - fitting is to use models with as few parameters as possible .nevertheless , the data may contain enough information for constraining more parameters and we may be using overly simple models to interpret the observations .we propose to use the mdl criterion to discriminate among different models that can be used to describe a set of observations .our aim is to introduce in the community an easy technique for confronting different models when applied to the same data set . in the framework of the mdl criterion ,the researcher is able to objectively discriminate one of the models among the others , making sure that the selected model explains the observations without over - fitting them .we demonstrate our approach by using the lte inversion code sir ( stokes inversion based on response functions ; * ? ? ?an example of the stokes profiles of the lines at 15648 and 15652 observed in a very quiet internetwork region are shown in fig .[ fig : observations ] @xcite .the inversion is carried out with a two - component model ( a magnetic one occupying a fraction of the resolution element and a non - magnetic one filling up the rest of the space ) .the observations clearly show strongly distorted stokes profiles that can not be correctly reproduced with this simple two - component model .although very simple , this test demonstrates the capabilities of the mdl criterion to pick up a model when none of the models of the proposed set is able to correctly fit the observations .sir represents the variation with depth of the thermodynamical and magnetic properties with the aid of nodes that are equidistant in the @xmath99 axis .it is of interest to note that when a new node is included for representing the depth variation of a physical quantity , the previous nodes are shifted so that the final distribution is again equidistant .splines are used to interpolate these quantities between the nodes .the number of nodes of the temperature ( or the magnetic field strength ) is increased and the mdl criterion is used for selecting the optimal values .we follow the prescriptions proposed in [sec : computer_oriented ] .the length of the model is calculated as the number of bits to encode the number of parameters plus their values .the length of the data given the model is chosen to be equal to the number of bits neccesary to encode the points in the profile for which the relative difference between the model and the data is above a certain threshold .this threshold represents the tolerance we allow in our model for considering that it fits our observations .the first experiment consists in selecting the optimal number of nodes in the temperature when the number of nodes in the rest of variables are kept constant .we only compare the stokes i observed and synthetic profiles in this experiment because it is the only stokes parameter that is almost unsensitive to the magnetic properties of the atmosphere .this is not generally the case since stokes i can be also sensitive to the magnetic field in the strong field regime . in analysing internetwork quiet sun stokes profiles , the filling factor of the magnetic component needs to be very small . in such a situation , the emergent stokes i profile fundamentally depends on the properties of the non - magnetic component , while the emergent stokes v profile depends only on the properties of the magnetic component .we consider that a point in the profile reproduces the observations if the relative error between the observed and the synthetic profiles is below 2% .the results for the message length are shown in the left panel of fig .[ fig : lte ] , while the fit is shown in the left panel of fig .[ fig : observations ] .the mdl criterion demonstrates that @xmath952 nodes are required in the temperature depth profile .fewer nodes give a fit to the stokes profiles that is too bad .more nodes give a model that takes more bits to communicate than the uncorrectly fitted points themselves .the second experiment is similar , but in this case we select the number of nodes of the magnetic field strength .the threshold for the relative error in the stokes i profiles is 2% , while we increase it to 10% for the stokes v profiles .this large relative error for the stokes v profile is motivated by the poor fit we obtain to the observed stokes profiles with the simple two - component model .the message length is shown in the right panel of fig .[ fig : lte ] , while the fit is shown in the right panel of fig .[ fig : observations ] .the optimum number of nodes for the magnetic field suggested by the mdl criterion is @xmath953 . in this sectionwe have presented a very simple problem concerning the selection of the optimal number of parameters in an lte inversion .however , we consider that this approach will be of great interest for solving such a difficult problem in a simple way and we suggest calculating the mdl criterion for all the fits performed to an observed profile .we have presented the minimum description length principle developed by @xcite to discriminate between a set of available models that can approximate a given data set .we have briefly presented its relation with the bayesian approach of model selection through the application of the shannon s theorem . in our opinion, the mdl principle presents a user friendly procedure for model selection .we have presented simple ways of estimating the message length that can be applied for communicating integer and real numbers . a more computer - oriented and easy to implement procedure has been also shown . for the sake of clarity , we have applied the mdl principle to simplified problems .the selection of the optimal number of pca components when de - noising observed stokes profiles and the selection of the optimal number of nodes in the temperature and magnetic field depth profile obtained from lte inversion of stokes profiles .the results show the potential of this technique for model selection , with the advantage of being very simple to calculate . as the main conclusion of this paper , we propose using the mdl principle as a way to quantitatively select among different competing models .we propose to include the description length as one of the final outputs of any inversion code .this makes it very easy to select the optimal model from a proposed set of models based on the framework of the mdl principle .it is of interest to stress that this principle can be used to select among models that are based on different scenarios .usually , most complex scenarios translate into more degrees of freedom that may not be constrained by the observations . using the mdl principle , the selected model among all the possibilities might not give the best fit to the observations but it represents the model that produces a good fit with a conservative number of parameters .as a final remark , a possible future way of research may be how the mdl principle can be implemented as a regularization of the merit function in the existing inversion techniques . as a consequence, one would end up with an inversion code that automatically selects the optimal model .we thank m. j. martnez gonzlez for helpful discussions on the subject of the paper .we also thank thornsten a. carroll for his careful review and useful comments .this research has been funded by the spanish ministerio de educacin y ciencia through project aya2004 - 05792 .

<S> it is shown that the two - part minimum description length principle can be used to discriminate among different models that can explain a given observed dataset . </S><S> the description length is chosen to be the sum of the lengths of the message needed to encode the model plus the message needed to encode the data when the model is applied to the dataset . </S><S> it is verified that the proposed principle can efficiently distinguish the model that correctly fits the observations while avoiding over - fitting . </S><S> the capabilities of this criterion are shown in two simple problems for the analysis of observed spectropolarimetric signals . </S><S> the first is the de - noising of observations with the aid of the pca technique . </S><S> the second is the selection of the optimal number of parameters in lte inversions . </S><S> we propose this criterion as a quantitative approach for distinguising the most plausible model among a set of proposed models . </S><S> this quantity is very easy to implement as an additional output on the existing inversion codes . </S>

1409.2771

construction of the universal potential energy function for molecules has been a challenging and active field of research in chemical physics .the reason for this is that the potential energy function succinctly carries the necessary informations relevant for a molecule .thus , an enormous number of such functions have been proposed over the years , after the publication of three - parameter , exponential morse potential @xcite about 85 years ago .the literature is huge ; the following reference gives some of the older as well as relatively newer empirical functions @xcite .usually , larger the number of parameters in the analytical potential energy function , better the fit with experimental data .while a few of these such as morse , mie - type and pseudoharmonic potentials offer _ exact _ analytic solutions @xcite , most of these unfortunately can not be analytically solved for arbitrary vibrational and rotational quantum numbers .this necessitates the use of approximation schemes for their solutions .recently some attention has been paid on an analytic tietz - hua ( th ) model potential @xcite for ro - vibrational levels in diatomic molecules , expressed in the following form , @xmath4 ^ 2 ; \ \ \ b_h=\beta ( 1-c_h),\ ] ] where @xmath5 relates to the molecular bond length , @xmath6 the morse constant , @xmath7 signifies the potential well depth , @xmath8 denotes the internuclear distance , while @xmath9 implies an optimization parameter obtained from _ ab initio _ or rydberg - klein - rees intramolecular potentials respectively . note that in the limit of potential constant @xmath9 approaching zero , th potential reduces to the familiar morse potential @xcite .this potential supposedly describes the molecular dynamics ( especially at high rotational and vibrational quantum numbers ) more realistically than the traditional morse potential @xcite .also it has been noted that this usually fits the experimental rydberg - klein - rees curve more closely than the morse function , especially near the dissociation limit @xcite . in another study , using hamilton - jacobi theory in conjunction with bohr - sommerfeld quantization rule , analytical expressions for rotational - vibrational levels of diatomic molecules within th model have been derived @xcite .radial probability distributions of some diatomic molecules in excited rotational - vibrational states have also been reported using this route @xcite .very recently , _ exact _ analytical solution of the radial schrdinger equation with th potential has been provided for @xmath10 waves within a parametric nikiforov - uvarov method @xcite . in another development ,approximate analytical solutions of the dirac equation with th potential were obtained for arbitrary spin - orbit quantum number using the pekeris scheme @xcite .the purpose of this work is to offer approximate solution of radial schrdinger equation with th potential for molecules .as already mentioned , the @xmath0 states of this potential can be obtained in closed analytic form ; while eigenvalues and eigenfunctions of @xmath11 states of th oscillator has not yet been reported in the literature , to the best of our knowledge .here we take the help of generalized pseudospectral method ( gps ) for an optimal effective discretization of the relevant schrdinger equation .this method has produced very promising results for a number of situations having physical , chemical interest , including structure , dynamics in atomic and molecular physics .accurate eigenvalues , eigenfunctions were obtained for low as well as higher states for a class of potentials such as spiked harmonic oscillators , logarithmic , rational , power - law , hulthn , yukawa , exponentially screened coulomb potentials , etc . @xcite .thus we make a detailed study on the bound - state spectrum of th oscillator with particular reference to diatomic molecules .ro - vibrational energies and radial densities are studied for both @xmath10-wave and _ rotational _ sates having _ arbitrary _ low and high vibrational quantum number .this will also enable us to judge the viability and feasibility of current approach in the context of diatomic molecular potentials .to this end , arbitrary @xmath12 states are reported for six molecules , _ viz ._ , h@xmath1 , hf , n@xmath1 , no , o@xmath1 , o@xmath2 . comparison with literature data are made wherever possible .the article is organized as follows . a brief overview of the adopted method is given in section ii .then a discussion of the results is presented in section iii , while we conclude with a few remarks in section iv .this method has been discussed in detail earlier ( see the references @xcite and therein ) .thus it suffices to present here only a brief summary of the essential steps involved . without loss of generality , the desired time - independent radial schrdinger equation , to be solved ,can be written as ( atomic units employed unless otherwise mentioned ) , @xmath13 \psi_{n,\ell}(r)=e_{n,\ell}\ \psi_{n,\ell}(r)\ ] ] where @xmath14 is the th potential , as given in eq .( 1 ) , while @xmath15 , @xmath16 signify the radial and angular momentum quantum numbers respectively .the gps formalism facilitates the use of a denser mesh at small distance and relatively coarser mesh at large distance preserving similar accuracy at both the regions .a key step in this approach is to approximate a function @xmath17 defined in the interval @xmath18 $ ] by an n - th order polynomial @xmath19 _ exactly _ at the discrete collocation points @xmath20 , @xmath21 within the legendre pseudospectral method that we are using currently , @xmath22 , @xmath23 , and @xmath20 s @xmath24 are determined from the roots of first derivative of the legendre polynomial @xmath25 with respect to @xmath26 , i.e. , @xmath27 .the @xmath28s in eq .( 3 ) are termed the cardinal functions expressed as , @xmath29 satisfying the relation @xmath30 . at this stage, a transformation @xmath31 is used to map the semi - infinite domain @xmath32 $ ] onto the finite domain @xmath18 $ ] , along with an algebraic nonlinear mapping , @xmath33 , with l , @xmath34 being two mapping parameters . finally introducing a symmetrization proceduregives a transformed hamiltonian of the following form , @xmath35 the advantage lies in the fact that this leads to a _ symmetric _ matrix eigenvalue problem which can be solved readily and efficiently to give accurate eigenvalues , eigenfunctions by using standard routines .note that @xmath36 for the above transformation and one finally obtains a set of discretized coupled equations .considerable checks have been made on the convergence of eigenvalues with respect to the mapping parameters for a decent number of molecular states .after a series of such test calculations , a choice has been made at the point where the results changed negligibly with such variations . in this way , a consistent and uniform set of parameters ( @xmath37 @xmath38 and @xmath39 ) has been used ..[tab : table1]spectroscopic parameters of the molecules , used in present calculation , taken from @xcite . [ cols="^,^,^,^,^,^,^",options="header " , ] next we proceed for a detailed investigation on the energy variations for three selected molecules _ viz . , _h@xmath1 , hf and no respectively .the top three panels ( a ) , ( b ) , ( c ) in fig .1 depict the variations of e@xmath40d@xmath41 ( in ev ) with respect to the angular quantum number @xmath16 for h@xmath1 , hf and no .these are given for six values of vibrational quantum number , _ viz . , _@xmath42 for h@xmath1 ; seven values of @xmath15 ( @xmath43 in addition to all the six @xmath15 in h@xmath1 ) for hf ; and nine values of @xmath15 ( @xmath44 in addition to all the seven @xmath15 in hf ) for no .note that the @xmath16 axis goes to 30 for h@xmath1 , while for the other two this is extended to 40 .this happens because of the fact that a limited number of bound states is supported by the potential .such bound states occur in larger number for no and hf than in h@xmath1. it may be worthwhile mentioning here that , the maximum vibrational quantum number @xmath45 and maximum rotational quantum number @xmath46 for h@xmath1 , hf , n@xmath1 , no , o@xmath1 , o@xmath2 are ( 22,39 ) , ( 28,66 ) , ( 66,260 ) , ( 56,230 ) , ( 52,220 ) and ( 56,235 ) respectively @xcite .the above @xmath45 s are to be contrasted with the corresponding values of 18 , 23 , 82 , 67 , 65 and 58 in morse potential for the six molecules under investigation @xcite .much larger differences in @xmath45 in th potential ( 96 ) and morse potential ( 174 ) have been observed in i@xmath1 , where the actual value is 107 @xcite .the plots for other three molecules are omitted , as their qualitative characteristic features remain similar to one of the three plots in ( a ) , ( b ) , ( c ) .as one moves along the h@xmath1-hf - no series , the plot for a given @xmath15 series tends to vary rather slowly ( rate of increase slows down ) , with h@xmath1 and no showing maximum and minimum increase respectively , such that for no the plots are quite flat . also for a given molecule , as one goes to higher @xmath15 values , the separation between two successive @xmath15 plots tends to decrease .similar qualitative feature has been recorded earlier in the energy versus @xmath16 plot for h@xmath1 within a semiclassical approach @xcite .now we turn to the e@xmath40d@xmath41 ( in ev ) versus @xmath15 for fixed @xmath16 quantum number for the same three molecules in the bottom panels ( d ) , ( e ) and ( f ) respectively .the @xmath15 axis in h@xmath1 , hf , no extends to 25 , 30 and 40 . for h@xmath1 ,these are studied at seven @xmath16 , _ viz . , _ 0,5,10,15,20,25,30 , while for the other two molecules two more @xmath16 values of 35 , 40 are considered besides the seven value of h@xmath1 . in going from h@xmath1-hf - no ,the plots for different @xmath16 become progressively more closely spaced , with h@xmath1 showing maximum sparsity and in no the successive separations are too small to be identified properly in the scale .another interesting feature is that , rate of increase in energy slowly increases as one moves the series , producing a nearly linear structure in no . also for h@xmath1 ,hf , as @xmath16 takes higher values , the separation between to adjacent @xmath16 tends to grow large . in both the @xmath16 and @xmath15 plots in top and bottom panels , individual @xmath15 and @xmath16series for a given molecule remain nearly parallel to each other . )( top panel ) and vibrational ( @xmath15 ) ( bottom panel ) quantum numbers respectively .the plots ( a ) , ( d ) correspond to h@xmath1 , while ( b ) , ( e ) and ( c ) , ( f ) correspond to hf and no molecule respectively .state indices are indicated in the figure .see text for details . ] ) ( top panel ) and vibrational ( @xmath15 ) ( bottom panel ) quantum numbers respectively .the plots ( a ) , ( d ) correspond to h@xmath1 , while ( b ) , ( e ) and ( c ) , ( f ) correspond to hf and no molecule respectively .state indices are indicated in the figure .see text for details . ] ) ( top panel ) and vibrational ( @xmath15 ) ( bottom panel ) quantum numbers respectively .the plots ( a ) , ( d ) correspond to h@xmath1 , while ( b ) , ( e ) and ( c ) , ( f ) correspond to hf and no molecule respectively .state indices are indicated in the figure .see text for details . ]+ ) ( top panel ) and vibrational ( @xmath15 ) ( bottom panel ) quantum numbers respectively .the plots ( a ) , ( d ) correspond to h@xmath1 , while ( b ) , ( e ) and ( c ) , ( f ) correspond to hf and no molecule respectively .state indices are indicated in the figure .see text for details . ] ) ( top panel ) and vibrational ( @xmath15 ) ( bottom panel ) quantum numbers respectively .the plots ( a ) , ( d ) correspond to h@xmath1 , while ( b ) , ( e ) and ( c ) , ( f ) correspond to hf and no molecule respectively .state indices are indicated in the figure .see text for details . ] ) ( top panel ) and vibrational ( @xmath15 ) ( bottom panel ) quantum numbers respectively .the plots ( a ) , ( d ) correspond to h@xmath1 , while ( b ) , ( e ) and ( c ) , ( f ) correspond to hf and no molecule respectively .state indices are indicated in the figure .see text for details . ]tietz - hua oscillator has been found to be a more realistic analytical potential than the familiar morse potential in describing molecular dynamics at moderate as well as high rotational and vibrational quantum number . in the present work ,we have presented both @xmath47wave and rotational bound states having arbitrary rotational and vibrational quantum numbers with excellent accuracy .a total of 18 low and moderately high - lying ro - vibrational levels are given for six diatomic molecules , namely , h@xmath1 , hf , n@xmath1 , no , o@xmath1 , o@xmath2 . while the lower states match quite well the lone literature result , many new states are given here for the first time .energy changes with respect to @xmath3 quantum numbers are discussed in detail for three molecules . in short ,a simple accurate and efficient scheme is offered for this and other similar potentials in molecular physics .99 p. m. morse , phys .34 * , 57 ( 1929 ) .r. rydberg , z. phys .* 73 * , 376 ( 1931 ) .n. rosen and p. m. morse , phys .42 * , 210 ( 1932 ) .g. pschl and e. teller , z. phys .* 83 * , 143 ( 1933 ) .m. f. manning and n. rosen , phys .rev . * 44 * , 953 ( 1933 ) .e. r. lippincott , j. chem .* 21 * , 2070 ( 1953 ) .a. a. frost and b. muslin , j. chem .* 22 * , 1017 ( 1954 ) .t. tietz , j. chem .* 38 * , 3036 ( 1963 ) .j. n. murrell and k. s. sorbie , j. chem . soc .faraday trans . *70 * , 1552 ( 1974 ) .d. schiberg , mol .phys . *59 * , 1123 ( 1986 ) .w. hua , phys .a * 42 * , 2524 ( 1990 ) .a. n. zavitsas , j. am .soc . * 113 * , 4755 ( 1991 ) . f. m. rafi , phys .a * 205 * , 383 ( 1995 ) .d. o. n. gardner and l. v. szentply , j. phys .a * 103 * , 9313 ( 1999 ) .p. g. hajigeorgiou and r. j. le roy , j. chem .phys . * 112 * , 3949 ( 2000 ) .s. m. ikhdair and r. sever , j. mol .struct .( theochem ) * 806 * , 155 ( 2007 ) .s. m. ikhdair and r. sever , j. mol .struct .( theochem ) * 855 * , 13 ( 2008 ) .g. hajigeorgiou , j. mol .spectrosc . * 263 * , 101 ( 2010 ) .r. t. pack , j. chem .phys . * 57 * , 4612 ( 1972 ) .e. levin , h. partridge , j. r. stallcop , j. thermophys .heat transfer * 4 * , 469 ( 1990 ) .g. a. natanson , phys .a * 44 * , 3377 ( 1991 ) .j. a. kunc , f. j. gordillo - vzquez , j. phys .chem . a * 101 * , 1595 ( 1997 ) .f. j. gordillo - vzquez and j. a. kunc , j. mol .struct .( theochem ) * 425 * , 263 ( 1998 ) .m. hamzavi , a. a. rajabi and k .- e .thylwe , int .j. quant .chem . * 112 * , 2701 ( 2012 ) .s. m. ikhdair and m. hamzavi , few - body syst . * 53 * , 473 ( 2012 ) .a. k. roy , j. phys .g * 30 * , 269 ( 2004 ) .a. k. roy , int .j. quant .chem . * 104 * , 861 ( 2005 ) .a. k. roy , pramana j. phys . * 65 * , 01 ( 2005 ) .a. k. roy , a. f. jalbout and e. i. proynov , j. math .chem . * 44 * , 260 ( 2008 ) .a. k. roy , int .j. quant .chem . * 113 * , 1503 ( 2013 ) .a. k. roy , results in physics .* 3 * , 103 ( 2013 ) .the nist reference on constants , units and uncertainty , physics.nist.gov/cuu/constants/index.html .d. r. t. appadoo , r. j. le roy , p. f. bernath , s. gerstenkorn , p. luc , j. verges , j. sinzelle , j. chevillard and y. j. daignaux , j. chem . phys .* 104 * , 903 ( 1996 ) .

<S> accurate low and high - lying bound states of tietz - hua oscillator potential are presented . </S><S> the radial schrdinger equation is solved efficiently by means of the generalized pseudospectral method that enables optimal spatial discretization . </S><S> both @xmath0 and rotational states are considered . </S><S> ro - vibrational levels of six diatomic molecules _ viz . _ , h@xmath1 , hf , </S><S> n@xmath1 , no , o@xmath1 , o@xmath2 are obtained with good accuracy . </S><S> most of the states are reported here for the first time . </S><S> a detailed analysis of variation of eigenvalues with @xmath3 quantum numbers is made . </S><S> results are compared with literature data , wherever possible . </S><S> these are also briefly contrasted with the morse potential results . </S>

1308.3685

the r - mode instability was first discovered in @xcite and it was predicted that the instability could become a significant source of gravitational radiation .this happens because the r - modes are generically unstable to the cfs gravitational - radiation - driven instability @xcite .the r - mode instability follows immeadiately from the fact that r - modes that are prograde with respect to a distant observer are retrograde in the comoving frame for all values of the angular velocity ( the canonical energy of the modes is negative ) . for some nice reviewssee @xcite . however , different mechanisms to damp the instability have to be considered : one of them is viscosity @xcite , another one could be sufficiently strong magnetic fields @xcite . the instability windows for non - magnetized newtoninan stars were initially calculated in @xcite .more recently , the effect of magnetic fields on r - modes was discussed in @xcite for a spherical shell and in @xcite for a neutron star with purely toroidal field , in all cases in the newtonian context .this paper is a first part of a project that is supposed to contribute to the understanding of the instability window for slowly rotating relativistic neutron stars with magnetic fields .we are here interested in and focused on the modification of the r - modes in the presence of magnetic fields and its effects on the gravitational wave emission. the effect of magnetic fields on the r - mode frequencies could be interesting for astrophysical objects such as magnetars , that have magnetic fields of the order of magnitude @xmath2 g and are very slow rotators with rotational periods of a few seconds .we consider here stars of comparable magnetic fields with rotational periods of a few milliseconds ( due to a numerical difficulty : longer periods would need longer time evolutions ) .the issue of viscous damping of the mode , which determines the instability window of the r - mode , is further complicated by the presence of the magnetic field .therefore we leave it for future work .we treat the problem within the realm of perturbation theory , first by deriving general perturbation equations and then by solving them numerically with a 2d lax - wendroff method .the same numerical methods were used and tested in our previous paper @xcite .the advantages of the 2d dynamical evolution in this case is that it avoids both the r - mode continuous spectrum problem @xcite and the need for truncating the solution at some @xmath9 ( as done for instance in @xcite for torsional modes of a relativistic star with a dipole magnetic field ) .we first calculate the r - mode frequencies for different values of the rotation parameter and magnetic fields .then we calculate the instability growth rate due to the emission of gravitational waves as a function of the magnetic field .( this gives both general relativistic and magnetic field corrections to the results of @xcite . )the paper is organized as follows : in the second section we describe our background model and in the third section we present the full perturbation equations of our model .this section is followed by the fourth section , where we compute the r - mode frequencies for different rotation rates of the star , as a function of the magnetic field . in the fifth sectionwe compute the growth time due to the r - mode gravitational wave emission as a function of magnetic field . in the sixth section we present the concluding remarks .( everywhere in the paper , unless explicitly mentioned , we use the units @xmath10 . )we work with a slowly and uniformly rotating magnetized star with a polytropic equation of state .our model neutron star has @xmath11 , @xmath12 km , and the pressure @xmath13 is given by the polytropic equation of state @xmath14 , taken with the parameters @xmath15 , @xmath16 and @xmath17 is the rest - mass density of the star . the keplerian frequency @xmath18 ( mass shedding limit ) that we use in our paper to normalize the rotation of the star is @xmath19 .the effect of the rotation is taken up to the linear order in the rotation parameter @xmath20 , which means one considers the effect of the rotation on the spacetime metric ( frame dragging function ) , but neglects the effect of the rotation on the stellar structure .( the deformations of the stellar fluid due to rotation are of the order @xmath21 . )we consider a dipole magnetic field and for the realistic neutron stars with magnetic fields ( up to the order of @xmath22 for magnetars ) , one can neglect the effect of the magnetic field on both the stellar structure and the background metric ( for a more detailed argumentation , see @xcite ) .this means the model follows from a line element : @xmath23^{2},\ ] ] where @xmath24 , @xmath25 and @xmath26 are functions of @xmath27 , and a stress energy tensor given by : @xmath28 here @xmath29 is the total energy density , the 4-velocity @xmath30 is given as @xmath31 and @xmath32 is related to the magnetic field @xmath33 , defined as usual in terms of the electromagnetic tensor @xmath34 as @xmath35 our background model is thus obtained by the tov equations supplemented by a polytropic equation of state , hartle s equation for the frame dragging function @xmath26 @xcite : @xmath36 and the magnetic field is given by the maxwell equations . therefore our dipole equilibrium magnetic field ( for our choice of equilibrium magnetic field, the induction equation is trivially fulfilled ) obeys the maxwell equations @xmath37 solved with a 4-current @xmath38 with the only non - zero component @xmath39 ( this choice is equivalent , for instance , to choosing the terms with parity @xmath40 in the parity decomposition given in @xcite and keeping only @xmath41 , @xmath42 ) with the radial profile @xmath43 that describes a ring current inside the star .moreover , this current profile allows us to consider the magnetic field as force - free at the surface ( the lorentz force goes to zero at the surface of the star as @xmath44 ) . choosing the vector potential as @xmath45 ( same parity and symmetry choices as we did for @xmath46 above ) ,the maxwell equations give : @xmath47a_{,r } - \frac{2}{r^2}a + \alpha = 0.\ ] ] here the components of our dipole magnetic field are obtained from @xmath48 as : @xmath49 @xmath50 it is known that the equation has outside the star an exact analytic solution @xcite : @xmath51 .\label{eq : a_exact}\ ] ] the full solution inside and outside the star for the dipole magnetic field was computed by numerically solving eq .( [ eq : a_gen ] ) inside the star , matching the regular series expansion of the solution near the center with the numerical solution up to the surface , where we require continuity of @xmath48 and its first derivative with the exterior analytic solution ( [ eq : a_exact ] ) .some representative plots of the behavior of the radial function @xmath48 can be seen in figure [ fig : teste_a ] , for increasing values of the current parameter @xmath52 .one typical solution for the amplitude of the magnetic field inside the star is given in figure [ fig : teste_b2 ] ( we note here that the solid lines are contour lines , and not magnetic field lines ) . throughout the paper we refer to the value of the magnetic field at the center of the star .but since the magnetic field at the pole of a star can be determined observationally , while the value of the magnetic field at the center of star must be calculated with some model , we present in figure [ fig : teste_bpole ] the relation between @xmath53 and @xmath54 given by our model . in figure[ fig : teste_omee ] we present some representative plots of the numerical solutions obtained for the frame dragging function @xmath26 , from hartle s equation ( [ eq : hartle ] ) . for different currents as a function of the radial coordinate divided by the radius of the star .the magnetic field is shown in the plane given by the rotational axis and a perpendicular direction to the rotational axis . ] as a funtion of the magnetic field at the center of the star @xmath55 as calculated with our model .we obtained @xmath56 . ] as a function of the radial coordinate normalized by the radius of the star .( the frame dragging function up to the linear order in @xmath20 is for a uniformly rotating star only a function of @xmath27 and outside the star behaves as @xmath57 @xcite . ) ]we are working in the cowling approximation and considering only barotropic perturbations , therefore the fundamental set of perturbation variables is given by @xmath58 .( for an analysis of the accuracy of the cowling approximation , but only for f and p modes , see @xcite . for r - modes , the cowling approximation gives more accurate frequencies , as these modes do not involve large density variations @xcite . )the perturbation equations are obtained by perturbing the euler equations ( @xmath59 ) : @xmath60 and the energy conservation equation @xmath61 together with the perturbed induction equations @xmath62 furthermore one uses as constraints the perturbed ideal mhd equation : @xmath63 and the 4-velocity normalization condition @xmath64 together with the fact that the magnetic field remains perpendicular to the 4-velocity @xmath65 the last two equations can be subsequently used to reduce the variables to the 7 independent variables @xmath66 , with @xmath59 .there are 3 independent components of the perturbed induction equation which turn into : * the @xmath27 component : @xmath67 * the @xmath68 component : + @xmath69 * the @xmath70 component : + @xmath71 the independent components of the perturbed euler equation : the supplementary three constraints are the following : let us mention that the upper equations reduce for the special case of @xmath79 to the equations shown in @xcite . for the numerical integration with the 2d lax - wendroff scheme we need to obtain the dynamical equations in a form containing only one time derivative in each equation .this is most convenient to achieve by proper linear combinations of the original equations and ommiting the @xmath80 terms .the 7 independent variables @xmath81 , ( @xmath59 ) , are further transformed into `` momentum - like '' variables @xmath82 ( for the introduction of these variables in the newtonian context see @xcite . )this transformation is done for the purpose of obtaining a simple boundary condition at the stellar surface as : @xmath83 furthermore we apply regularity conditions at the center of the star and at the rotational axis , together with the correct symmetry conditions at the equatorial plane .( for the details about these conditions see our previous work @xcite . )another constraint that has to be fulfilled is the time independent mhd equation , that is checked to be satisfied in each step of the calculation ( up to certain determined numerical error ) .the time independent mhd constraint is obtained from the perturbed ideal mhd equation by subtracting the appropriate multiple of the @xmath70 component of the perturbed induction equation .the constraint reads : our integration domain occupies only the first quadrant , since we take advantage of the symmetries at the equatorial plane .our numerical grid typically has @xmath85 points in @xmath86 , where @xmath27 varies in @xmath87 $ ] and @xmath68 in @xmath88 $ ] .we take usually 10.000 - 50.000 time steps in the evolution of the equations , depending on the rotating rate and , consequently , the frequency of the r - mode . in each time evolutionwe observe at least several periods of oscillation of the perturbations .we point out here that our time evolutions were so far stable , and we did not see signs of the hydromagnetic instability observed in @xcite for axial - led perturbations in newtonian gravity .the investigation of this issue in our relativistic treatment is left for a future work .we limited the maximum rotation rate considered here by @xmath89 , motivated by the results of @xcite , where they see corrections of the order @xmath90 in the r - mode frequencies for @xmath91 .we also limited our minimum rotation rate at @xmath92 because of numerical reasons , as already stated in the introduction ( lower rotation rates would demand longer simulations ) . in the next sectionwe discuss the numerical limits on the magnetic field .the final equations obtained via the linear combinations and redefinitions of variables can be found in the appendix [ a1 ] . in appendix[ a2 ] we present simplified equations obtained from the equations in appendix [ a1 ] by neglecting the coefficients of the order @xmath93 and @xmath94 .these simplified equations can be useful for sufficiently weak magnetic fields and sufficiently slow rotation rates , where neglecting the @xmath93 and @xmath94 terms could be justified .we used also these equations to compute the r - mode frequencies and the results are compared in the figure [ fig : rmode1 ] .the r - modes ( @xmath95 ) were computed using the equations - .we solve the system of equations with a 2d lax - wendroff scheme with non - constant coefficients @xcite .we refer the reader to a previous work @xcite for further details on the numerical setup used for obtaining the r - mode frequency and eigenfunction .( in @xcite it was used for non - magnetized and differentially rotating stars . )we calculated the r - modes first for zero magnetic field .the dependence of the r - mode frequencies on the rotation parameter for the non - magnetized field case is shown in figure [ fig : rmode0 ] .we compared results with @xcite for the star with @xmath96 , and found that our results match with less than 3% error .( for more results on r - modes of non - magnetized stars see also the papers @xcite . ) in figures [ fig : rmode1 ] and [ fig : rmode2 ] we present r - modes as a function of magnetic field . for comparison we present in the figure [ fig : rmode1 ] also r - modes calculated via the simplified equations from the appendix [ a2 ] for the star with @xmath96 .the approximation of the simplified equations is shown to break down in this case at the value of magnetic field around @xmath97 g , while the results obtained from the full equations _ seem _ to breakup at a larger magnetic field around @xmath98 g. for larger magnetic fields than @xmath99 g ( where we do not entirely trust our results ) , we still see that the r - mode disappears completely due to the growth of another mode ( possibly an alfvn mode ) .we believe that the breakdown in the behavior of the r - mode frequencies is caused by the growth of this other mode and the subsequent deformation of the r - mode .this is consistent with the expectations based on the results of @xcite .as can be seen in the figure [ fig : rmode2 ] , the r - mode frequencies change very little when one turns on the magnetic field .this is consistent with the observation of @xcite for newtonian stars .the change of the frequencies is more pronounced for smaller values of the rotation parameter . for @xmath100 and magnetic field @xmath101g , the r - mode frequency changes by a little less than 4% . in case of larger rotation@xmath102 the same value of magnetic field changes the r - mode frequency by 1% . even though the variations are small , one can still clearly observe from fig.[fig : rmode2 ] the behavior of the frequencies with respect to the increasing magnetic field . such a behavior seems to have remarkable features : the r - mode frequencies behave for sufficiently large @xmath20 ( @xmath103 ) as @xmath104 , whereas for a smaller value of @xmath20 , ( @xmath105 ) , the behavior of the frequencies with respect to the magnetic field is given as @xmath106 .let us note that the @xmath106 dependence was observed for the r - modes of the spherical shell in @xcite . in figuresand we show the plots of the r - mode eigenfunctions for all the variables .( the eigenfunctions are shown in the plane given by the rotational axis and a perpendicular axis to the rotation . )the complicated interplay between r - modes and magnetic fields is more visible in the @xmath107 eigenfunction , where we can see a sort of double peak .this happens for all rotation rates , and it is more pronounced for larger magnetic fields .we believe that this shows the deformation of the r - mode eigenfunction caused by other modes excited for large enough magnetic fields . for more details on the procedure used for extracting the eigenfunctions , see again @xcite . ) fits . ]the r - mode instability growth times ( for @xmath95 ) are calculated by using the usual quadrupole formula ( for the details see for example @xcite ) .the characteristic timescale is calculated from the equation : @xmath108 with @xmath109 , @xmath59 .the energy time derivative is calculated from the quadrupole formula as : @xmath110 with @xmath111^{2}},\ ] ] where @xmath112 and @xmath113 are the mass and the current multipoles defined as in @xcite .( we use both mass and current multipoles , but because the r - modes involve only a perturbed velocity field , to the lowest order in @xmath20 , gravitational radiation through current multipoles dominates over that produced by mass multipoles @xcite . ) in particular the multipoles can be expressed as : @xmath114 and @xmath115 ( here @xmath116 and @xmath117 are the multipoles defined in @xcite . )we were able to fit the function @xmath1 for zero magnetic fields as a functions of the rotation period @xmath118 as @xmath119 with the dimensionless parameters @xmath120 and @xmath121 taking the values @xmath122 and @xmath123 .we compare our values for @xmath120 and @xmath121 with the values obtained in @xcite in table [ tab : par ] ( see also figure [ fig : logtau ] ) ..the @xmath120 and @xmath121 parameters for the case of zero magnetic fields .we are comparing our results with @xcite where the calculations were done for the newtonian polytropes with stellar parameters close to ours .we obtained 27 - 34 % faster emission of gravitational waves compared to the newtonian setting . [ cols="^,^,^,^",options="header " , ] the instability growth time scale @xmath1 relative change due to the magnetic field is shown in figure .we can see that the relative change becomes positive for lower values of magnetic fields ( increasing the growth time and slowing down the emission of gravitational waves ) and then becomes negative for larger values of the magnetic field ( with the opposite effect ) , causing a relative change of up to @xmath3 5% .similarly to the r - mode frequencies , the relative effect of the magnetic field is more pronounced for the lower rotation rates . however , to estimate the amount of gravitational waves emitted and the window of the instability we would need to calculate the viscosity damping rates and that is left for future work .we presented here a model for a rotating magnetized star , in which we neglect the distortion of both the geometry and the fluid by a _ dipolar _ magnetic field .we derived the full perturbation equations in the cowling approximation . after solving the 2d time evolution problem with a lax - wendroff method , we computed the r - mode frequencies using the fourier spectrum of the solution and we were able to extract the eigenfunctions of the perturbations . the frequencies and eigenfunctions of the r - modewere then used to calculate the growth time scale due to gravitational radiation , using the newtonian quadrupole formalism .we found that the effect of the magnetic field on the frequencies is very small , ( up to 5% for the lowest rotation rates ) . for lower rotation rates the frequencies follow a @xmath124 dependence and , for higher rotation rates , a @xmath125 dependence .the effect on the r - mode growth time @xmath1 indicates a faster emission of gravitational waves , compared to the newtonian non - magnetized calculations of @xcite .we found that @xmath1 is more significantly affected by the presence of general relativity ( @xmath126% ) and less significantly by the presence of magnetic field ( up to @xmath127% ) .our results indicate that the relative effects of the magnetic field are more pronounced for more slowly rotating stars. therefore it could be possible that they achieve higher values for magnetars , that have rotation periods about 1000 times larger than the ones considered here ( due to numerical limitations ) .however , it is not trivial to estimate how large these corrections would be , given the complicated dependence on both rotation and magnetic field of the solutions .the effect of viscosity will play of course a key role in determining the actual instability window and is left for the future work .a more realistic description of the star would also need to include work with realistic equations of state and a stellar crust , together with considering the backreaction from the production of toroidal magnetic field @xcite .this is also left for the future .the authors are especially thankful to luciano rezzolla and shinichirou yoshida for many useful discussions in different stages of this project .this work was supported by fapesp and the max planck society .the final dynamical equations for the numerical evolution are given as :@xmath136\delta\tilde u^{\theta}+h^{r}(\delta\tilde u^{\phi}_{,\phi}+\delta\tilde u^{\theta}_{,\theta})-(h^{\theta}_{,\theta}+\cot(\theta)h^{\theta})\delta\tilde u^{r}-h^{\theta}\delta\tilde u^{r}_{,\theta}=0.~~~~~~~~~~~~~~~~~~~~~~~~~~~ \end{aligned}\ ] ] @xmath137\delta\tilde u^{r}-~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~\\ \left[h^{r}_{,r}+h^{r}\left(\nu_{,r}\left\{\frac{p+\epsilon}{2\gamma p}+1\right\}+\frac{\lambda_{,r}}{2}+\frac{2}{r}\right)\right]\delta\tilde u^{\theta}-h^{r}\delta\tilde u^{\theta}_{,r}=0.~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~\nonumber \end{aligned}\ ] ] @xmath138 \delta\tilde u^{\phi}_{,\phi } + ~~~~~~~~~~~~~~~~~~~~~\nonumber\\ + \left [ -(\epsilon+p)\omega_{,r}+(\omega-\omega ) \left\ { ( \epsilon+p ) \left ( \frac{2}{r}-\nu_{,r } \right ) - \left ( \frac{\lambda_{,r}}{2}+\nu_{,r}+\frac{2}{r } \right ) \gamma p \right\ } \right ] \delta\tilde u^{r}+~~~~~~~~~~~~\\ + ( \omega-\omega ) \{2(\epsilon+p)-\gamma p\}\cot(\theta ) \delta\tilde u^{\theta}- ( \omega-\omega ) \gamma p \delta\tilde u^{\theta}_{,\theta}- ( \omega-\omega ) \gamma p \delta\tilde u^{r}_{,r}+ ( \epsilon+p)\delta\tilde u^{\phi}_{,t}+e^{\nu/2 } \left [ \frac{\epsilon+p}{r^{2}\sin^{2}(\theta)}\delta p_{,\phi}- \right .h^{\phi}_{,r}-h^{\theta}\delta\tilde h^{\phi}_{,\theta}- \left .\left [ \left ( \frac{\nu_{,r}}{2}+\frac{2}{r } \right ) h^{r}+2\cot(\theta)h^{\theta } \right ] \delta\tilde h^{\phi}+\frac{e^{\lambda}h^{r}}{r^{2}\sin^{2}(\theta)}\delta\tilde h^{r}_{,\phi}+\frac{h^{\theta}}{\sin^{2}(\theta)}\delta\tilde h^{\theta}_{,\phi } \right ] = 0.~~~~~~\nonumber\end{aligned}\ ] ] @xmath139 \delta p+~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~\nonumber\\ + ( \epsilon+p)e^{-\lambda}\delta p_{,r}-h^{r}(\delta\tilde h^{r}_{,r}+\delta\tilde h^{\phi}_{,\phi}+\delta\tilde h^{\theta}_{,\theta})-h^{\theta}\delta\tilde h^{r}_{,\theta}+~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~\\ + \left [ e^{-\lambda}r^{2}h^{\theta } \left ( \frac{\nu_{,r}}{2}+\frac{2}{r } \right ) -\cot(\theta)h^{r}-r^{2}h^{\theta}e^{-\lambda}\frac{\nu_{,r}}{2}\left\{\frac{\epsilon+p}{\gamma p}+1\right\ } \right ] \delta\tilde h^{\theta}+~~~~~~~~~~~~~~~~~~~~~~~~~\nonumber\\ + h^{\theta}e^{-\lambda}r^{2}\delta\tilde h^{\theta}_{,r}-h^{r } \left ( \frac{\lambda_{,r}}{2}+\frac{\nu_{,r}}{2}\left\{\frac{\epsilon+p}{\gamma p}+1\right\}+\frac{2}{r } \right ) \delta\tilde h^{r}=0~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~\nonumber\end{aligned}\ ] ] @xmath140 \delta\tilde h^{\theta}-~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~\nonumber\\ h^{r}\delta\tilde h^{\theta}_{,r}-h^{\theta } \left [ \frac{\lambda_{,r}}{2}+\frac{\nu_{,r}}{2}\left\{\frac{p+\epsilon}{\gamma p}+1\right\}+\frac{2}{r } \right ] \delta\tilde h^{r}+\frac{\epsilon+p}{r^{2}}\delta p_{,\theta}=0~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~\nonumber\end{aligned}\ ] ] l. rezzolla , f.k .lamb , d. markovic and s. shapiro , `` properties of r modes in rotating magnetic neutron stars .1 . kinematic secular effects and magnetic evolution equations '' , phys . rev .d , * 64 * 104013 ( 2001 ) , arxiv : gr - qc/0107061 l. rezzolla , f.k .lamb , d. markovic and s. shapiro , `` properties of r modes in rotating magnetic neutron stars .evolution of the r modes and stellar magnetic field '' , phys .d , * 64 * 104014 ( 2001 ) , arxiv : gr - qc/0107062 c. chirenti , j. skakala and s. i .yoshida , `` slowly rotating neutron stars with small differential rotation : equilibrium models and oscillations in the cowling approximation , '' phys .d * 87 * , 044043 ( 2013 ) , arxiv:1301.3111 [ gr - qc ] .s. yoshida , s. yoshida and y. eriguchi , `` r - mode oscillations of rapidly rotating barotropic stars in general relativity : analysis by the relativistic cowling approximation '' , mon.not.roy.astron.soc .* 356 * ( 2005 ) 217 - 224 , arxiv : astro - ph/0406283 c. cuofano , s. dallosso , a. drago and l. stella , `` generation of strong magnetic fields by r - modes in millisecond accreting neutron stars : induced deformations and gravitational wave emission '' , phys.rev .d * 86 * ( 2012 ) 044004 , arxiv : astro - ph/1203.0891

<S> we study here the r - modes in the cowling approximation of a slowly rotating and magnetized neutron star with a poloidal magnetic field , where we neglect any deformations of the spherical symmetry of the star . </S><S> we were able to quantify the influence of the magnetic field in both the oscillation frequency @xmath0 of the r - modes and the growth time @xmath1 of the gravitational radiation emission . </S><S> we conclude that magnetic fields of the order @xmath2 g at the center of the star are necessary to produce any changes . our results for @xmath0 show a decrease of up to @xmath3 5% in the frequency with increasing magnetic field , with a @xmath4 dependence for rotation rates @xmath5 and @xmath6 for @xmath7 . </S><S> ( these results should be trusted only within slow rotation approximation and we kept @xmath8 . ) for @xmath1 , we find that it is approximately 30% smaller than previous newtonian results for non - magnetized stars , which would mean a faster growth of the emission of gravitational radiation . </S><S> the effect of the magnetic field in @xmath1 causes a non - monotonic effect , that first slightly increases @xmath1 and then decreases it further by another @xmath3 5% . </S><S> ( the value of magnetic field for which @xmath1 starts to decrease depends on the rotational frequency , but it is generally around @xmath2 g . ) future work should be dedicated to the study of the effect of viscosity in the presence of magnetic fields , in order to establish the magnetic correction to the instability window . </S>

physics0512241

noise at the dynamical and measurement level often encourages the choice of stochastic interpretation of time series data . detrended fluctuation analysis ( dfa ) @xcite and its extensions @xcitehave been used widely to determine the nature of correlations from stochastic processes obtained from a wide - range of systems @xcite .these studies have subsequently suggested the choice of stochastic models to explain complex behavior observed in synthetic and natural data sets @xcite .the objective of the present study is to understand the scaling behavior of the first - order stochastic processes and their volatility series @xcite as function of the process parameter .based on the results obtained for the various parameter regime , we evaluate the choice of these processes in modeling real - world temperature data @xcite .classical ar ( 1 ) is given by the expression @xmath1 where @xmath2 represents independent and identically distributed ( i.i.d ) gaussian innovations with zero mean and unit variance .the corresponding volatility series of ( 1 ) is given by @xmath3 , where @xmath4 represents the absolute value of @xmath5 .correlations in @xmath6 imply that clusters of ( small ) big changes are likely to be followed by clusters of ( small ) big changes .it can be shown analytically that eqn .( 1 ) is stationary when @xmath7 i.e. @xmath8 .the autocorrelation function ( acf ) of ( 1 ) decays exponentially and is of the form @xmath9 for lag @xmath10 .the nature of the decay is governed by the process parameter @xmath0 .while @xmath11 exhibits an exponential decay for ( @xmath12 ) , it exhibits an exponential decay with oscillations for ( @xmath13 ) .it is important to note that @xmath14 and @xmath15 represent cases where the current sample is negatively and positively correlated with its immediate past , respectively .the strength of the short - term correlated and short - term anti - correlated behavior is dictated by the process parameter @xmath0 .we also consider the case where @xmath2 in ( 1 ) is generated by a long - range correlated gaussian process @xcite . this extended ar ( 1 ) with long - range correlated gaussian innovations shall be referred to as car ( 1 ) in the subsequent sections .the choice of the term car was encouraged by recent studies @xcite . in the light of recent findings@xcite , we critically evaluate the choice of car ( 1 ) for modeling real - world daily temperature records .a complete description of dfa and its extensions can be found elsewhere @xcite . in the present study ,the length of the data sets is chosen sufficiently larger @xmath16 , in order to avoid bias in the estimation procedure due to finite sample - size effects .dfa with fourth order polynomial detrending is used in order to minimize bias due to local polynomial trends @xcite .we also integrate the data prior to dfa estimation and scale the resulting fluctuation function by the window size @xcite . the fluctuation function of uncorrelated noise @xmath17 and the long - range correlated noise @xmath18 shall be used as reference in the subsequent discussions . as noted earlier ( sec .2 ) ar ( 1 ) process is stationary for @xmath8 . in the present study, we consider parameters @xmath19 and @xmath20 . while the former is close to the uncorrelated regime @xmath21 , the latter is close to the non - stationary regimes , @xmath22 .log - log plots of the fluctuation function versus time scale for ar ( 1 ) and its volatility series with @xmath23 , fig .1a , is parallel to exponent 0.5 in the asymptotic regime , with a slight deviation from 0.5 in the short time scales for @xmath24 .a similar analysis was carried for process @xmath25 fig .unlike @xmath26 whose fluctuation function is relatively homogenous , characteristic crossovers are observed for @xmath27 and their volatility series , fig .1b . for @xmath28 , the log - log plot of the ar ( 1 ) process displayed correlated behavior at short - time scales and uncorrelated behavior at the larger time scales @xmath29 .however , for @xmath30 , one observes a characteristic crossover from anti - correlated behavior for time scales @xmath31 to uncorrelated behavior for @xmath32 .the volatility series for @xmath28 was homogenous and exhibited an exponent of @xmath33 characteristic of monofractal data @xcite , whereas that of @xmath30 , displayed a marked crossover from correlated to uncorrelated behavior ._ from the above discussion , it is clear that the process parameter @xmath0 has a significance impact on the scaling behavior at shorter time scales .this is reflected by marked distortions and cross - overs in the fluctuation plots .it can also be noted that irrespective of the choice of @xmath0 , the scaling in the asymptotic regime resembles that of an uncorrelated noise , @xmath17_. long - range correlated gaussian innovations @xmath2 with scaling exponent of ( @xmath34 ) @xcite was used to generate car ( 1 ) .the fluctuation function of car ( 1 ) process for @xmath35 were parallel to each other and to @xmath2 ( @xmath34 ) , see fig .the corresponding volatility series displayed an exponent of 0.5 characteristic of monofractal data .thus the scaling behavior of car ( 1 ) is similar to that of ar ( 1 ) for process parameter @xmath0 close to zero . for @xmath30 ,at short time scale @xmath36 , car ( 1 ) is highly anti - correlated however the behavior mimicked that of @xmath2 ( @xmath34 ) for @xmath37 , fig .2b . for @xmath38 ,short time scales had an exponent @xmath39 close to the integrated version of @xmath2 ( @xmath34 ) .however , in the asymptotic regime the scaling exponent was that of @xmath2 ( @xmath34 ) , fig .fluctuation function obtained for the volatility series ( @xmath38 ) was parallel to the reference line with slope 0.5 characteristic of monofractal data .however , the fluctuation function of the volatility series ( @xmath40 ) displayed correlated behavior at short time scales @xmath41 and uncorrelated behavior @xmath42 in the asymptotic regime .as in the case of ar ( 1 ) , the process parameter @xmath0 has a significant impact on the scaling behavior of car ( 1 ) process , revealed by marked distortion and crossovers in the log - log plots .[ [ critical - note - on - the - choice - of - car-1-to - model - temperature - correlations ] ] critical note on the choice of car ( 1 ) to model temperature correlations ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ it has been shown @xcite that atmospheric temperature data ( from a randomly chosen 14 continental regions ) display power correlations with dfa exponent @xmath43 and this is considered as universal persistence behavior of the temperature data . using this result as a benchmark ,the quality of the temperature data simulated by the global climate models has been evaluated @xcite . in a later study @xciteit has been shown that temperature data from continental and costal zones display lesser deviations from the previously observed universal behavior ( exponent ) of 0.65 while the islands display greater variability with the average exponent being close to 0.8 @xcite . in a recent study , @xcite car(1 ) process was proposed as a plausible model for explaining correlations in atmospheric daily temperature fluctuations collected from fourteen meteorological centers in hungary .the choice of the car ( 1 ) model was inspired by the fact that temperature records exhibit both short - term and long - term correlations @xcite .we generated the data using the car ( 1 ) model proposed by in @xcite @xmath44 the model parameters were chosen as @xmath45 , @xmath46 is zero - mean , unit variance long - range correlated gaussian innovations with scaling exponent @xmath43 , @xmath47 as in @xcite .log - log plot of the fluctuation function of eqn .2 , fig . 3 , displayed strong correlations with exponent ( @xmath48 ) at short scales @xmath49 .however , in the asymptotic regime it displayed an exponent @xmath50 close to that of the long - range correlated gaussian innovations .fluctuation function of the corresponding volatility series was parallel to @xmath51 , characteristic of monofractal processes .this has to be contrasted with recent studies @xcite , which provided compelling evidence of long - range correlations @xmath52 in the volatility series of daily temperature fluctuations .such long - range volatile correlations were observed across randomly chosen and spatially separated meteorological sites across the world . this in turn was attributed to universality in daily temperature records @xcite . it has been shown that long range correlated data with exponents @xmath53 will exhibit uncorrelated behavior in volatility analysis while the multifractal signals exhibit long range correlations in volatility analysis @xcite . as the car ( 1 ) model , eqn .2 , is driven by a linear process ( see above ) , one would expect the model to display uncorrelated behavior in the volatility analysis and is indeed the case obtained for this model . since , the car ( 1 ) model eqn.2 @xcite captures the linear features of the temperature data ( as the volatility exponent is 0.5 for this model ) , it will be unable to account for the nonlinear features in the data irrespective of the choice of the process parameter , as demonstrated by figs . 2 and 3 .equation ( 2 ) @xcite does show distinct scaling behavior with a marked crossover , however , it is not appropriate for modeling the dynamics of daily temperature records . in @xcite ,the authors subsequently proposed a nonlinear counterpart of car(1 ) model , termed nlcar(1 ) in order to capture the probability distribution of the daily temperature records .the dfa results obtained in @xcite for nlcar(1 ) and car(1 ) were indistinguishable ( for details we refer to @xcite ) .however , this does not necessarily imply nlcar(1 ) and car(1 ) have similar volatile correlations .recent studies @xcite have used nonlinear stochastic models successfully to model glacial dynamics also reflected in long - range volatile correlations . encouraged by previous findings @xcite and results presented in @xcite , we believe that nlcar(1 ) may be a more appropriate model to capture dynamics of daily temperature records .a. kirly , i.m .jnosi , phys .e 65 ( 2002 ) 0511021 .peng , s. v. buldyrev , a. l. goldberger , s. havlin , f. sciortino , m. simon , h. e. stanley , nature 356 ( 1992 ) 168 - 170 .k . peng , s. v. buldyrev , s. havlin , m. simons , h. e. stanley , a. l. goldberger , phys rev e 49 ( 1994 ) 1685 .y. ashkenazy , p. ch .ivanov , s. havlin , c .- k .peng , a.l .goldberger , h.e .stanley , phys .( 2001 ) 1900 .j. w. kantelhardt , s. a. zschiegner , e. koscielny bunde , s. havlin , h. e. stanley , physica a 316 ( 2002 ) 87 . t. kalisky , y. ashkenazy , s. havlin , phys . rev .e 72 ( 2005 ) 011913 .r. b. govindan , a. bunde , s. havlin , physica a 318 ( 2003 ) 529. h. a. makse , s. havlin , m. schwartz , h. .e .stanley , phys .e 53 ( 1996 ) 5445 . j. w. kantelhardt , e. koscielny bunde , h. h. a. rego , s. havlin , a. bunde , physica a 295 ( 2001 ) 441 .e. koscielny bunde , a. bunde , s. havlin , y. goldreich , physica a 231 ( 1996 ) 393 .e. koscielny bunde , a. bunde , s. havlin , h.e .roman , y. goldreich , h .- j .schellenhuber , phys .81 ( 1998 ) 729 .r. b. govindan , d. vjushin , a. bunde , s .brenner , s. havlin , h .- j .schellnhuber , phys .lett . 89 ( 2002 ) 028501. j. f. eichner , e. koscielny bunde , a. bunde , s. havlin , h .- j .schellnhuber , phys .e 68 ( 2003 ) 062902 .monetti , s. havlin , a. bunde , physica a 320 ( 2003 ) 581 .y. ashkenazy , d. r. baker , h. gildor , j. geophys .110 ( 2005 ) doi:10.1029/2004jc002548 . and @xmath54 is shown by ( solid curve with open circles ) and ( solid curve with ) respectively .corresponding volatility analysis for @xmath24 and @xmath24 are shown with ( open circles and ) respectively .( b ) fluctuation function for @xmath30 and @xmath28 is shown by ( solid curve with open circles ) and ( solid curve with ) respectively .corresponding volatility analysis for @xmath30 and @xmath28 are shown with ( open circles and ) respectively .fluctuation function of the uncorrelated noise with slope 0.5 is shown at the bottom of fig 1 ( a ) and ( b ) as reference . in the asymptotic regimeall the curves are parallel to the reference line indicating uncorrelated nature of the process.,width=480,height=336 ] and @xmath55 ) and their volatility series is shown in ( a ) and ( b ) respectively .explanations are same as in fig .fluctuation function of the uncorrelated noise ( slope 0.5 , solid line ) and long - range correlated noise ( slope = 0.8 , dotted line ) is shown at the bottom in ( a ) and ( b ) as reference . in the asymptotic regime dfa curves for the original dataare parallel to the reference line with slope 0.8 and the dfa curves of the volatile data are parallel to the reference to line with slope 0.5 . , width=480,height=336 ] the data are correlated with an exponent @xmath56 .the dfa curves of the volatile data are parallel to the reference line with slope 0.5 indicating uncorrelated nature of the volatile correlations.,width=480,height=480 ]

<S> the present study investigates linear and volatile ( nonlinear ) correlations of first - order autoregressive process with uncorrelated ar ( 1 ) and long - range correlated car ( 1 ) gaussian innovations as a function of the process parameter ( @xmath0 ) . in the light of recent findings </S><S> @xcite , we discuss the choice of car ( 1 ) in modeling daily temperature records . </S><S> we demonstrate that while car ( 1 ) is able to capture linear correlations it is unable to capture nonlinear ( volatile ) correlations in daily temperature records . , time series , auto - regressive process , detrended fluctuation analysis , temperature records 05.45.tp </S>

cond-mat0011008

the coupled cluster method ( ccm ) @xcite has previously been applied @xcite to various lattice quantum spin systems with much success . in particular , high - order @xcite ccm calculations using a localised approximation scheme called the lsub@xmath1 approximation scheme have previously been utilised to determine the ground - state properties of the spin - half square lattice _ xxz_ model to high accuracy ( see especially refs .these calculations were performed both in a regime characterised by non - zero nel order @xcite in the @xmath0-direction , and also in a regime characterised by non - zero nel ordering in the @xmath5-plane @xcite .the results were found to be in excellent agreement with the known results of the _ xxz _ model ( discussed below ) .the spin - half _ xxz _ model is defined by the following hamiltonian , @xmath6 ~~ , \label{xxz_hamiltonian}\ ] ] where the index _ i _ in eq .( [ xxz_hamiltonian ] ) runs over all @xmath7 lattice sites , and @xmath8 runs over all _ z _ nearest - neighbours lattice vectors with respect to @xmath9 on the chain ( _ z_=2 ) , the square lattice ( _ z_=4 ) , and the simple cubic lattice ( _ z_=6 ) .periodic boundary conditions are also assumed .we note that @xmath10=0 $ ] for this model , where @xmath11 .the spin - half _ xxz _ model contains three regimes with respect to the anisotropy parameter @xmath12 . for @xmath13 there is a ferromagnetic regime in which the classical ground state is also the quantum - mechanical ground state . at @xmath14 ,there is a first - order phase transition to a planar regime in which the classical ground state is a nel state in the @xmath5-plane . for the square and cubic lattices , the quantum - mechanical ground state in this regimeis furthermore characterised by non - zero , in - plane , long - range ordering . for the linear chain quantum fluctuationsare strong enough to destroy all long - range ordering and so this region in 1d is sometimes referred to as the `` critical regime . '' in the 1d case the planar regime spans the range @xmath15 , and at @xmath16 there is another phase transition to a phase characterised by non - zero nel ordering in the @xmath0-direction for the ground state . for the linear chain , the bethe ansatz @xciteprovides an exact solution of the quantum spin - half _ xxz _ model , although for higher spatial dimensionality no exact solutions have as yet been determined .however , extensive approximate calculations have been carried out for this model , in particular for the heisenberg antiferromagnet ( haf ) at @xmath16 .it is believed that the 1d phase transition at @xmath16 also occurs for higher spatial dimensionality than one .examples of such approximate calculations for the spin - half _ xxz _ and haf models are spin - wave theory @xcite ( swt ) , exact diagonalisations of finite - sized systems @xcite , exact cumulant series expansions @xcite , and quantum monte carlo @xcite ( qmc ) calculations . in particular, we note that such results for the spin - half isotropic haf on the square lattice typically predict that approximately 61 - 62@xmath17 of the classical nel - ordering remains in the quantum case . in this articlewe present results of high - order coupled ccm calculations for the ground- and excited - state properties of the spin - half _ xxz _ model on the linear chain , square lattice , and simple cubic lattice based on the systematic incorporation of many - spin correlations on top of a nel model state with spins aligned in the @xmath0-direction . in sec .ii , we describe the high - order ccm formalisms for both the ground and excited states .although the high - order ccm formalism has been described previously @xcite , we present a very brief overview of it in order to give a necessary background for the new high - order formalism for the excited state .we also discuss in sec .ii the manner in which the ccm results for a localised approximation scheme called the lsub@xmath1 scheme ( in which we retain all @xmath2-body correlations defined on all possible locales of @xmath1 adjacent lattice sites ( @xmath3 ) ) are extrapolated to the limit @xmath18 . in sec .iii , we present the results of our high - order treatment of the _ xxz _ model for these lattices , and finally our conclusions are presented in sec . iv .in this section , we firstly describe the general ground - state ccm formalism @xcite , and then show how to apply it to the specific case of the spin - half _ xxz _ model .this is then extended to deal with excited states .the exact ket and bra ground - state energy eigenvectors , @xmath19 and @xmath20 , of a many - body system described by a hamiltonian @xmath21 , @xmath22 are parametrised within the single - reference ccm as follows : @xmath23 the single model or reference state @xmath24 is required to have the property of being a cyclic vector with respect to two well - defined abelian subalgebras of _ multi - configurational _ creation operators @xmath25 and their hermitian - adjoint destruction counterparts @xmath26 .thus , @xmath24 plays the role of a vacuum state with respect to a suitable set of ( mutually commuting ) many - body creation operators @xmath25 , @xmath27 with @xmath28 , the identity operator .these operators are complete in the many - body hilbert ( or fock ) space , @xmath29 the _ correlation operator _@xmath30 is decomposed entirely in terms of these creation operators @xmath25 , which , when acting on the model state ( @xmath31 ) , create multi - particle excitations on top of the model state .we note that although the manifest hermiticity , ( @xmath32 is explicitly imposed .the _ correlation coefficients _@xmath33 are regarded as being independent variables , even though formally we have the relation , @xmath34 the full set @xmath33 thus provides a complete description of the ground state . for instance , an arbitrary operator @xmath35 will have a ground - state expectation value given as , @xmath36 we note that the exponentiated form of the ground - state ccm parametrisation of eq .( [ eq2 ] ) ensures the correct counting of the _ independent _ and excited correlated many - body clusters with respect to @xmath24 which are present in the exact ground state @xmath19 .it also ensures the exact incorporation of the goldstone linked - cluster theorem , which itself guarantees the size - extensivity of all relevant extensive physical quantities .the determination of the correlation coefficients @xmath33 is achieved by taking appropriate projections onto the ground - state schrdinger equations of eq .( [ eq1 ] ) .equivalently , they may be determined variationally by requiring the ground - state energy expectation functional @xmath37 , defined as in eq .( [ eq6 ] ) , to be stationary with respect to variations in each of the ( independent ) variables of the full set .we thereby easily derive the following coupled set of equations , @xmath38 { \rm e}^s|\phi\rangle = 0 , \;\ ; \forall i \neq 0 \;\ ; . \label{eq8}\end{aligned}\ ] ] equation ( [ eq7 ] ) also shows that the ground - state energy at the stationary point has the simple form @xmath39 it is important to realize that this ( bi-)variational formulation does _ not _ lead to an upper bound for @xmath40 when the summations for @xmath30 and @xmath41 in eq .( [ eq2 ] ) are truncated , due to the lack of exact hermiticity when such approximations are made .however , it is clear that the important hellmann - feynman theorem _ is _ preserved in all such approximations .we also note that eq .( [ eq7 ] ) represents a coupled set of nonlinear multinomial equations for the _ c_-number correlation coefficients @xmath42 . the nested commutator expansion of the similarity - transformed hamiltonian , @xmath43 + { 1\over2 ! } [ [ h , s],s ] + \cdots \;\ ; , \label{eq10}\ ] ] together with the fact that all of the individual components of @xmath30 in the sum in eq .( [ eq2 ] ) commute with one another , imply that each element of @xmath30 in eq .( [ eq2 ] ) is linked directly to the hamiltonian in each of the terms in eq .( [ eq10 ] ) .thus , each of the coupled equations ( [ eq7 ] ) is of linked - cluster type .furthermore , each of these equations is of finite length when expanded , since the otherwise infinite series of eq .( [ eq10 ] ) will always terminate at a finite order , provided ( as is usually the case ) only that each term in the second - quantised form of the hamiltonian @xmath21 contains a finite number of single - body destruction operators , defined with respect to the reference ( vacuum ) state @xmath24 .therefore , the ccm parametrisation naturally leads to a workable scheme which can be efficiently implemented computationally .it is also important to note that at the heart of the ccm lies a similarity transformation , in contrast with the unitary transformation in a standard variational formulation in which the bra state @xmath44 is simply taken as the explicit hermitian adjoint of @xmath19 .we now wish to apply the general ccm formalism outlined above to the specific case of the spin - half _ xxz _ model , and we choose the nel state , in which the spins lie along the @xmath0-axis , to be the model state .furthermore , we perform a rotation of the local axes of the up - pointing spins by 180@xmath45 about the @xmath46 axis , so that spins on both sublattices may be treated equivalently .the ( canonical ) transformation is described by , @xmath47 the model state now appears @xmath48 to consist of purely down - pointing spins in these rotated local axes . in terms of the spin raising and lowering operators @xmath49 the hamiltonian may be written in these local axes as , @xmath50 \ ; . \label{eq : newh}\ ] ] in this article , we also wish to perform high - order ccm calculations for this model , and in order to do this we firstly define the @xmath51 operators more explicitly , as @xmath52 , where the set - index @xmath53 .each of the single - site indices is allowed to cover _ all _ lattice sites , although double ( or greater ) occupancy of any particular site in any set @xmath54 is explicitly prohibited for the spin - half case , since @xmath55 .we analogously define the ket - state correlation coefficients as @xmath56 for a cluster @xmath54 defined above . by constructionthe coefficients @xmath57 are completely symmetric under the interchange of any two indices .it is also useful to define the following two operators , @xmath58 the similarity transform of the hamiltonian of eq .( [ eq6 ] ) may now be written in terms of these operators .we find explicitly that @xmath59 , where @xmath60 we note that the similarity - transformed hamiltonian acting on the model state @xmath61 has now been written purely in terms of spin - raising operators .the problem of determining a given ccm ket - state equation , defined by eq .( [ eq7 ] ) for a given cluster of index @xmath54 , thus becomes an exercise in pattern - matching the spin - lowering operators in @xmath62 to the terms in @xmath63 contained in eqs .( [ eq13]-[eq15 ] ) .this task is perfectly suited to implementation using computer algebra techniques , and the ensuing set of coupled , non - linear ket - state equations is then easily solved ( for example , using the newton - raphson method ) .once the ket - state correlation coefficients have been determined it is then possible to find the bra - state coefficients similarly , as described in ref .one may finally calculate any ground - state expectation values that one wishes to obtain in terms of the coefficients @xmath64 .an example of this is the sublattice magnetisation , given by @xmath65 in the rotated spin coordinates defined above .the quantity provides a measure for the amount of nel ordering in the @xmath0-direction remaining in the ccm ground state , @xmath66 , with respect to the perfect ordering ( @xmath67 ) of the model state .the ccm formalism is exact in the limit of inclusion of all possible multi - spin cluster correlations within @xmath30 and @xmath68 , although in any real application this is usually impossible to achieve .it is therefore necessary to utilise various approximation schemes within @xmath30 and @xmath41 .the three most commonly employed schemes have been : ( 1 ) the sub@xmath69 scheme , in which all correlations involving only @xmath69 or fewer spins are retained , but no further restriction is made concerning their spatial separation on the lattice ; ( 2 ) the sub@xmath69-@xmath1 sub - approximation , in which all sub@xmath69 correlations spanning a range of no more than @xmath1 adjacent lattice sites are retained ; and ( 3 ) the localised lsub@xmath1 scheme , in which all multi - spin correlations over distinct locales on the lattice defined by @xmath1 or fewer contiguous sites are retained .note that for this system we also make the specific and explicit restriction that the creation operators @xmath70 in @xmath30 preserve the relationship that , in the original ( unrotated ) spin coordinates , @xmath71 in order to keep the approximate ccm ground - state wave function in the correct ( @xmath72 ) subspace . the number of such distinct ( or fundamental ) configurations for the ground state at a given level of approximation is labelled by @xmath73 .we denote as distinct configurations those which are inequivalent under the point and space group symmetries of the both the lattice and the hamiltonian . in practicewe find that the ccm equations often terminate at specific critical values ( denoted @xmath74 ) of the anisotropy parameter which are dependent on the particular approximation scheme chosen , such that no solution based on this model state exists for @xmath75 . at these _critical points _ the second derivative of the ground - state energy is found to diverge , and the critical points have been shown to reflect the corresponding phase transition point in the real system .we now turn our attention to the ccm parametrisation of the excited state developed by emrich @xcite , and as a specific example we present results for the high - order treatment of the excited states of the spin - half _ xxz _ model .an excited - state wave function , @xmath76 , is determined by linearly applying an excitation operator @xmath77 to the ket - state wave function of eq .( [ eq2 ] ) , such that @xmath78 this equation may now be used to determine the low - lying excitation energies , such that the schrdinger equation , @xmath79 , may be combined with its ground - state counterpart of eq .( [ eq1 ] ) to give the result , @xmath80 e^s | \phi \rangle ~ ( \equiv \hat r |\phi \rangle ) ~~ , \label{eq17}\ ] ] where @xmath81 is the excitation energy . by analogy with the ground - state formalism, the excited - state correlation operator is written as , @xmath82 where the set @xmath83 of multi - spin creation operators may differ from those used in the ground - state parametrisation in eq .( [ eq2 ] ) if the excited state has different quantum numbers than the ground state .we note that eq .( [ eq18 ] ) implies the overlap relation @xmath84 . by applying @xmath85 to eq .( [ eq17 ] ) we find that , @xmath86 e^s | \phi \rangle ~~ , \forall i \ne 0 ~~ , \label{temp1}\ ] ] which is a generalised set of eigenvalue equations with eigenvalues @xmath87 and corresponding eigenvectors @xmath88 , for each of the excited states which satisfy @xmath84 .analogously to the ground - state case , we define the excited - state operators @xmath89 the state , @xmath90e^s | \phi \rangle$ ] , may now be divided into three elements , @xmath91 , such that ( after collecting together like terms ) we find the following expressions , @xmath92 we note again that @xmath8 runs over all nearest - neighbour lattice vectors and that terms which are equivalent under a lattice translation which connects one sublattice to the other have been collected together in eqs .( [ eq21])-([eq23 ] ) . in order to determine explicitly the eigenvalue equation of eq .( [ temp1 ] ) we now _ pattern match _ the configurations in the set @xmath93 to the spin - raising operators contained in @xmath94 , analogously to the ground - state procedure .for low orders of approximation , this may be performed readily by hand , although for higher orders of approximation may one must again use computational methods .we note that the lowest - lying excited states for the _ xxz _ model lie in the @xmath95 and @xmath96 subspaces with respect to the `` unrotated '' ground state .we thus restrict the `` fundamental '' clusters in the set @xmath70 used in eq .( [ eq18 ] ) to be those which reflect this property , and the number of such fundamental configurations for the excited state at a given level of lsub@xmath1 approximation is denoted by @xmath97 . in order to solve fully the eigenvalue problem at a given value of @xmath12, we note furthermore that one must firstly fully determine and solve the ground ket - state equations in order to obtain numerical values for the set @xmath42 which are then used as input to the eigenvalue problem of eq .( [ temp1 ] ) .the level of lsub@xmath1 approximation is also explicitly restricted to be the same for the ground and excited states , such that the calculation is kept as systematic and self - consistent as possible .finally , we note that for the specific case of the spin - half _ xxz _ model considered here , we only consider the lowest eigenvalue of eq .( [ eq17 ] ) in order to calculate the excitation energy gap .we furthermore note that this eigenvalue is not specifically restricted by our truncation procedure to be a real number as the generalised eigenvalue problem of eq .( [ eq17 ] ) is not constrained to be symmetric .however , in practice we find that this eigenvalue is real at every value of @xmath12 considered , for each lattice , and at each level of lsub@xmath1 approximation .this provides a rather stringent check on the approximation scheme .we present results below of high - order ccm calculations for the spin - half _ xxz _ model for the linear chain , square lattice , and simple cubic lattice .in particular , we determine the ground - state energy , the sublattice magnetisation , and the lowest - lying excited - state energy for these models using the lsub@xmath1 approximation scheme .it is clearly useful to be able to extrapolate the `` raw '' ccm results , at each value of @xmath12 separately , to the limit @xmath98 . in the absence of any rigorous scaling theory for the results of lsub@xmath1 approximations, two empirical approaches are outlined below and utilised .the first approach , denoted to as `` extrapolated(1 ) ccm , '' assumes a leading `` power - law '' dependence of the lsub@xmath1 expectation values with @xmath1 .the value for the ground - state energy , sublattice magnetisation , or excitation energy ( at a given value of @xmath12 ) is denoted by @xmath99 for a given value of @xmath1 where @xmath100 .the index @xmath9 denotes the i@xmath101 data element of @xmath102 such elements , although we note that @xmath9 and @xmath1 do not have to be explicitly equivalent .the leading `` power - law '' extrapolation is now described by , @xmath103 we plot @xmath104 against @xmath105 and the best fit of the data set , @xmath106 , to the power - law dependence given above is obtained when the absolute value of the _ linear correlation _ of these points is maximised with respect to the variable @xmath107 .this value of @xmath107 is then assumed to be the extrapolated value for @xmath99 in the limit @xmath98 .the second such extrapolation scheme of the lsub@xmath1 data , denoted the `` extrapolated(2 ) ccm ''scheme , uses pad approximants in which the data set is modeled by the ratio of two polynomials , given by @xmath108 this furthermore implies that , @xmath109 we now wish to determine the coefficients @xmath110 and @xmath111 in order to find the polynomials in the equation above , and the above equation is rewritten in terms of a matrix given by , @xmath112 ( note that @xmath113 . )this problem may be easily solved by inverting the matrix in order to determine the coefficients @xmath114 and @xmath115 at each value of @xmath12 separately . in the limit , @xmath98, it is seen that the extrapolated value of @xmath99 is given by @xmath116 .we furthermore note that the case @xmath117 corresponds also to a `` least - squares '' fit of the @xmath102 data points to a @xmath118-order polynomial .previous extrapolation results @xcite for the haf on the square lattice fitted @xmath99 against @xmath119 for the ground - state energy , @xmath120 , and the critical points , @xmath74 , while the sublattice magnetisation was fitted against @xmath121 .these `` rules '' were chosen because the ccm sub2-@xmath69 calculations for the square lattice haf were found to converge as a function of @xmath69 to their full sub2 solution in these ways , and so analogous rules were used for the lsub@xmath1 data .note that results for the haf on the square and simple cubic lattices using these procedures are also quoted below . for the spin - half square latticehaf , we find that the present rules are reasonable approximations for the scaling of the lsub@xmath1 results .finally , we note that the lsub2 results for @xmath99 and @xmath122 generally fit in rather poorly to the asymptotic behaviour of @xmath99 and @xmath122 as a function of @xmath9 .as we are interested in the asymptotic value of @xmath99 , it is found that we obtain better extrapolated results by discarding the lowest order lsub2 results where possible , i.e. , for the linear chain and square lattice results . hence for both extrapolation schemes used here , lsub@xmath1 results are used with @xmath123 for the linear chain , with @xmath124 for the square lattice , but with @xmath125 for the simple cubic lattice .the results for the ground - state energy , @xmath120 , of the ccm treatment of the spin - half linear chain _ xxz _ model regime @xmath126 are found to be in excellent agreement with exact results @xcite over the whole of this regime , ( and therefore no graphical plot of this data is given ) .table [ tab1 ] indicates the accuracy of the raw and extrapolated lsub@xmath1 for the spin - half linear chain haf ( @xmath16 ) in comparison with exact bethe ansatz @xcite results .it is seen that all of the extrapolated results for the ground - state energy , apart from those of the extrapolated(2 ) ccm scheme with @xmath127 , are accurate to at least five decimal places .indeed , the spin - half linear chain haf is the `` worst - case - scenario '' within this regime for such ccm calculations .thus , for @xmath128 even better accuracy is obtained , and indeed the ccm results are exact at all levels of approximation in the limit @xmath129 , in which case the model state is the exact quantum ground state .the results for the sublattice magnetisation , @xmath130 , of this model using the ccm are shown in fig .it is seen that the ccm results tend to the exact results as @xmath12 increases , although they are non - zero and positive at @xmath16 . the raw ccm results may be greatly improved by extrapolation , and the extrapolated(1 ) ccm results of @xmath131 is in good agreement with the exact result that @xmath130 becomes zero at @xmath16 .the extrapolated(2 ) ccm results do not perform quite as well as the extrapolated(1 ) ccm results , although they do present a very significant improvement on the raw lsub@xmath1 results . however , fig .[ fig1 ] shows that the results of both extrapolations are in good qualitative agreement with exact results , and we note that the spin - half haf on the linear chain presents a particularly difficult challenge , since not only does all of the nel - like long - range order inherent in the model state disappear at this point but also there is an infinite - order phase transition at this point ( i.e. , one in which the energy and all of its derivatives with respect to @xmath12 are continuous ) .an indication of the efficacy of the ccm treatment of this model is also given for the lowest - lying excitation energy , as shown in fig .it is seen that the extrapolated results are in excellent agreement with the exact result , and are very much better than those results of linear spin - wave theory ( lswt ) over the whole of this regime . indeed , the manner in which the ccm results for the excitation energy behave as @xmath132 is in good qualitative and quantitative agreement with the exact results .this is in stark contrast to the results of lswt in 1d which clearly show completely incorrect qualitative behaviour for @xmath87 near to @xmath16 .furthermore , table [ tab1 ] indicates that the extrapolated(2 ) ccm results for the excitation energy of the spin - half linear chain haf are zero to three decimal places of accuracy .the extrapolated(1 ) results are also consistent with a gapless spectrum at @xmath16 . the various ccm results for the ground - state energy of the spin - half square lattice _ xxz_ model are found to be in excellent mutual agreement over a wide range of @xmath12 .this may be seen in fig .[ fig3 ] , in which both the `` raw '' ground - state energy ccm lsub@xmath1 results for @xmath133 and extrapolated ccm results are plotted for this model .it is also seen from fig .[ fig3 ] that our results are in excellent agreement with those results of the qmc method given in ref .@xcite for this model .table [ tab2 ] illustrates the ccm results for the special case of the spin - half haf , and it is seen that ccm results are in excellent agreement ( see also for example refs .@xcite ) with those results of swt @xcite , finite - sized calculations @xcite , series expansions @xcite , and qmc techniques @xcite .however , our results for the spin - half square lattice haf perhaps indicate that the ground - state energy might be slightly lower than the best qmc estimate obtained to date of sandvik @xcite which is , namely , @xmath134 .the extrapolated(1 ) ccm and extrapolated(2 ) ccm results are in also good agreement with the previous extrapolation @xcite of ccm lsub@xmath1 results at this point , which predict that @xmath135 .( we note again that the previous extrapolated value @xcite of @xmath135 was obtained by fitting to a quadratic in @xmath136 . ) as @xmath12 approaches the haf ( @xmath16 ) , we find that the ground - state energy exponent , @xmath137 , approaches the value @xmath138 .we also note that the expansion coefficient of the @xmath117 pad approximant ( viz . , the series in integral powers of @xmath139 ) which becomes overwhelmingly dominant as @xmath132 is the coefficient which corresponds to the @xmath136 term .thus , the extrapolated(1 ) and extrapolated(2 ) results have independently reinforced what were the admittedly `` naive '' assumptions made in the `` previous '' extrapolation @xcite in which the ground - state energy for the square lattice haf is fitted to a quadratic function in @xmath136 .( note that similar results for the scaling of the ground - state energy are also observed for both the linear chain and simple cubic lattice . )the various ccm results for the sublattice magnetisation of the spin - half square lattice _ xxz _ model are similarly in very good mutual agreement over a wide range of @xmath12 , as seen in fig .they are furthermore found to agree very well with the results of the best of other approximate methods for the specific case of the haf at @xmath140 , as shown in table [ tab2 ] , which typically yield results of @xmath141 .however , we note that the extrapolated(1 ) ccm result for the haf of @xmath142 given in table [ tab2 ] probably lies slightly too low .this is because the critical exponent @xmath137 for the sublattice magnetisation appears to change very rapidly near to the haf point for the square lattice , where , for example , we find that @xmath143 at @xmath144 and that @xmath145 at @xmath16 .it is expected however that the inclusion of the second - order term to the extrapolation using a leading - order power - law scaling would remedy this situation .we may see from fig .[ fig4 ] , however , that the extrapolated(2 ) ccm results appear to work extremely well up to and including the haf itself .( note that qualitatively similar results are observed for the simple cubic lattice as for the square lattice . )the ccm results for the lowest - lying excitation energy of the _ xxz _ model in the @xmath128 regime also appear to be in excellent mutual agreement , as shown in fig .the extrapolated ccm results seem to be qualitatively similar to results of lswt , although they lie significantly lower and are believed to be much more accurate than the lswt results in this regime .the mutual agreement of the extrapolated ccm results only ever falters very near to the haf point ( @xmath16 ) .indeed , the spin - half square lattice haf is believed to be gapless .the extrapolated(1 ) ccm results predict that the excitation energy gap disappears at @xmath146 .furthermore , the extrapolated(2 ) ccm @xmath117 and @xmath147 results strongly indicate that the spin - half square lattice isotropic haf ( @xmath16 ) is gapless , with results of @xmath148 and @xmath149 , respectively .indeed , we believe that the extrapolated(2 ) ccm results provide excellent results for the excitation gap of the spin - half square lattice _ xxz _ model across the whole of the regime , @xmath126 . the results for the critical points predicted by the ccm for the spin - half square lattice _ xxz_ model are also presented in table [ tab2 ] .these results strongly indicate that the phase transition point is at ( or very near to ) the haf point ( @xmath16 ) , which is fully consistent with the behaviour that is believed to occur in this model . we have also determined completely new ccm results for the spin - half _ xxz _ model on the simple cubic lattice .these results look qualitatively very similar to those for the square lattice model and so are not plotted here .it is seen from table [ tab3 ] that the results for the ground- and excited - state properties of the cubic - lattice haf are in good agreement with those results of lswt .the results for ground - state energy of the spin - half cubic - lattice haf are shown in table [ tab3 ] , and it is seen that the lsub6 approximation is already quite near the converged limit .the extrapolated(1 ) and extrapolated(2 ) ccm results and the `` previous extrapolated ccm '' result for the ground - state energy are thus seen to be in mutual agreement with each other . for the sublattice magnetisation of the spin - half cubic - lattice haf ,the extrapolated ccm results give values of @xmath150 , and this result is in excellent agreement with the result of lswt of @xmath151 .by contrast , the extrapolated(1 ) ccm result appears to lie too low for the lowest - lying excitation energy of the simple cubic - lattice haf .however , the extrapolated(2 ) ccm @xmath117 and @xmath147 results indicate that the spin - half cubic - lattice haf is gapless with results of @xmath152 and @xmath153 , respectively , which is once again consistent with the expected behaviour of this model . the results for the critical point of the spin - half cubic - lattice _ xxz_ model appear to converge to a value near to @xmath16 , which is again believed to be ( or be near to ) the correct result .( note that with only two points an assumption must be made as to the scaling of @xmath74 with @xmath1 .hence , in analogy with sub2-@xmath69 results the lsub@xmath1 values for @xmath74 were plotted against @xmath136 and linearly extrapolated , and so these results are referred to as the `` previous extrapolated ccm '' result in table [ tab3 ] . )in this article new results have been presented for the ground - state properties of the spin - half _ xxz _ model on the linear chain , the square lattice , and the simple cubic lattice .a localised approximation scheme was utilised within the ccm ground - state formalism , and this allowed high - order calculations to be very efficiently performed .the results were seen to be in excellent agreement with exact bethe ansatz calculations for the linear chain and the best results of other approximate methods for the square and simple cubic lattices .these results were extrapolated using a number of techniques which extend previous heuristic attempts at extrapolation .the results so obtained are of very high quality across a very wide range of the anisotropy parameter @xmath12 , thus reinforcing our results at the specific point for the heisenberg antiferromagnet at @xmath16 .indeed , it was found from these calculations that our previous heuristic extrapolations for the ground - state energy of the spin - half square lattice heisenberg antiferromagnet did , in fact , provide very reasonable results , and that the naive assumptions made in this extrapolation were largely justified .hence , for example , the previous `` best estimate '' of the ground - state energy of the spin - half square - lattice heisenberg antiferromagnet of @xmath154 was justified .furthermore , our ccm calculations for spin - half heisenberg antiferromagnet indicate that about 62@xmath17 of the classical nel ordering remains for the square lattice case , and that 83.6@xmath17 of the classical nel ordering remains for the simple cubic lattice case .a new formalism has also been introduced in order to treat for the first time the excited states of lattice quantum spin systems via the same high - order localised approximation scheme in a computationally efficient manner .the approximation level was kept at the same subapproximation level for both the ground and excited states , although the ccm ground - state correlations preserved the @xmath155 property of the ground state and the ccm excited - state correlations produced a change of @xmath156 with respect to the ccm approximation to the ket ground state .extrapolations were again attempted in the @xmath126 regime and very good correspondence of the extrapolated ccm results for the lowest - lying excitation energy were observed for the linear chain in comparison with the exact bethe ansatz results .furthermore , the results for the spin - half heisenberg antiferromagnet , for all of the lattices considered here , strongly indicate that the excitation energy gaps of these models are zero .future applications of the high - order formalism presented in this article will be to systems of complex crystallographic point- and space - group symmetries , e.g. , those of the cavo .another extension will be to apply these methods to systems of higher quantum spin number , @xmath157 .in particular , it would be of interest to apply the new high - order formalism for the excited state to the spin - one linear chain heisenberg antiferromagnet which contains an excitation energy gap . the high - order formalismmight also be applied to systems with electronic ( rather than just spin ) degrees of freedom .finally , the application of the ccm to spin systems at non - zero temperature remains a future goal .we wish to thank dr .ligterink for his useful and interesting discussions .one of us ( rfb ) gratefully acknowledges a research grant from the engineering and physical sciences research council ( epsrc ) of great britain . this work has also been supported by the deutsche forschungsgemeinschaft ( grk 14 , graduiertenkolleg on ` classification of phase transitions in crystalline materials ' andalso project no .ri 615/9 - 1 ) .one of us ( rfb ) also acknowledges the support of the isaac newton institute for mathematical sciences , university of cambridge , during a stay at which the final version of this paper was written .99 f. coester , _ nucl .phys . _ * 7 * , 421 ( 1958 ) ; f. coester and h. kmmel , _ ibid . _ * 17 * , 477 ( 1960 ) .h. kmmel , k.h .lhrmann , and j.g .zabolitzky , _ phys rep . _* 36c * , 1 ( 1978 ) .r.f . bishop and k.h .lhrmann , _ phys .b _ * 17 * , 3757 ( 1978 ) . j.s .arponen , _ ann .phys . _ _ ( n.y . ) _ * 151 * , 311 ( 1983 ) .arponen , r.f .bishop , and e. pajanne , _ phys . rev .a _ * 36 * , 2519 ( 1987 ) ; _ ibid . _* 36 * , 2539 ( 1987 ) ; _ ibid . _ * 37 * , 1065 ( 1988 ) .bishop , _ theor .. acta _ * 80 * , 95 ( 1991 ) .arponen , and r.f .bishop , _ ann .( n.y . ) * 207 * , 171 ( 1991 ) ; _ ibid . _* 227 * , 275 ( 1993 ) ; _ ibid . _* 227 * , 2334 ( 1993 ) .bishop , in _ microscopic quantum - many - body theories and their applications _ , edited by j. navarro and a. polls , lecture notes in physics , vol .* 510 * ( springer - verlag , berlin , 1998 ) , p. 1 .m. roger and j.h .hetherington , _ phys .b _ * 41 * , 200 ( 1990 ) ; m. roger and j.h . hetherington , _ europhys .lett . _ * 11 * , 255 ( 1990 ) .bishop , j.b .parkinson , and y. xian , _ phys .b _ * 44 * , 9425 ( 1991 ) .bishop , j.b .parkinson , and y. xian , _ j. phys . : condens. matter _ * 5 * , 9169 ( 1993 ) .farnell and j.b .parkinson , _ j. phys . : condens. matter _ * 6 * , 5521 ( 1994 ) . y. xian , _ j. phys .: condens .matter _ * 6 * , 5965 ( 1994 ) .r. bursill , g.a .gehring , d.j.j .farnell , j.b .parkinson , t. xiang , and c. zeng , _ j. phys .: condens . matter _ * 7 * , 8605 ( 1995 ) .bishop , r.g .hale , and y. xian , _ phys .lett . _ * 73 * , 3157 ( 1994 ) ; _ int .j. quantum chem . _ * 57 * , 919 ( 1996 ) .bishop , d.j.j .farnell , and j.b .parkinson , _ j. phys . : condens .matter _ * 8 * , 11153 ( 1996 ) . d.j.j .farnell , s.a .krger , and j.b .parkinson , _ j. phys . : condens* 9 * , 7601 ( 1997 ) .bishop , y. xian , and c. zeng , in _ condensed matter theories _ , edited by e.v .ludea , p. vashishta , and r.f .bishop vol .* 11 * ( nova science publ . , commack , new york , 1996 ) , p. 91 . c. zeng , d.j.j .farnell , and r.f .bishop , _ j. stat ._ , * 90 * , 327 ( 1998 ) . r. f. bishop , d.j.j .farnell , and j.b .parkinson , _ phys .b _ * 58 * , 6394 ( 1998 ) . r. f. bishop , d. j. j. farnell , and chen zeng , _ phys . rev . b _ * 59 * , 1000 ( 1999 ) .j. rosenfeld , n.e .ligterink , and r.f .bishop , _ phys .b _ * 60 * , 4030 ( 1999 ) . h. a. bethe , _ z. phys . _* 71 * , 205 ( 1931 ) .l. hulthn , _ ark . mat .* 26 * , no .11 ( 1938 ) .r. orbach , _ phys .rev . _ * 112 * , 309 ( 1958 ) ; c. n. yang and c. p.yang , _ ibid . _ * 150 * , 321 ( 1966 ) ; _ ibid . _ * 150 * , 327 ( 1966 ) .j. des cloiseaux and pearson , _ phys .* 128 * , 2131 ( 1962 ) ; l. d. faddeev and l. a. takhtajan , _ phys .* 85a * , 375 ( 1981 ) .p. w. anderson , _ phys .rev . _ * 86 * , 694 ( 1952 ) ; t. oguchi , _ phys ._ * 117 * , 117 ( 1960 ) .hamer , z. weihong , p. arndt , _ phys .b _ * 46 * , 6276 ( 1992 ) .s. tang and j.e .hirsch , _ phys .b _ * 39 * , 4548 ( 1989 ) .d. d. betts and g. e. stewart , _ can ._ * 75 * , 47 ( 1997 ) .w. zeng , j. oitmaa , and c. j. hamer , _ phys .b _ * 43 * , 8321 ( 1991 ) .t. barnes , d. kotchan , and e. s. swanson , _ phys .b _ * 39 * , 4357 ( 1989 ) .j. carlson , _ phys .b _ * 40 * , 846 ( 1989 ) ; n. trivedi and d. m. ceperley , _ ibid . _* 41 * , 4552 ( 1990 ) .k. j. runge , _ phys .b _ * 45 * , 12292 ( 1992 ) ; _ ibid . _ * 45 * , 7229 ( 1992 ) . a.w .sandvik , _ phys .b _ * 56 * , 11678 ( 1997 ) .k. emrich , _ nucl .phys . _ * a351 * , 379 , 397 ( 1981 ) ..results obtained for the spin - half haf on the linear chain using the ccm lsub@xmath1 approximation with @xmath158 .@xmath73 denotes the number of fundamental configurations for the ground state , and @xmath97 denotes the number of fundamental configurations for the excited state .the ground - state energy per spin , @xmath120 , the sublattice magnetisation , @xmath159 , and the lowest - lying excitation energy , @xmath87 , are shown .the lsub@xmath1 results for @xmath123 are extrapolated using the two extrapolation schemes described in the text , and these results are compared to exact results from the bethe ansatz . [ cols="<,^,^,^,^,^",options="header " , ] [ tab3 ]* figure 1 : * ccm results for the sublattice magnetisation of the spin - half _ xxz _ model on the linear chain using the lsub@xmath1 approximation based on the @xmath0-aligned nel model state , compared to exact bethe ansatz results .* figure 2 : * ccm results for the lowest - lying excitation energy of the spin - half _ xxz _ model on the linear chain using the lsub@xmath1 approximation based on the @xmath0-aligned nel model state , compared to results of exact bethe ansatz and results of linear spin - wave theory .* figure 3 : * ccm results for the ground - state energy of the spin - half _ xxz _ model on the square lattice using the lsub@xmath1 approximation based on the @xmath0-aligned nel model state , compared to quantum monte carlo ( qmc ) calculations @xcite .* figure 5 : * ccm results for the lowest - lying excitation energy of the spin - half _ xxz _ model for the square lattice , using the lsub@xmath1 approximation based on the @xmath0-aligned nel model state , compared to results of linear spin - wave theory .

<S> in this article , we present new results of high - order coupled cluster method ( ccm ) calculations , based on a nel model state with spins aligned in the @xmath0-direction , for both the ground- and excited - state properties of the spin - half _ xxz _ model on the linear chain , the square lattice , and the simple cubic lattice . in particular , </S><S> the high - order ccm formalism is extended to treat the excited states of lattice quantum spin systems for the first time . </S><S> completely new results for the excitation energy gap of the spin - half _ xxz _ model for these lattices are thus determined . </S><S> these high - order calculations are based on a localised approximation scheme called the lsub@xmath1 scheme in which we retain all @xmath2-body correlations defined on all possible locales of @xmath1 adjacent lattice sites ( @xmath3 ) . </S><S> the `` raw '' ccm lsub@xmath1 results are seen to provide very good results for the ground - state energy , sublattice magnetisation , and the value of the lowest - lying excitation energy for each of these systems . </S><S> however , in order to obtain even better results , two types of extrapolation scheme of the lsub@xmath1 results to the limit @xmath4 ( i.e. , the exact solution in the thermodynamic limit ) are presented . </S><S> the extrapolated results provide extremely accurate results for the ground- and excited - state properties of these systems across a wide range of values of the anisotropy parameter . </S><S> pacs numbers : 75.10.jm , 75.50ee , 03.65.ca </S>

astro-ph0510746

to collect the largest possible samples of blue stragglers and comparison stars for low - luminosity globular clusters , we used high quality ground - based photometry from the literature and photometry from the _ hubble space telescope ( hst ) _ archives ( all using the wfpc2 camera ) . in all cases , we chose datasets with little photometric scatter at the turnoff , so that contamination of the blue straggler sample by normal main sequence stars was minimal .all _ hst _ data was reduced in a uniform manner using the hstphot package @xcite .our final sample consists of 13 globular clusters with @xmath6 ( 9 with hst data ) .color - magnitude diagrams ( cmds ) for the primary _ hst _ fields are shown in fig . [ cmds ] .the blue stragglers discussed here are all brighter than the cluster turnoff and bluer ( by 0.05 in @xmath7 ) , and there was typically no difficulty with field star contamination of the portion of the cmd populated by stragglers . although most of our straggler identifications are based on @xmath8 colors ( as opposed to @xmath9 for p04 ) , there were few stars for which the choice of color could affect the selection . in some clusters , a significant field star population redder than the cluster turnoff and brighter than the subgiant branch made it impractical to use cluster giants as a comparison population ,so we used hb stars instead . clusters with large field contamination could sometimes be used if they had blue hb morphologies .the complete sample of blue stragglers and hb stars will be presented in a companion paper .we also examined the blue straggler populations of two well - studied old open clusters : m67 and ngc 188 .both of these clusters have extensive proper motion membership information available that allows us to identify stragglers , and to estimate the absolute magnitude of the cluster by summing the fluxes of member stars ( @xmath10 and @xmath11 , respectively ) .we have also estimated the central luminosity densities ( @xmath12 and 2.2 , respectively ) using structural information from @xcite and @xcite .although these clusters have lower ages ( and more luminous stars on average ) , these two clusters have characteristics comparable to the faintest globular clusters . p04 used two measures to judge the relative frequency of blue stragglers in a cluster : the population ratio @xmath13 , and the number of stragglers per absolute visual flux unit @xmath14 . in fig .[ fbss ] , we plot @xmath15 versus absolute cluster magnitude and central density .values for these cluster properties were taken from the compilation of @xcite . with the inclusion of data for these 13 clusters ,we extend the cluster sample fainter about 3 magnitudes in @xmath16 and more than 3 orders of magnitude in central density .the significant anti - correlation between @xmath15 and @xmath16 seen by p04 continues to the low luminosity end of the sample .that end is anchored by the clusters e 3 ( @xmath17 ) and palomar 13 ( @xmath18 ) .e 3 has no known hb stars ( one unconfirmed candidate ) , and so produces a lower limit .the majority of the clusters have higher relative frequencies than all but a handful of the clusters in the p04 sample .( we also examined the cluster m71 , which was the faintest cluster in the p04 data , to verify that our straggler selection produced a straggler frequency equal to theirs to within the errors . )none of the clusters examined here has @xmath15@xmath19 , which is a noticeable departure from the behavior of more luminous clusters in spite of the significant variations from cluster to cluster . although the open clusters that we have been able to include are significantly younger than the globular clusters in the sample ( 4 gyr for m67 , 6 gyr for ngc 188 ) , they appear to follow the relation defined by the globular clusters , and agree with the high value found for pal 13 . we have used red clump stars ( in ngc 188 , including two blue subdwarfs ) as the comparison populations for the open clusters ,but this should not be an issue since the lifetimes of stars in this phase is within about 10% of the lifetimes of red hb stars in globular clusters @xcite .this puts the open cluster straggler populations in a new context normal for their total masses , not abnormally large .because our sample of low - luminosity clusters also contains some of the lowest density clusters , we compare to the population of field blue stragglers @xcite .the prevalence of single - lined low - eccentricity spectroscopic binaries in the field sample implies that stable binary mass transfer is the main production mechanism among field stars .( coalescence of close binaries like w uma systems would produce the rest . )the 4 lowest - luminosity clusters in our sample are known to have binary fractions comparable to that of field stars , based on precise photometry of the main sequence [ @xmath20 for pal 13 @xcite ; @xmath21 for e3 @xcite ; @xmath22 for m67 @xcite ; @xmath23 for ngc 188 @xcite ] .@xcite found that low - luminosity ( @xmath24 ) globular clusters have larger @xmath15 values than higher luminosity clusters and estimated that @xmath15@xmath25 in the field .we find that _ only _ the four least luminous clusters have bss frequencies that are consistent with ( palomar 13 , m67 , ngc 188 ) or exceed ( e3 ) the preston & sneden estimate for the field .thus it seems that all but the least luminous globular clusters are less efficient at producing stragglers than the field .binary systems that transfer mass are quite `` hard '' systems with binding energies that are considerably larger than kinetic energies of the stars they encounter in most clusters @xcite .a figure of merit is @xmath26 , where @xmath27 , @xmath28 , and @xmath29 are masses of the binary components and the average single star , @xmath30 is the binary separation , and @xmath31 is the cluster velocity dispersion @xcite .the boundary between `` hard '' and `` soft '' binaries is @xmath32 , while mass transfer binaries in the clusters described here have @xmath33 . in general, such hard binaries should survive in a cluster and produce stragglers via mass transfer .do stellar collisions still produce stragglers though ? for the nearby open clusters , studies of _ individual _ blue stragglers provide information on the formation mechanisms . in ngc 188, the straggler pkm 4330 has a high membership probability ( @xmath34% ; platais et al .2003 ) , but stands more than 3 magnitudes above the turnoff magnitude , which probably means it has more than twice the turnoff mass and results from a multi - star collision .there is a much larger body of information on m67 stars . like pkm 4330 ,the brightest straggler ( s977 ) appears to have a mass more than twice the cluster turnoff mass .there are many variable blue stragglers , including s1036 ( w uma variable ) , s1082 ( rs cvn variable , x - ray source , possible triple system containing _ two _ blue stragglers ) , and s1280 and s1284 ( @xmath35 scuti pulsators , s1284 in a short - period eccentric binary ) .several more blue stragglers are in long - period spectroscopic binaries : s752 , s975 , and s1195 ( low - eccentricity orbits , s975 with hot companion ) , and s1267 and s997 ( eccentric orbits ) .roughly half of these systems probably formed via normal binary mass transfer ( s752 , s975 , s1036 , and s1195 ) , but an equally large number of systems probably require a collisional mechanism involving more than 2 stars ( most notably , s977 , s997 , s1082 , s1267 , and s1284 ) gyr ; @xcite ) , we must keep in mind that these stragglers would have evolved and died by the time the cluster s age reached that of a globular cluster ( @xmath36 gyr ) . ] .the clusters in this study reveal a continuation of the strong anti - correlation between straggler frequency and cluster luminosity seen by p04 , and a break in the weaker anti - correlation with cluster density ( lower panel of fig .an important question is whether the dependence on cluster density has physical meaning or whether it is an artifact of a correlation between cluster properties .[ pars ] shows the relation between integrated magnitude and central density @xcite .the similarity of this plot and the bottom panel of fig .[ fbss ] leads us to believe that the density relationship is an artifact the clusters e3 and pal 13 most clearly stand out as clusters with high straggler frequencies , low luminosity , and relatively high central densities .low - luminosity clusters ( @xmath37 ) cover almost the entire range of central densities , while straggler frequencies remain uniformly high .we must keep in mind that for most of the clusters surveyed by p04 the blue straggler samples are mostly confined to the core , and selection effects in the p04 cluster sample ( aside from @xmath6 ) may influence the interpretation of these figures .however , central density is not as good a predictor of the size of the straggler population as integrated cluster luminosity .does this mean that strong gravitational interactions between stars are unimportant in the production of stragglers in low - luminosity clusters ? as mentioned in [ field ] , a significant fraction of stragglers in m67 are in binaries showing evidence of strong multiple - star interactions .stars at the bright end of the straggler luminosity function should also reflect the relative importance of stellar collisions and binary coalescences since mass transfer is very unlikely to be 100% efficient during the formation of a straggler in a wide binary .as shown in fig .[ cmds ] , the clusters in our low - luminosity sample still produced a substantial fraction of stragglers at the high luminosity end ( more than 10% have @xmath38 ) .9 of the 13 globular clusters ( as well as m71 ) and both open clusters have at least one straggler more than 2 magnitudes brighter than the turnoff , implying a collision of at least three stars . we can attempt to qualitatively understand the problem by examining the timescale for collisions @xmath39 @xcite .the dominant factor in @xmath39 is number density ( @xmath40 ) .our cluster sample covers more than three orders of magnitude in central density , and when combined with the sample of p04 , the total range is more than six orders of magnitude .so , it might be tempting to believe that the near constancy of @xmath15 for many clusters with @xmath41 is an indication that primordial binary star mechanisms dominate and strong interactions have become unimportant .the velocity dispersion @xmath31 is also a few times lower for low - luminosity clusters compared to most clusters in the p04 sample : from a maximum of 2.4 km s@xmath42for ngc 6535 @xcite to @xmath43 km s@xmath42for palomar 13 @xcite and 0.8 km s@xmath42for m67 @xcite , to @xmath44 km s@xmath42for palomar 5 @xcite .low encounter velocities increase the time between interactions but more strongly increase the probability of collision via gravitational focusing , so that @xmath45 @xcite .velocity dispersion only varies by a factor of about 100 among all milky way globular clusters , so its effects on the timescale are usually overwhelmed by those of cluster density .the velocity dispersion has a larger effect on the statistical properties of the binary star population though .as velocity dispersion decreases , wider binaries qualify as `` hard '' ( @xmath46 ) . for @xmath47 km s@xmath48 ,the hard - soft boundary is already at several au . for most globular clusters ( and all of the ones in the present sample ) , binary systems that would ultimately produce blue stragglers via mass transfer in a giant evolutionary phase are likely to survive due to their hardness , even though collisions are fairly frequent . for the moment , we only consider collisions ( i.e. strong gravitational interactions ) involving binaries , and not physical collisions of stars . in a cluster like ngc 6535 at the bright end of our sample , the collision timescale for the largest binaries that transfer mass in the giant phase implies that they have a few collisions in the age of the cluster .( keep in mind that these collision timescales use central densities , and so are overestimates for the cluster as a whole . ) for lower luminosity clusters , there are competing effects .lower velocity dispersion allows wider binaries to survive to the present day .photometric measurements of large binary fractions among some of the low - luminosity clusters in our sample @xcite support this .lower cluster density increases the collision timescale for binaries of a given size , although this is partly offset by lower @xmath31 and increased gravitational focussing .however , the changes in the nature of the binary population ( increased binary fraction and increased average orbital separation ) help enhance the overall binary collision rate .the collision timescale for binaries at the hard - soft boundary has @xmath49 , which allows lower velocity dispersion to partially cancel the effects of lower number density .if we assume that the present - day binary period distribution is uniform in @xmath50 between a minimum period ( set by star sizes ) and the hard - soft boundary , then the typical binary involved in a strong interaction will have @xmath51 , where @xmath52 and @xmath53 are orbital separations of the hardest binaries and the hard - soft boundary , respectively .thus , the collision timescale for these systems should only be a factor of @xmath54 more than the collision timescale for binary systems at the hard - soft boundary .a strong interaction involving a close binary is very likely to produce at least one physical collision of stars ( @xmath55 , where @xmath56 is the critical velocity needed to unbind the binary ; fregeau et al .2004 ) . however , for wide but hard binaries in low - mass clusters , the main sequence star components present small targets compared to the binary . for a ratio of stellar radius to orbital separation @xmath57 ,the cross section for collisions decreases approximately as @xmath58 , according to figure 9 of @xcite . because the binary collision timescale @xmath59 , the timescale for stellar collisions involving binaries then _ decreases _ slowly with increasing @xmath30a combination of effects may therefore conspire to produce a significant number of stellar collisions involving binaries even within low luminosity clusters .the lack of a stronger correlation between @xmath15 and central density may result from the effects of changing velocity dispersion and binary population .this might also partially explain the apparent lack of correlation between @xmath15 and the probability @xmath60 that a given star will have a collision in one year ( p04 ) since the calculation of @xmath60 from king models does not account for changes in the binary star distributions between clusters .populations of low - luminosity x - ray sources in globular clusters [ such as quiescent low - mass x - ray binaries ( qlmxbs ) , cataclysmic variables , and millisecond pulsars ] appear to behave differently .@xcite found strong correlations between number of sources and normalized encounter rate @xmath61 , and @xcite identified probable qlmxbs that are consistent with predictions based on tidal capture or exchanges during multiple star interactions .it is not clear that this is consistent with the scenario of @xcite in which massive stars exchange into previously - existing binaries in the most massive clusters , forming stragglers that die off before the present day .the blue straggler and low - luminosity x - ray populations may be giving us complimentary information about the importance of collisions and the prevalence of binaries in star clusters .we suggest two directions to disentangle the important factors driving the dynamical evolution of clusters .first , the binary star populations of more clusters need to be observationally characterized particularly the binary fraction and the period distribution .second , models of cluster evolution that examine the production of stragglers and x - ray sources are needed .the recent hybrid models ( binary population synthesis with dynamics ) of @xcite provide some guidance , although they modelled clusters that are denser and more massive than the ones in our sample .they find that the present - day core binary fraction is inversely related to cluster density , and that the binary period distribution remains nearly flat in @xmath62 up to periods of @xmath63 d and more for their lowest - density clusters .the low - luminosity clusters examined in this study make perfect subjects for numerical studies of cluster dynamics with current generations of computer codes .we would like to thank the anonymous referee for helpful suggestions , k. cudworth , d. dinescu , b. mochejska , a. rosenberg , and a. sarajedini for copies of their datasets , and useful conversations with o. de marco and j. fregeau .this work has been funded through grant ast 00 - 98696 from the national science foundation to e.l.s . and m. bolte .binney , j. & tremaine , s. 1987 , galactic dynamics ( princeton university press , princeton ) blecha , a. , meylan , g. , north , p. , & royer , f. 2004 , , 419 , 533 bonatto , c. , & bica , e. 2003 , , 405 , 525 bonatto , c. , bica , e. , & santos , j. f. c. 2005 , , 433 , 917 carney , b. w. , latham , d. w. , laird , j. b. , grant , c. e. , & morse , j. a. 2001 , , 122 , 3419 carney , b. w. , latham , d. w. , & laird , j. b. 2005 , , 129 , 466 clark , l. l. , sandquist , e. l. , & bolte , m. 2004 , , 128 , 3019 davies , m. b. , piotto , g. , & de angeli , f. 2004 , , 349 , 129 dolphin , a. e. 2000a , , 112 , 1383 ferraro , f. r. , fusi pecci , f. , & bellazzini , m. 1995 , , 294 , 80 fregeau , j. m. , cheung , p. , portegies zwart , s. f. , & rasio , f. a. 2004 , , 352 , 1 girard , t. m. , grundy , w. m. , lopez , c. e. , & van altena , w. f. 1989 , , 98 , 227 girardi , l. , bressan , a. , bertelli , g. , & chiosi , c. 2000 , , 141 , 371 harris , w. e. 1996 , , 112 , 1487 heinke , c. o. , grindlay , j. e. , lugger , p. m. , cohn , h. n. , edmonds , p. d. , lloyd , d. a. , & cool , a. m. 2003 , , 598 , 501 ivanova , n. , belczynski , k. , fregeau , j. m. , & rasio , f. a. 2005 , , 358 , 572 montgomery , k. a. , marschall , l. a. , & janes , k. a. 1993 , , 106 , 181 odenkirchen , m. , grebel , e. k. , dehnen , w. , rix , h .- w . , & cudworth , k. m. 2002 , , 124 , 1497 piotto , g. , et al .2004 , , 604 , l109 ( p04 ) platais , i. , kozhurina - platais , v. , mathieu , r. d. , girard , t. m. , & van altena , w. f. 2003 , , 126 , 2922 pooley , d. , et al . 2003 , , 591 , l131 preston , g. w. , & sneden , c. 2000 , , 120 , 1014 pryor , c. , & meylan , g. 1993 , in asp conf .50 , structure and dynamics of globular clusters , eds .s. g. djorgovski and g. meylan , ( asp : san francisco ) , 357 vandenberg , d. a. , & stetson , p. b. 2004 , , 116 , 997 veronesi , c. , zaggia , s. , piotto , g. , ferraro , f. r. , & bellazzini , m. 1996 , in asp conf .92 , formation of the galactic halo ... inside and out , ed .morrison & a. sarajedini ( san francisco : asp ) , 301 von hippel , t. , & sarajedini , a. 1998 , , 116 , 1789

<S> we examine the blue straggler populations of 13 low - luminosity ( @xmath0 ) globular clusters and 2 old open clusters . </S><S> these clusters test blue straggler formation in environments intermediate between higher luminosity ( and usually higher density ) clusters and the galactic field . the anti - correlation between the relative frequency of blue stragglers ( @xmath1 ) and cluster luminosity continues to the lowest luminosity clusters , which have frequencies meeting or exceeding that of field stars . </S><S> in addition we find that the anti - correlation between straggler frequency and central density disappears for clusters with density less than about @xmath2 pc@xmath3 , although this appears to be an artifact of the correlation between cluster luminosity and central density . </S><S> we argue on observational ( wide , eccentric binaries containing blue stragglers in m67 , and the existence of very bright stragglers in most of the clusters in our sample ) and theoretical grounds that stellar collisions still produce a significant fraction of the blue stragglers in low luminosity star clusters due to the long - term survival of wide binaries . </S><S> blue stragglers are one of several exotic stellar populations whose creation seems to be intimately tied to the dynamical environment that stars find themselves in . </S><S> these stars are probably created when the mass of a normal main sequence star is increased by external means : the longest - surviving explanations involve either mass transfer within a binary system or the merger of stars in a binary or during a collision . </S><S> the main debate is over which mechanisms play the largest roles in different environments . in the galactic field at one extreme , </S><S> the probability that two field stars ( or binary systems ) will strongly interact after they leave their birth environment is expected to be vanishingly small . </S><S> indeed , radial velocity surveys have identified a high fraction of single - line spectroscopic binaries with low eccentricity among field blue stragglers , which supports the idea that these stragglers were formed via binary mass transfer @xcite . at the other extreme </S><S> are the most luminous ( @xmath4 ) globular clusters . </S><S> ( * ? ? ? </S><S> * hereafter p04 ) found a strong _ anti - correlation _ between blue straggler frequency [ relative to horizontal branch ( hb ) stars ] and integrated cluster magnitude , and a weaker anti - correlation with central density . </S><S> ( a similar relation can be seen in @xcite . ) because expected collision rates correlate with cluster mass , @xcite proposed that primordial binaries that would form stragglers were being depleted earlier in the history of a massive cluster , thanks to exchanges of massive stars into the binaries and subsequent mass transfer . </S><S> for the most massive clusters ( @xmath5 ) , they predicted that star collisions would take over as the dominant mechanism for forming stragglers . </S><S> regardless of the interpretation of the p04 results , the connection between the field and cluster environments has not been made . to this end , we have examined low - mass globular clusters and old open clusters to see whether the straggler populations still show the influences of the cluster environment , or whether that influence has begun to disappear . </S>

1102.5171

over the past decades , multi - electrode recordings ( taketani and baudry , 2006 ) have unveiled the nature of the activity of populations of neural cells in various systems , such as the vertebrate retina ( schnitzer and meister , 2003 ) , cortical cultures ( tang et al , 2008 ) , or the prefrontal cortex ( peyrache et al , 2009 ) .the observation of substantial correlations in the firing activities of neurons has raised fundamental issues on their functional role ( romo , hernandez , zainos and salinas , 2003 ; averbeck , latham and pouget , 2006 ) . from a structural point of view ,a challenging problem is to infer the network and the strengths of the functional interactions between the neural cells from the spiking activity ( fig .[ fig-1]a ) .powerful inference procedures are needed , capable to handle massive data sets , with millions of spikes emitted by tens or hundreds of neurons .a classical approach to infer functional neural connectivity is through the study of pairwise cross - correlations ( perkel , gerstein and moore , 1967 ; aersten and gerstein , 1985 ) .the approach was applied in a variety of neural systems , including the auditory midbrain of the grassfrog ( epping and eggermont , 1987 ) , the salamander retina ( brivanlou , warland and meister , 1998 ) , the primate and rat prefrontal cortex ( constantidinidis , franowicz and goldman - raking , 2001 ; fujisawa , amarasingham , harrison and buzsaki , 2008 ) .other approaches , capable of taking into account network - mediated effects , were proposed based on concepts issued from statistics and graph theory ( seth and edelman , 2007 ; dahlhaus , eichler and sandkhler , 1997 ; sameshima and baccal , 1999 ; jung , nam and lee , 2010 ) , information theory ( bettencourt , stephens , ham and gross , 2007 ) , or statistical physics ( schneidman , berry , segev and bialek , 2006 ; shlens et al , 2006 ) .an alternative approach is to assume a particular dynamical model for the spike generation .the generalized linear model , which represents the generation of spikes as a poisson process with a time - dependent rate is a popular framework ( brown , nguyen , frank , wilson and solo , 2001 ; truccolo et al , 2005 ; pillow et al , 2008 ) . the integrate - and - fire ( if ) model , where spikes are emitted according to the dynamics of the membrane potential is another natural candidate ( gerstner and kistler , 2002 ; jolivet , lewis and gerstner , 2004 ) .the problem of estimating the model parameters ( external current , variance of the noise , capacitance and conductance of the membrane ) of a single stochastic if neuron from the observation of a spike train has received a lot of attention ( paninski , pillow and simoncelli , 2004 ; pillow et al ., 2005 ; mullowney and iyengar , 2008 ; lansky and ditlevsen , 2008 ) .few studies have focused on the inference of interactions in an assembly of if neurons ( makarov , panetsos and de feo , 2005 ) .recently , we proposed a bayesian algorithm to infer the interactions in a network of stochastic perfect integrators when the synaptic integration is instantaneous and the noise is vanishingly small ( cocco , leibler and monasson , 2009 ) . in the present work we introduce a bayesian algorithm to infer the couplings and the external currents in an assembly of leaky if neurons , and in presence of moderate input noise ( fig .[ fig-1]a ) .the computational time grows as the product of the number of recorded spikes , and the square of the number of neurons .we validate the algorithm on synthetic data , and apply it to real recordings of the ganglion cell activity in the salamander retina , presented with natural visual stimuli , and in the absence of stimulus ( spontaneous activity ) .in the leaky integrate - and - fire ( lif ) model , the membrane potential @xmath4 of neuron @xmath5 at time @xmath6 obeys the first - order differential equation , @xmath7 where @xmath8 and @xmath9 are , respectively , the capacitance and conductance of the membrane .the ratio @xmath10 is the membrane leaking time .@xmath11 is the synaptic current coming from the other neurons and entering the neuron @xmath5 at time @xmath6 : @xmath12 where @xmath13 is the strength of the connection from neuron @xmath14 onto neuron @xmath5 ( figure [ fig-1]a ) ; @xmath15 is the time at which neuron @xmath14 fires its @xmath16 spike .we assume that synaptic inputs are instantaneously integrated , _i.e. _ that the synaptic integration time is much smaller than all the other time scales , including @xmath17 .our method for inferring the interactions relies on this assumption , and should be modified in the presence of synaptic integration kernels with temporal filtering .@xmath18 is a constant external current flowing into neuron @xmath5 ( fig .[ fig-1]a ) , and @xmath19 is a fluctuating current , modeled as a gaussian noise process : @xmath20 , @xmath21 .the noise standard deviation , @xmath22 , has here the dimension of a current times the square root of a time .an alternative definition would consist in rescaling @xmath22 with a time dependent factor , _e.g. _ @xmath23 ; our definition allows us to reach the perfect integrator limit ( @xmath24 ) while keeping @xmath22 fixed .the neuron @xmath5 remains silent as long as @xmath25 remains below a threshold potential @xmath26 .if the threshold is crossed at some time @xmath27 , _ i.e. _ @xmath28 , then a spike is emitted , and the potential is reset to its resting value : @xmath29 .the dynamics then resumes following ( [ ode ] ) .let @xmath30 and @xmath31 denote the sets of , respectively , the interactions and currents .let @xmath32 $ ] be the time at which neuron @xmath5 emits its @xmath16 spike ; @xmath33 is the duration of the recording .how can we infer the interactions and currents from the observation of the spiking activity ?consider the raster plots in fig .[ fig-1]b . in pattern @xmath34, the timings of the spikes of neuron 1 do not seem to be correlated to the activity of neuron 2 .hence , we may guess that there is no interaction from neuron 2 to neuron 1 ( @xmath35 ) . in pattern @xmath36 ,a spike of neuron 1 is likely to follow a spike of neuron 2 , which suggests that the interaction @xmath37 is positive .conversely , in pattern @xmath38 , it seems that the firing of neuron 2 hinders the firing of neuron 1 , which indicates that @xmath37 has a negative value .this crude reasoning can be made mathematically rigorous in the framework of statistical inference ( cover and thomas , 2006 ) .let us define the likelihood @xmath39 of a set of spiking times , @xmath40 , given @xmath41 and @xmath42 . according to bayes rule the most likely couplings and currents , @xmath43 and @xmath44 , given the set of spiking times @xmath45 can be inferred through the maximization of @xmath39 .due to the statistical independence of the noises @xmath46 from neuron to neuron , the likelihood @xmath47 of @xmath48 given @xmath49 can be written as the product of first - passage time ( fpt ) probabilities , @xmath50 here @xmath51 denotes the probability that @xmath25 crosses @xmath26 for the first time at time @xmath52 , starting from @xmath53 at time @xmath54 and conditioned to the inputs from the other neurons at times @xmath55 , with @xmath56 . as the synaptic integration is instantaneous , an incoming spike from neuron @xmath14 results in a ( positive or negative ) jump of the potential @xmath25 by @xmath57 ; @xmath51 can therefore be interpreted as the fpt density probability for a one - dimensional ornstein - uhlenbeck process with a time - dependent force .it is important to stress that the presence of the products over the spike intervals in ( [ prodfpt ] ) does not entail that the spiking times are independent .consider now the potential @xmath4 during the inter - spike interval ( isi ) @xmath58 $ ] .the boundary conditions are @xmath59 ( reset of the potential right after a spike ) , and @xmath60 ( condition for firing ) . at intermediate times, the potential can take any value smaller than @xmath26 .the logarithm of the probability of a dynamical path ( time course ) of the potential over the @xmath16 isi of neuron @xmath5 is , after multiplication by the variance @xmath61 of the noise , @xmath62&= & -\frac 1{2 } \int _ { t_{i , k}}^{t_{i , k+1 } } dt\ \eta_i(t)^2 \\&= & -\frac 1{2 } \int _ { t_{i , k}}^{t_{i , k+1 } } dt\ \left [ c\,\frac{d v_i}{dt}(t ) + g\ , v_i(t ) -i_i^{syn } ( t)-i_i\right]^2 \ , \nonumber\end{aligned}\ ] ] according to the gaussian nature of the noise @xmath19 and to the dynamical equation of the lif ( [ ode ] ) .while no exact expression is known for @xmath51 , it can be analytically approximated by the contribution of the most probable dynamical path for the potential , @xmath63 ( paninski , 2006 ) .this approximation becomes exact when the standard deviation @xmath22 of the noise is small .the idea is to replace the distribution of paths for the potential @xmath4 with a single , most likely path @xmath63 , which we call optimal .we now explain how to derive @xmath64 through the condition that the log probability @xmath65 ( [ pathint2 ] ) is maximal .let us assume first that @xmath66 .then , the derivative of @xmath65 in ( [ pathint2 ] ) with respect to @xmath64 must vanish , which gives @xmath67 we now turn this second order differential equation for the optimal potential into a first order differential equation at the price of introducing a new function , @xmath68 , and a new first order differential equation for this function .it is straightforward to check that the solution of @xmath69 is a solution of the optimization equation ( [ optima ] ) if @xmath70 fulfills @xmath71 where @xmath17 is the membrane leaking time .the similarity between eqns ( [ ode ] ) and ( [ optimb ] ) allows us to interpret @xmath72 as a current noise .however , this noise is no longer stochastic , but rather it follows the deterministic path solution of ( [ 2.8 ] ) .we will , therefore , in the following refer to @xmath70 as the optimal noise .@xmath70 corresponds to the most likely value the noise takes given the set of spiking times . solving ( [ 2.8 ] )shows that the optimal noise is an exponential function of the time : @xmath73 where @xmath74 is a constant , which we call noise coefficient .it may happen that the optimal potential only reaches the threshold without actually crossing it at intermediate times .when this is the case , the optimal potential equals @xmath75 and its derivative with respect to the time vanishes .the expression for the optimal noise can be then read from ( [ optimb ] ) , and is given by @xmath76 equation ( [ eq2 ] ) ensures that the potential does not cross the threshold value at a time @xmath77 . despite their apparent simplicity , eqns ( [ optimb],[eq1],[eq2 ] )are not easy to solve , due mainly to the interplay between the two regimes , @xmath78 and @xmath79 , mentioned above . the determination of @xmath64 was achieved numerically by paninski for a single neuron ( paninski , 2006 ) .we now sketch the procedure to determine @xmath64 rapidly , even for tens of neurons .the procedure relies on the search for contacts , that is , times at which the optimal potential touches the threshold .there are two types of contacts : contacts coinciding with a synaptic input ( the potential touches the threshold at time @xmath15 ) , and contacts arising in between two inputs . in the absence of leakage ,only the former type of contacts matter , and a search procedure to locate those isolated - time contacts was proposed by cocco , leibler and monasson ( 2009 ) . in the presence of leakage ,both types of contacts have to be taken into account .the search procedure is more complex , and is explained below .[ fig - fluctuphi ] [ fig - compar - optpath ] [ fig - scheme ] we assume in this section that the couplings and currents are known . consider neuron @xmath5 at time @xmath80 $ ] , where @xmath81 is the index of the isi .the initial and final conditions for the optimal potential are : @xmath82 and @xmath83 . in between, @xmath84 obeys the lif evolution equation ( [ optimb ] ) with an optimal ` noise ' @xmath68 .@xmath70 can be interpreted as a non - stochastic , external , time - dependent current to be fed into the neuron in order to drive its potential from 0 to @xmath26 , given the synaptic couplings .the expressions for the optimal ` noise ' are given by ( [ eq1 ] ) when @xmath85 , and ( [ eq2 ] ) when the optimal potential @xmath64 is equal to the threshold value .when @xmath64 reaches the threshold at a time coinciding with an incoming spike , the coefficient @xmath74 in ( [ eq1 ] ) may abruptly change through an _ active contact _ ; the notion of active contact is illustrated in the simple case of a neuron receiving a single spike in appendix [ illustration ] .the potential @xmath64 may also touch the threshold without crossing it , and the noise may remain constant over some time interval ; we call such an event _ passive contact_. that the potential can brush , or remain at the threshold level without producing a spike is made possible by the @xmath86 limit .we will discuss later on the validity of this calculation , and how to modify it when the noise standard deviation , @xmath22 , does not vanish .both types of contacts are shown in fig .[ fig - scheme]a .let us explain how the positions of active and passive contacts can be determined .let @xmath87 be the emission times of the spikes arriving from the neurons interacting with @xmath5 during the time interval @xmath88 $ ] , and @xmath89 the corresponding synaptic strengths and @xmath90 can apparently arrive at the same time ; if so , we consider , based on model ( [ ode ] ) and ( [ defisyn ] ) , that a single input of amplitude @xmath91 enters the neuron . ] .let @xmath92 be the initial value of the potential , and @xmath93 be the index of the first input spike .if the time is small enough the optimal potential is surely below the threshold value . according to ( [ eq1 ] )the optimal noise is an exponential with noise coefficient @xmath74 , and the optimal potential is obtained by solving ( [ optimb ] ) with the result , @xmath94 where @xmath95 is the heaviside function .it is tempting to look for the value of @xmath74 such that a spike is emitted at time @xmath96 , defined by the implicit equation @xmath97 .however , the corresponding potential might not be below threshold at all intermediate times @xmath98 .instead , we look for the smallest noise capable of driving the potential from its initial value @xmath99 into contact with the threshold : @xmath100 as the potential ( [ defv ] ) is a monotonically increasing function of @xmath74 , a value of the noise smaller than @xmath101 would not be able to bring the potential to the threshold and to trigger a spike at any time , while a value larger than @xmath101 would violate the condition that the potential can not cross the threshold on the time interval @xmath98 , see appendices [ seccp - sec ] and [ seccp2 ] .we denote by @xmath102 the time at which the threshold is reached : @xmath103 . the solution to the minimization problem ( [ defnoise ] )can be found following the procedure described below . briefly speaking , the procedure identifies candidates for the contact points , selects the best one , and is iterated until the isi is completed . *_ active candidates : _ we first consider the possibility that the contact time @xmath104 coincides with a synaptic input .we therefore calculate for each @xmath105 , the root @xmath106 of the implicit equation @xmath107 .the smallest of those @xmath108 noise coefficients is called @xmath109 . * _ passive candidates : _ we then consider the case where the contact time @xmath104 may not be simultaneous to any input , but rather fall between two successive spikes . for each @xmath110 , we look for a noise coefficient @xmath111 and a contact time @xmath112 $ ] fulfilling the set of coupled equations @xmath113 , expressing that the potential reaches and does not cross the threshold .these two equations can be solved analytically , see expressions ( [ etap ] ) and ( [ etap2 ] ) in appendix [ secgnon ] .we call @xmath114 the smallest noise coefficient corresponding to those possible passive contacts . *_ selection of the best candidate : _ if @xmath115 , the contact is active and takes place at time @xmath116 for a certain @xmath117 comprised between @xmath118 and @xmath119 ( fig .[ fig - scheme]a ) . the optimal potential and noise in the time interval @xmath120 $ ]are given by , respectively , eqns ( [ defv ] ) and ( [ eq1 ] ) with @xmath121 . if @xmath122 , the contact is passive , and takes place in the time interval @xmath123 $ ] for a certain @xmath124 comprised between @xmath118 and @xmath119 .the potential will remain equal the threshold , and the noise will remain constant according to ( [ eq2 ] ) over a finite time interval @xmath125 $ ] , after which both @xmath64 and @xmath70 resume their course ( fig .[ fig - scheme]a ) .@xmath126 is the smallest delay allowing the potential to be in active contact with the threshold at a later time @xmath127 , with @xmath128 .the correctness of this statement is ensured by the fact that there can be at most one passive contact between two active contacts ( appendix [ secgnon ] ) ; hence , a passive contact is necessarily followed by an active contact ( or by the spike at the end of the isi ) . for every integer @xmath129 comprised between @xmath130 and @xmath119 , we calculate analytically the delay @xmath131 such that the potential reaches the threshold in @xmath132 , see eqn ( [ eqdeltac ] ) in appendix [ secgnon ] ; the smallest among those delays and the corresponding value of @xmath129 are , respectively , @xmath133 and @xmath117 . * _ iteration : _ we are left with the calculation of @xmath70 and @xmath64 on the remaining part of the inter - spike interval , @xmath134 $ ] .to do so , we iterate the previous steps .we first update @xmath135 , @xmath136 , @xmath137 in ( [ defv ] ) , and look for the lowest noise producing a new contact over the interval @xmath138 $ ] using ( [ defnoise ] ) again . the procedure is repeated until the whole inter - spike time interval is exhausted . as a result a sequence of values for @xmath101 is built , each value corresponding to the noise coefficient ( [ defnoise ] ) between two successive active contact points . the lif dynamical equation ( [ ode ] ) involves quantities , such as the membrane potential , the membrane conductance , the input current , which have different physical units .a straightforward algebra shows that ( [ ode ] ) is equivalent to the following differential equation , @xmath139 which involves only dimensionless variables ( denoted with overbars ) : @xmath140 the noise has zero mean , and covariance @xmath141 , where @xmath142 intuitively , we expect that the potential @xmath4 will not depart much from the optimal path @xmath64 , and , hence , that our inference algorithm will be accurate if the dimensionless standard deviation of the noise , @xmath143 , is small .we illustrate this claim on the simple case of a neuron receiving a few inputs from two other neurons during two inter - spike intervals ( isi ) of length @xmath144 , see fig .[ fig - compar - optpath]b .the times of the input spikes were randomly chosen , once for all , before the simulations started .then , we numerically integrated the lif equation for the potential ( [ ode ] ) for @xmath145 random realizations of the noise @xmath146 .the realizations such that the neuron spiked twice , with isis falling in the range @xmath147\times \tau$ ] were considered as successful .the number of successful realizations was comprised between @xmath148 and @xmath149 , depending on the noise level , @xmath22 .we show in fig .[ fig - compar - optpath]b the paths of the potential and of the noise , averaged over successful realizations , and compare them to the optimal potential , @xmath150 , and noise , @xmath151 .as expected the agreement is very good for small @xmath22 .we now make this observation quantitative .consider the @xmath16 inter - spike interval @xmath58 $ ] of neuron @xmath5 .the optimal potential @xmath64 is the time - course followed by the lif membrane potential @xmath4 in the @xmath152 limit . when the noise variance is not vanishing, the potential @xmath4 can slightly deviate from the optimal path ( fig .[ fig - fluctuphi]c ) .deviations are null at the extremities of the inter - spike interval due to the boundary constraints on the potential . a measure of the magnitude of the fluctuations of the potential is thus given by the variance of @xmath153 at the middle of the isi , _ i.e. _ @xmath154 ( fig .[ fig - fluctuphi]c ) .this variance can be calculated when the constraint that the fluctuating potential @xmath155 does not cross the threshold at times @xmath156 is relaxed , see appendix [ app - corrections ] .we obtain @xmath157 where @xmath17 is the membrane leaking time . as expected , if @xmath158 is small , so are the fluctuations of the potential around the optimal path . however , the reverse statement is false .consider , for instance , the case of a perfect integrator , for which the dimensionless @xmath158 ( [ defnounit ] ) is virtually infinite . sending @xmath159 in ( [ result - fluctu ] ), we obtain @xmath160 hence , the relative fluctuations of the potential are small if the typical amplitude of the electrical charge entering the neuron during the isi due to the noise , @xmath161 , is small compared to the total charge @xmath162 necessary to reach the threshold from the rest state .it is interesting to note that this statement applies to the lif , too .whatever the level of the noise , @xmath143 , the relative fluctuations of the potential ( [ result - fluctu ] ) can be made small if the duration of the isi is short enough compared to the membrane leaking time , @xmath17 . for large values of @xmath22 ,a discrepancy between the optimal potential and the potential obtained in simulations appears ( fig .[ fig - pro]b ) .a general observation is that the optimal potential calculated by the fixed threshold procedure can get very close to @xmath26 , while the true potential stays further away from the threshold to avoid premature firing . to further illustrate this effect ,consider a system of two if neurons , 1 and 2 , both fed with an external current .in addition , neuron 1 receives positive inputs from neuron 2 ( @xmath163 ) , and neuron 2 is independent from the activity of neuron 1 ( @xmath164 ) . in presence of a strong noise, the optimal potential calculated from the fixed threshold procedure quickly reaches a stationary value close to @xmath26 , while the random potential obtained from simulations fluctuates around a much lower level ( fig .[ fig - pro]e ) .the presence of a strong noise biases the membrane potential to lower values to prevent early spiking . a heuristic approach to reproduce this biasconsists in decreasing the threshold from @xmath26 to a time- and context - dependent value , @xmath165 .we now explain how this moving threshold , @xmath165 , is determined .consider first a neuron with no synaptic input , fed with an external current @xmath166 , during the inter - spike interval @xmath58 $ ] .we call @xmath167 the probability that the potential , taking value @xmath168 at time @xmath169 , remains below the threshold at any larger time @xmath6 , with @xmath170 .this probability depends on the current @xmath166 , and can be expressed for an arbitrary level of noise , @xmath22 , as a series of parabolic cylinder functions ( alili , patie and pedersen , 2005 ) .figure [ fig - pro]d show @xmath171 as a function of @xmath168 for some characteristic values of the parameters .the probability of survival , @xmath171 , sharply decreases to zero when @xmath168 gets close to the threshold , @xmath172 .we model this outcome by the following approximation , which involves a new , effective threshold @xmath165 : we consider that the processes starting from a value of the potential @xmath173 will not survive for a time delay @xmath174 . in other words , the true threshold , @xmath26 , is changed into a moving threshold , which is a function of the current @xmath166 , the time @xmath174 , and the parameters @xmath175 . a simple way to define @xmath165 is to look at the intersection of the tangent line to @xmath171 in @xmath172 with , say , the @xmath176 line to @xmath177 have been tried , do not qualitatively affect the results presented later in this article . ] ; the resulting expression for @xmath165 is given in appendix [ essai2 ] .figure [ fig - pro]e shows the output of the moving threshold procedure on the simple 2-neuron system described above .the optimal potential , pushed down by the moving threshold @xmath165 is much lower than in the fixed threshold approach and in much better agreement with the random realization of the membrane potential . more details are given in section [ secsubmt ] . to account for the existence of synaptic inputs, we may choose the parameter @xmath166 entering the calculation of @xmath171 and @xmath178 to be the value of the effective current @xmath179 , rather than the external current @xmath18 itself .here , @xmath180 is the average firing rate , defined as the number of spikes fired by neuron @xmath14 divided by the duration @xmath33 .contrary to the external current @xmath18 , the effective current @xmath181 takes into account the ( average ) input current coming from other neurons .this choice was done in the numerical experiments reported in the results section . to further speed up the calculations, we derive the value of @xmath165 for discrete - values delays @xmath174 only ; in a discrete interval , @xmath165 is kept to a constant value .alternative heuristic approaches to deal with the presence of moderate noise can be proposed . in appendix [ essai2 ]we introduce a cost - function for the effective current , whose effect is also to decrease the optimal potential .these approaches are effective when the optimal potential calculated by the fixed threshold procedure quickly saturates to a level close to @xmath26 . more precisely , we expect the moving threshold procedure to be efficient if the membrane leaking time is smaller or comparable to the isi , and the leaking current , @xmath182 , is larger or equal to the external current , @xmath166 .the fixed or moving threshold procedures allow us to calculate the optimal paths for the potential and the noise , given the couplings and currents .knowledge of those paths gives us also access to the logarithm of the likelihood @xmath47 in the @xmath152 limit , @xmath183 = -\frac 12 \sum _ { i ,k } \int _ { t_{i , k } } ^{t_{i , k+1 } } dt \ , \eta_i^*(t)^2\end{aligned}\ ] ] since @xmath184 in ( [ deflpcc ] ) involves the sum over different neurons , the maximization over the couplings @xmath13 and the current @xmath18 of neuron @xmath5 can be done independently of the other couplings @xmath185 and currents @xmath186 ( @xmath187 ) .formally , we are left with @xmath1 independent inferences of the most likely couplings and current for a single neuron , in presence of the spikes emitted by the @xmath188 other neurons . as a consequence neurons ` decouple ' in the inverse problem : the couplings @xmath13 and the current @xmath18 of neuron @xmath5 can be inferred independently of the other couplings @xmath185 and currents @xmath186 ( @xmath187 ) .@xmath65 defined in ( [ pathint2 ] ) is a negative - semidefinite quadratic function of its arguments @xmath189 .it is thus a concave function of the couplings and the currents .this property holds for @xmath184 ( [ deflpcc ] ) ( boyd and vandenberghe , 2004 ) . in order to infer the most likely current @xmath18 and couplings @xmath13 , we start from an arbitrary initial value _ e.g. _ @xmath190 .the full path of the optimal noise , @xmath68 , over all the inter - spike intervals @xmath81 of neuron @xmath5 , is calculated following the above procedure .we then update the couplings and the current using the newton - raphson method to maximize @xmath191 , _i.e. _ to minimize the integral of the squared optimal noise , see ( [ deflpcc ] ) .convergence follows from the concavity property stated above .the procedure requires the expressions for the gradient and the hessian matrix of @xmath191 with respect to the couplings @xmath13 and the current @xmath18 , which can be calculated exactly from ( [ deflpcc ] ) and ( [ defv ] ) .note that @xmath191 is piecewise continuously twice - differentiable ; while the gradient is continuous for all @xmath13 and @xmath18 , the hessian matrix is bounded and negative , and may discontinuously jump due to a change of the contact points .knowledge of the hessian matrix is also important to determine how reliable are the values of the inferred parameters .when the variance of the noise , @xmath192 , vanishes the inferred parameters can not deviate from their most likely values .however , for small but non zero @xmath22 , deviations are possible .the reason is that the corrections to the log - likelihood @xmath184 , to the lowest order in the noise variance @xmath61 , do not depend on the current and interactions ( appendix [ app - corrections ] ) . ] .the probability for such deviations can be estimated from the expansion of @xmath184 around its maximum .we introduce for each neuron @xmath5 , the @xmath1-dimensional vector @xmath193 whose components are : @xmath194 , and @xmath195 for @xmath196 .the multiplication of the current by the membrane leaking time ensures that all components can be expressed in units of a coupling .similarly we call @xmath197 the vector obtained when the current and couplings take their most likely values , that is , maximize @xmath184 .let us call @xmath198 the hessian matrix of @xmath184 .the parameters @xmath199 are normally distributed around their most likely values , with a covariance matrix given by @xmath200 ^{-1}_{j , j ' } \ .\ ] ] in particular , the error bars on the inferred parameters are given by the diagonal elements of the inverse of @xmath201 .note that , if the value of @xmath22 is not known , formulas ( [ fluch ] ) and ( [ flucov ] ) can still be used to compare the error bars between each other .as the entries of @xmath201 scale linearly with the duration @xmath33 of the recording , or , more precisely , the number @xmath0 of recorded spikes the uncertainty on the inferred parameters will decrease as @xmath202 .a detailed spectral analysis of @xmath203 in the case of weak couplings , reported in appendix [ aeigen ] , shows that the largest eigenvalue , @xmath204 , is related to the fluctuations of the effective current , @xmath205 where @xmath206 is the average firing rate of neuron @xmath14 , calculated over the time scale @xmath207 preceding a spike of neuron @xmath5 .the smallest eigenvalue , @xmath208 , corresponds to the fluctuations of the current @xmath18 alone . in other words ,the uncertainty on the inferred value for @xmath181 is much smaller than the one on the current @xmath18 .the intermediate eigenmodes describe correlated fluctuations of the couplings .explicit expressions for the largest and smallest eigenvalues , @xmath209 and @xmath208 , are derived in appendix [ aeigen ] .when a small change of @xmath41 and @xmath210 causes a modification of the set of contact points the second derivative of @xmath184 may be discontinuous .a simple illustration is provided by the the case of a single input , whose log - likelihood @xmath184 is reported in appendix [ illustration ] .if the maximum is located at , or very close to the boundary dividing two or more sets of contacts , the value of the hessian matrix will depend on the direction along which the maximum @xmath211 is approached .this phenomenon is also encountered in the analysis of real data , see section [ sec - inf - neg ] .in this section , we test our inference procedure on synthetic data generated from networks with known interactions and currents . we compare the results obtained from our two inference algorithms , the fixed and moving threshold procedures , respectively defined in sections [ sec - algo ] and [ sec - beyond ] ..5 cm [ fig - error ] we first consider @xmath1 ( ranging from 20 to 160 ) neurons , with no leakage ( @xmath212 ) .the neurons are uncoupled ( @xmath213 for @xmath214 ) , and fed with identical currents ( @xmath215 for all @xmath5 ) .the choice of the noise variance , @xmath61 , is specified later .the lif equation is solved numerically , using a fourth - order runge - kutta integration scheme .we choose the elementary time step to be @xmath216 sec , while the average duration of the isi is @xmath148 to @xmath217 longer . for each realization of the noise , the simulation is run until a set of @xmath218 spikes is generated .we then use the first @xmath0 spikes in this set to infer the currents and the couplings ( not fixed to zero _ a priori _ ) with the fixed threshold procedure .the algorithm stops if the log - likelihood @xmath184 increases by less than @xmath219 after an iteration of the newton - raphson procedure .alternatively , the algorithm may halt when the overall change in the couplings and current becomes smaller than a certain _ a priori _ bound .figures [ fig - time]a&b show how the running time scales with , respectively , the number @xmath0 of spikes , and the number @xmath1 of neurons . the empirically found scaling , @xmath220 , can be understood as follows .consider one neuron , say , @xmath5 .the number of spikes of neuron @xmath5 is , on average , equal to @xmath221 , where @xmath33 is the duration of the recording and @xmath222 is the average firing rate .the number of contact points , @xmath223 , is found to scale as the number of spikes , @xmath224 .the calculation of the contribution to the hessian @xmath225 coming from the interval between two successive contact points of @xmath226 takes @xmath227 time .the total calculation of @xmath225 thus requires @xmath228 operations over the time required for the inversion of the hessian matrix is equal to @xmath229 , and is generally much larger than one .the reason is that the number of parameters to be inferred , @xmath1 , has to be smaller than the number of constraints over the optimal potential , @xmath223 . for the real data analyzed in section [ sec - real ], we have @xmath230 and @xmath231 for , respectively , dark and natural movie data sets . ] . the loop over the neuron index , @xmath5 ,gives an extra ( multiplicative ) factor @xmath1 .the running time of the moving threshold algorithm grows as @xmath232 , too .however the proportionality constant is generally higher than for the fixed threshold procedure , due to the extra computational burden to calculate @xmath165 . for fixed@xmath1 and @xmath0 , the running times of both procedures increase with the number of contacts , _e.g. _ when the membrane conductance @xmath9 increases .this effect is described in section [ sec - real ] .we define the inference errors as the root mean square of the difference between the inferred parameters , @xmath233 and the true values , @xmath234 : @xmath235 together with a similar definition for the effective current , @xmath236 , with @xmath237 replaced with the inferred value for @xmath238 .the inference errors depend on the dimensionless noise ratio , changing the value of the current @xmath166 amounts to changing the time - scale of the evolution of the potential in ( [ ode ] ) .hence , the errors @xmath239 depend on the parameters @xmath240 through the value of @xmath241 only ( as long as @xmath242 ) .] , @xmath243 figure [ fig - error]c shows the inference errors found for different noise ratios @xmath241 , and their dependence on the number @xmath0 of spikes , in the absence of membrane leakage . for small data sets ,the inference error is mainly due to the imperfect sampling .as the number @xmath0 of spikes increases , @xmath244 decreases as @xmath202 , as expected from section [ sec - accuracy ] .when @xmath0 is very large , the errors saturate to a residual value , @xmath245 . the presence of the residual error @xmath246 results from the dominant - path approximation done in our calculation of the likelihood @xmath47 .the value of @xmath245 decreases with @xmath241 as expected .the cross - over between the sampling - dominated and residual error regimes takes place for a certain value of the number of spikes , @xmath247 . both @xmath247 and @xmath245depend on the observable , _ i.e. _ @xmath248 , and on the noise ratio @xmath241 . with the values of @xmath0reached in the simulations , the onset of the cross - over is clearly visible for @xmath249 , can be guessed for @xmath166 , and is not observable for @xmath250 .the existence of a cross - over , and an estimate of @xmath247 can be derived from the discussion of section [ sec - accuracy ] .when @xmath0 is large , the _ a posteriori _ distribution of the inferred parameter , @xmath251 or @xmath250 , becomes gaussian , with a variance @xmath252 where @xmath253 is the eigenvalue of the hessian matrix of @xmath184 attached to the fluctuations of the parameter @xmath254 .the inference error sums up contributions coming from both the sampling fluctuations and the residual error .the cross - over takes place when both contributions are comparable , @xmath255 , that is , for @xmath256 figure [ fig - error]c confirms that @xmath247 diminishes with @xmath22 ( or @xmath241 ) , and is much smaller for @xmath249 than for @xmath166 ( as expected from the dependence on the eigenvalue @xmath253 ) ; moreover , the residual error on the couplings is extremely small ( or might be even zero ) .as a conclusion , our inference algorithm is very accurate in the absence of membrane leakage . with @xmath148 spikes per neuron only and @xmath257 , for instance , the errors on the currents and on the couplings are , respectively , @xmath258 and @xmath259 . even in the presence of strong noise ( @xmath260 ) , and with the same number of spikes per neuron , the errors on the effective currents and on the couplings are less than 1% ..3 cm [ fig - error - inf - true ] [ fig - coupled ] [ fig - mb0 ] [ fig - mb ] [ fig - negpos ] we now study the ability of the algorithm to infer the interactions between coupled neurons .to do so , we consider random connection graphs built in the following way ( bollobas , 2001 ) .we start from a complete oriented graph over @xmath1 neurons , and erase each one of the @xmath261 link with probability @xmath262 , independently of each other .the removal process is not symmetric : the link @xmath263 may be removed , while the connection @xmath264 is preserved . at the end of the construction process , the average number of outgoing ( or incoming ) neighbors of a neuron is @xmath265 .each existing connection is then assigned a synaptic weight , uniformly at random over the interval @xmath266 $ ] .all neurons receive the same external current @xmath166 .in addition , the membrane conductance , @xmath9 , is now different from zero .the values of @xmath267 , and @xmath22 are chosen so that the network remains below saturation .we have also performed simulations where the interaction graph is drawn as above , but each neuron @xmath5 is chosen to be either excitatory or inhibitory with equal probabilities .the outgoing interactions from @xmath5 have all the same sign , and random amplitudes in @xmath268 $ ] .the performance of our inference algorithms are qualitatively similar for both models .figure [ fig - errorvsp]a shows the error on the couplings inferred with the fixed threshold algorithm , @xmath269 , as a function of the fraction @xmath270 of connections , for three values of the membrane conductance over current ratio .the error roughly increases as @xmath271 , that is , the number of connections in the network .this scaling suggests that much of the inference error is due to non - zero couplings .this finding agrees with fig .[ fig - error]c , which showed that the inferred interactions between uncoupled neurons was very small in the @xmath212 case . to better understand the performance of the algorithm, we compare in fig .[ fig - error - inf - true]c the inferred interactions @xmath13 with their true values for the 1560 oriented pairs @xmath264 of a randomly drawn network of @xmath272 neurons , with @xmath273 and @xmath274 .when the ratio @xmath275 is small compared to unity , the quality of the inference is very good ( fig .[ fig - error - inf - true]c-1 ) . for larger ratios @xmath275the inferred couplings are still strongly correlated with their true values , but are approximately rescaled by an overall factor @xmath276 , corresponding to the average slope of the linear regression in fig .[ fig - error - inf - true]c-2 .as @xmath275 increases , this factor decreases and the inference error grows ( fig . [ fig - errorvsp]a ) .figure [ fig - coupled]b shows that the inference error on the interactions increases not only with @xmath275 but also with the noise ratio @xmath241 . for large values of @xmath241 ,the network can sustain activity even when @xmath277 , and the inference error can take large values ( upper curve in fig .[ fig - coupled]b ) . in this regime , the couplings found by the fixed threshold algorithm become small , and the inferred current @xmath18 gets close to @xmath278 .the corresponding potential @xmath84 rises sharply , in a time @xmath17 , to a value slightly below threshold , @xmath279 , with small fluctuations due to the synaptic inputs .this phenomenon can be seen in fig .[ fig - mb0]d , which compares the optimal potential of a neuron for two different values of membrane conductance . as discussed in the methods section ,this behavior is a consequence of the @xmath86 limit taken in the calculation of the optimal potential ; when @xmath22 , or @xmath241 , is not small , the potential is unlikely to stay close to the threshold for a long time without producing a spike , see fig .[ fig - compar - optpath]b . in the next paragraph, we analyze the results of the moving threshold inference procedure . as a conclusion ,zero couplings are perfectly inferred , while the amplitude of large ( positive or negative ) interactions can be underestimated by the fixed threshold algorithm , especially so when the noise is strong .however , the relative ordering of the interactions is essentially preserved by the inference procedure .the moving threshold procedure was tested in fig .[ fig - pro]e on an asymmetric system of two if neurons ( @xmath280 ) in the presence of a strong noise , see description in caption and section [ sec - beyond ] .while the fixed threshold procedure erroneously inferred that both interactions vanish , the moving threshold correctly inferred the sign and the order of the magnitude of the coupling : @xmath281 .the inferred currents were within 10% of their true values .these results were obtained from a large number @xmath0 of spikes to avoid finite-@xmath0 effects .the synthetic data used in fig .[ fig - coupled]b were generated with two different values of the noise ratio , @xmath241 .we estimate the relative fluctuations of the potential around the optimal path , averaged over all the inter - spike intervals in the data set , using formula ( [ result - fluctu ] ) , and find @xmath282 for all values of @xmath275 comprised between .9 and 1.25 .hence , the relative fluctuations can not be neglected when @xmath283 .figure [ fig - coupled]b shows the inference error obtained from the moving threshold algorithm as a function of the membrane conductance for that value of the noise ratio . not surprisingly , the moving threshold procedure is more accurate than the fixed threshold algorithm . in the moving threshold algorithm , the optimal potential is constrained to remain below a certain threshold , @xmath178 , which depends on the time preceding the next spike and on the effective current @xmath181 .figure [ fig - mb]e shows the values of the moving threshold @xmath178 and of the optimal potential @xmath284 for a few spike intervals of the same neuron as in fig . [ fig - mb0]d .as expected , the value of @xmath84 lies substantially further away from the threshold @xmath26 than in the fixed threshold procedure .in addition , fig .[ fig - mb]e shows a random realization of the potential @xmath4 , obtained through numerical integration of the lif differential equation ( [ ode ] ) , for the same neuron @xmath5 .although @xmath25 is stochastic , the comparison of several inter - spike intervals indicates that @xmath84 and @xmath4 are in fair statistical agreement .to investigate in more details the origin of the inference error on the couplings for large values of @xmath241 and @xmath275 , we plot in fig .[ fig - negpos]c the inferred values of the interaction @xmath13 vs. the true value for every pairs @xmath264 of a randomly drawn network of @xmath272 neurons .the interactions inferred by the fixed threshold algorithm are about ten times smaller than their true values ( fig .[ fig - negpos]c-3 ) .the use of the moving threshold procedure leads to a spectacular improvement for positive - valued couplings ( fig .[ fig - negpos]c-4 ) .while positive couplings are accurately inferred , the magnitude of negative couplings is often overestimated .these negative couplings are responsible for most of the error @xmath244 in fig .[ fig - coupled]b . from the bayesian point of view ,when @xmath17 is smaller than the average isi , negative - valued couplings are indeed intrinsically harder to infer than positive - valued ones .a positive input drives the potential closer to the threshold , which strongly reduces the isi .conversely , a negative input drives the potential down , and a spike is unlikely to occur before the potential first relaxes to its average level @xmath285 after a time of the order of @xmath17 .hence , the influence of a negative input is hardly seen in the increase of the isi when @xmath17 is smaller than the average isi .we present an analytical calculation supporting this argument in section [ sec - inf - neg ] .we now apply our algorithm to multi - electrode recordings of the ganglion cell activity of the salamander retina .two data sets were considered .the first one , hereafter referred to as dark ( data courtesy of m. meister ) , reports the spontaneous activity of 32 neurons for 2,000 seconds , and consists of @xmath286 spikes ( schnitzer and meister , 2003 ) . in the second experiment , referred to as natural movie ( data courtesy of m. berry ) , a retina was presented a 26.5 second - long movie , repeated 120 times , and the activity of 40 neurons was registered for the whole duration of 3,180 seconds ( schneidman , berry , segev and bialek , 2006 ) .natural movie includes @xmath287 spikes .the firing rates , averaged over the population of recorded neurons , have similar values in the two data sets : @xmath288 spikes / sec in dark , @xmath289 spikes / sec in natural movie .these two data sets were analyzed in a previous work ( cocco , leibler and monasson , 2009 ) with the perfect integrator model ( @xmath212 ) and the fixed threshold algorithm . in this sectionwe extend the analysis to the case of the lif model and use both the fixed and moving threshold approaches . in particular we show that the lif model is capable of inferring the asymmetry of the interactions , which is seen in the cross - correlograms but was not obtained with the perfect integrator model .moreover we discuss error bars on the inferred couplings and the fact that strong negative interactions are more difficult to infer than positive - valued couplings .we stress that the couplings we infer _ a priori _depend on the stimulus .cocco , leibler and monasson ( 2009 ) have studied how the interactions inferred with the perfect integrator model depended on the stimulus based on the analysis of two recordings on the same retina , namely the spontaneous activity and random flickering squares . an alternative approach to disentangle stimulus - induced and structural contributions to the couplingswould be to consider a time- and stimulus - dependent external current @xmath290 ( section [ realistic ] ) .the value of the membrane leaking time @xmath17 strongly affects the number of contacts and the running time of the algorithm .it takes about 40 seconds to infer the currents and the interactions from either dark or natural movie when @xmath291 sec with one core of a 2.8 ghz intel core 2 quad desktop computer , and about 10 times longer when @xmath292 msec .the number of passive contacts of the optimal potential computed by the fixed threshold procedure quickly decreases as @xmath17 increases .it is divided by @xmath293 when the membrane leaking time increases from @xmath294 msec to @xmath295 sec for both data sets . in comparison ,the number of active contacts is less sensitive to the value of @xmath17 .we find that the ratio of the number of contacts per neuron and per second over the average firing rate takes similar values for both data sets .for @xmath296 msec , this ratio is @xmath297 for dark , and @xmath298 for natural movie . the number of passive contacts is smaller with the moving threshold algorithm , while the number of active contacts remains rather unchanged compared to its value with the fixed threshold procedure . on the overall ,the running time of the moving threshold procedure is higher due to the calculation of the time - dependent threshold @xmath165 .knowledge of the variance of the noise is required for the moving threshold algorithm .the value of @xmath22 could , in principle , be determined from experimental measures of the fluctuations of the synaptic current , but is unknown for the two recorded data sets available to us .we choose @xmath22 so that the relative fluctuations of the potential around the optimal path @xmath284 are less than @xmath299 .we compute these fluctuations by averaging ( [ result - fluctu ] ) over all isi and all neurons @xmath5 in the population .the corresponding value of the dimensionless standard deviation of the noise ( [ defnounit ] ) are : for dark , @xmath300 for , respectively , @xmath301 msec ; for natural movie , @xmath302 for , respectively , @xmath301 msec .figure [ fig - amplitude]a shows the average value of the current and of the interaction strength as a function of the membrane leaking time . as expected with the fixed threshold inference procedure , we find that the average value of the couplings decreases as @xmath17 gets small .this effect varies from neuron to neuron : the closer @xmath18 is to @xmath278 , the smaller are the couplings @xmath13 . to compare the matrices of couplings@xmath303 inferred with the fixed threshold algorithm for different values of @xmath17 , we use the correlation coefficient ( hubert and baker , 1979 ) @xmath304 where @xmath305 identical matrices correspond to @xmath306 , and uncorrelated matrices give @xmath307 .@xmath308 is independent of the scale of the coupling matrices @xmath250 and @xmath90 , _ i.e. _@xmath309 for any @xmath310 ; therefore , @xmath308 is sensitive to the relative amplitudes of the couplings @xmath90 and @xmath250 and not to their absolute differences . we choose @xmath250 to be the coupling matrix in the absence of leakage and @xmath90 to be the coupling matrix for a given @xmath17 . the value of @xmath308 as a function of @xmath17 is shown in fig .[ fig - compar - j]b . even for @xmath311 msec, the coupling matrix is substantially similar to the one obtained with the perfect integrator model ( @xmath312 for dark , @xmath313 for natural movie ) . despite the overall change in the amplitude of the inferred couplings , the relative ordering of the couplings with the pair indices @xmath314is largely independent of @xmath17 , especially so for dark .however , for specific pairs of neurons , the interactions may strongly depend on @xmath17 .such a dependence effect will be illustrated in section [ sec - symmetry ] , and can be related to the temporal structure of the corresponding cross - correlograms ..2 cm .2 cm .3 cm .2 cm [ fig - compar - j ] [ fig - fm ] [ fig - scaling - tau ] [ fig - latency ] the average value of the interactions calculated by the moving threshold algorithm does not decrease when @xmath17 gets smaller , and is larger than the one obtained from the fixed threshold procedure ( fig .[ fig - amplitude]a ) . to better understand this discrepancy, we compare in fig .[ fig - fm]c the interactions inferred with both algorithms for every pairs of neurons in the dark and natural movie data sets when @xmath315 msec .the agreement between both procedures is very good for positive and strong couplings .couplings which are slightly positive with the fixed threshold procedure generally have a larger value with the moving threshold procedure .this offset is responsible for the differences in the average values of the interactions found in fig .[ fig - amplitude]a .in addition , in natural movie , negative - valued couplings often have a stronger amplitude with the moving threshold procedure .we find , in both approaches , a few negative and very strong couplings .the amplitude of those extreme couplings increases very quickly as the membrane leaking time decreases .the emergence of strong negative interactions with the lowering of @xmath17 can be related to the presence of long latencies between the emission of spikes .we define the latency of neuron @xmath5 with respect to neuron @xmath14 as the smallest delay between a spike emitted by @xmath14 and a later spike fired by @xmath5 , @xmath316 a large value of the latency of neuron @xmath5 with respect to @xmath14 is interpreted by the inference procedure as the consequence of a strongly inhibiting coupling from @xmath14 to @xmath5 .however , the effect of a synaptic input of amplitude @xmath13 on the potential @xmath25 of the neuron @xmath5 decays exponentially with the ratio of the time elapsed from the input over the membrane leaking time .hence , to keep the latency fixed while @xmath17 is changed , the strong and negative interaction must change accordingly , @xmath317 where the constant has a negative value .figure [ fig - scaling - tau]d shows the negative couplings @xmath13 vs. the latencies of the corresponding pairs @xmath314 divided by @xmath17 , for three values of @xmath17 .the outcome suggests that relation ( [ jtau ] ) is indeed correct , see fig .[ fig - scaling - tau]d and its caption .the above mechanism explains why strongly negative couplings are less frequent in dark than in natural movie . for @xmath318 msec , there are 10 interactions ( out of 1560 ) smaller than @xmath319 in natural movie , and none ( out of 992 ) in dark . for @xmath311 msec, these two numbers are equal to , respectively , 23 and 1 .figure [ fig - latency]e shows the histograms of latencies for both data sets . in natural movie , we find 17 pairs with latencies larger than 25 msec . in dark , only one pair @xmath314 has a latency larger than 25 msec .the corresponding interaction , @xmath13 , is the only one smaller than @xmath319 for @xmath311 msec . as discussed in the methods section , the uncertainty on the inferred parameterscan be obtained from the hessian matrix of @xmath184 , that is , from the curvature of the log - likelihood around its maximum .to quantify those uncertainties , we use the following procedure .assume for instance we want to know how reliable is the inferred value , @xmath320 , of the interaction @xmath321 from neuron @xmath322 to neuron @xmath5 .we fix @xmath321 to an arbitrary value , and maximize @xmath323 ( [ deflpcc ] ) over all the couplings @xmath13 with @xmath324 and over the current @xmath18 .the outcome is a function of @xmath321 , which we denote by @xmath325 and call marginal log - likelihood .@xmath326 has , by definition , a maximum in @xmath327 .its second derivative in the maximum , @xmath328 , is related to the error bar @xmath329 on the interaction through , see ( [ fluch ] ) and ( [ flucov ] ) , @xmath330 the same procedure can obviously be used to obtain the error bar on the current @xmath18 .we now illustrate this approach on the natural movie data set , and one arbitrarily chosen neuron , @xmath331 .three interactions , representative of , respectively , positive , weak , and negative couplings , were singled out among the 39 couplings incoming onto neuron 1 .figure [ fig - curvature]a shows the marginal log - likelihoods @xmath332 , @xmath333 , and @xmath334 , in addition to @xmath335 .for all four parameters , the marginal likelihoods can be approximated with parabolas in the vicinity of their maxima .estimating the second derivatives from those best quadratic fits and using ( [ eb09 ] ) , we obtain @xmath336 where @xmath158 is the dimensionless noise level defined in ( [ defnounit ] ) .hence , the error bars on the couplings and currents have very similar values .this common value depends on the noise level , @xmath143 .as discussed in the next section [ sec - amplitude ] , @xmath143 is expected to be close to , or smaller than unity when @xmath337 msec .consequently , the value for @xmath338 is compatible with zero , while the interactions @xmath339 and @xmath340 are non zero , with 99.9999% confidence . a closer inspection of fig .[ fig - curvature]a shows that the quality of the quadratic fit of @xmath325 is excellent for @xmath339 and @xmath340 , but less so for @xmath341 and @xmath338 . for the latter parameters, it seems that the curvature of @xmath325 takes two different values , depending on whether the maximum is approached from the left of from the right .this phenomenon results from the piece - wise structure of the @xmath184 function , see methods section .a practical consequence is that the errors @xmath342 and @xmath343 are not evenly distributed around zero ; for instance @xmath338 is more likely to be larger than @xmath344 than it is to be smaller .-2 cm -4.2 cm 9.cm + 9.cm [ fig - tfpt ] note that strong , negative interactions may be harder to infer than positive - valued couplings , a phenomenon already underlined by aersten and gerstein ( 1985 ) .the underlying intuition is that the duration of the isi is less affected by an inhibitory input than by an excitatory input when the membrane leaking time , @xmath17 , is small compared to the average value of the isi .we now present an analytical argument supporting this intuition . consider a neuron , fed with an external current @xmath166 and with noise variance equal to @xmath61 . assume a synaptic input of amplitude @xmath250 is received at time @xmath345 .we call @xmath346 the average value of the time at which the neuron will emit a spike ; the calculation of @xmath346 can be done using a series of parabolic cylinder functions ( alili , patie and pedersen 2005 ) . figures [ fig - tfpt]b&c shows that the dependence of @xmath346 on @xmath250 is much weaker for negative - valued @xmath250 than for positive couplings . as the set of spiking times is the only information we have at our disposal , the difficulty in inferring negative couplings is intrinsic to the bayesian approach , and can not be circumvented by any particular algorithm ..5 cm [ fig - coefficient_r ] [ fig - symmetry ] the dependence of the symmetry of couplings upon the membrane leaking time @xmath17 can be understood , to some extent , from the structure of the cross - correlograms , that is , the histograms @xmath347 of the delays @xmath348 between the times of the spikes fired by the two neurons @xmath349 in each pair .to do so , we consider two pairs of neurons in dark , called pairs @xmath350 and @xmath351 .figure [ fig - rapportj]a shows the cross - correlograms @xmath352 and @xmath353 .pair @xmath350 is characterized by a positive peak in @xmath354 , centered in @xmath345 , and of width @xmath293 msec .pair @xmath351 exhibits a positive peak of correlations , of the same width , but centered around @xmath355 msec .we plot in fig .[ fig - rapportj]b the symmetry ratios of the interactions in the pairs , @xmath356 and @xmath357 .we find that @xmath358 is , to a large extent , independent of @xmath17 .conversely , @xmath359 sharply decreases with decreasing @xmath17 and is close to zero when @xmath311 msec , which coincides with the typical delay in the cross - correlogram @xmath353 shown in fig .[ fig - rapportj]a .we conclude that the inference procedure is capable of capturing the directionality of the interaction between the neurons 1 and 22 , if @xmath17 is small enough .this results shed some light on the correspondence between the interactions inferred within the lif model and within the ising model ( schneidman , berry , segev and bialek , 2006 ; shlens et al , 2006 ) .couplings inferred with the perfect integrator model for dark are in good agreement with the ising interactions , when the time is binned into windows of width @xmath360 msec ( cocco , leibler and monasson , 2009 ) . by construction ,the ising model produces symmetric interactions from the pair - wise correlations of the activities , averaged of the binning window . in the absence of leakage , the integrate - and - fire inference algorithm hardly distinguishes between a post - synaptic and pre - synaptic firing pattern , and produces rather symmetric couplings .but as @xmath17 decreases , the lif couplings may become strongly asymmetric ( fig .[ fig - rapportj]b ) . in this case, the correspondence between the ising and lif couplings breaks down .the same phenomenon was observed in natural movie , where delays in the cross - correlograms are even stronger .in this article , we have presented a procedure to infer the interactions and currents in a network of leaky integrate - and - fire neurons from their spiking activity .the validity of the procedure was established through numerical tests on synthetic data generated from networks with known couplings .we have also applied our algorithm to real recordings of the activity of tens of ganglion neurons in the salamander retina . though our algorithm is limited to moderate noise levels and instantaneous synaptic integration, it is fast and can , to our knowledge , handle much bigger data sets than the existing inference methods for the stochastic if model .it is our intention to make this algorithm available to the neurobiology community in a near future .cross - correlation analysis ( perkel , gerstein and moore , 1967 ; aersten and gerstein , 1985 ) consists in studying the distribution of delays between the spikes of neurons in a pair .this approach has been used to characterize the connections between neurons ( amplitude , time - scale , dependence on distance ) , or their dynamical evolution ( fujisawa , amarasingham , harrison and buzsaki , 2008 ) . the analysisdo not require any combinatorial processing of the activity of a large part of the neural assembly . as a result, the approach is not limited to small networks .however , cross - correlation analysis may find difficult to separate direct correlations from indirect correlations modulated through interactions with neurons in the surrounding network ( ostojic , brunel and hakim , 2009 ; cocco , leibler and monasson , 2009 ) , or due to common inputs ( constantidinidis , franowicz and goldman - rakic , 2001 ; trong and rieke , 2008 ) . in statistical approachesa widely - used concept is the one of causality ( seth and edelman , 2007 ) .a causal interaction exists from neuron @xmath5 to neuron @xmath14 if the knowledge of the activity of @xmath5 helps predict the firing of @xmath14 beyond what can be achieved from the activity of @xmath14 alone . in practice, causal relationships are detected through linear multivariate statistical regressions ( sameshima and baccal , 1999 ) , and may overlook non - linear dependencies .causal analysis have also difficulties in evaluating the strength of the interactions . maximum entropy models , which deduce interactions from pairwise correlations only , have been shown to accurately reproduce higher - order correlations between neurons in the vertebrate retina ( schneidman , berry , segev and bialek , 2006 ; shlens et al , 2006 ; cocco , leibler and monasson , 2009 ) .these models , however , suffer from some limitations .interactions are constrained to be symmetric , and temporal correlations are partially discarded ( marre , el boustani , frgnac and destexhe , 2009 ) . in additionobtaining the interactions from the correlations may be computationally very hard for large networks , though efficient approximate algorithms have recently been developed ( cocco and monasson , 2010 ) .generalized linear models ( glm ) , which represents the generation of spikes as a poisson process with a time - dependent rate , have been applied to various neural systems ( brown , nguyen , frank , wilson and solo , 2001 ; truccolo et al , 2005 ; pillow et al , 2008 ) .the inference of parameters in the glm framework is apparently easier to solve than for if models , which has made the glm framework very attractive . whether glm are better than if models to account for real neural activity , regardless of the computational complexity of both inference framework , is an important issue ( gertsner and naud , 2009 ) .we hope that our work , which makes possible to apply the if model to large data sets , will help to answer this question .approaches to infer model parameters in the if framework have been so far capable of processing a very limited number of neurons or of spikes .pillow et al .( 2005 ) inferred the model parameters of one stochastic if neuron based on a 50 second - long recording with a procedure tolerating any level of noise ; makarov , panetsos and de feo ( 2005 ) inferred the connections between 5 deterministic if neurons from a 60 second - long synthetic spike train . in comparisonwe have analyzed a 3180-second long recording of the activity of 40 neurons .the running time of our procedure increases as @xmath361 , where @xmath33 is the duration of the recording and @xmath222 is the average firing rate .recently , koyama and paninski ( 2009 ) have proposed a numerical procedure for calculating the optimal potential and inferring the interactions . in their approach, the time is discretized into many time - bins of small duration @xmath362 , and the values of the optimal potentials at those discrete times can be found by means of the interior - point method for discrete constrained optimization problems .the running time of the procedure , @xmath363 , is approximately @xmath364 times larger than ours . in practice, @xmath222 is of the order of 1 to 10 hz , while the discretization time , @xmath362 , is of the order of 1 msec ; hence , @xmath364 ranges from 100 to 1000 .however , this order of magnitude does not take into account the existence of multiplicative constants ; a comparative test of the two approaches on the same synthetic or real data would be useful to accurately estimate their running times .furthermore , the algorithm introduced by koyama and paninski can easily incorporate the presence of temporal filtering in the interactions .our procedure is , in its present form , valid when the integration kernel is instantaneous only ; considering other synaptic kernels would require _ ad hoc _ modifications to the expressions of the optimal noise and potential and to the search procedure for contacts .one of the major assumptions in our approach is that the synaptic integration time , @xmath365 , is vanishingly small . in practice , @xmath366 does not vanish , but might often be smaller than the membrane leaking time , @xmath17 , and the average isi .assume that neuron @xmath5 , whose potential @xmath25 is close to the threshold @xmath26 , receives a spike at time @xmath6 from another neuron , @xmath14 , through a strongly excitatory connection @xmath367 .then , neuron @xmath5 will reach the threshold level after having received a charge @xmath368 , smaller than @xmath13 . as a consequence, large positive interactions can be underestimated when the latency of neuron @xmath5 from neuron @xmath14 ( [ deflat ] ) is smaller than @xmath369 . to compensate for this effectwe could introduce a time - dependent value for the interaction , @xmath370 where @xmath371 is the closest firing time of neuron @xmath5 .hence the effective interaction @xmath372 is equal to its nominal value @xmath13 only if the synaptic current has enough time to enter the neuron @xmath14 , and is a fraction of @xmath13 otherwise. the modified procedure will be correct as long as @xmath373 .if the synaptic and membrane time - scales are comparable , one needs to take into account the complete shape of the synaptic integration kernel , @xmath374 .choosing simple enough integration synaptic kernel , such as the piece - wise linear function @xmath375 if @xmath376 or @xmath377 , @xmath378if @xmath379 , could lead to tractable dynamical equations for the optimal potential and noise .the resolution of those equations is left for future work .the inference procedure that we have introduced here can be extended to include realistic features such as a refractory period , @xmath380 .to do so , we restrict the sum in ( [ defv ] ) to the spikes @xmath129 entering the neuron @xmath5 at times larger than @xmath381 .we have run the modified inference procedure on the recordings of the retinal activity , for values of @xmath380 ranging from 2 to 5 milliseconds .the couplings did not change much with respect to the values found with @xmath382 .note that the introduction of a propagation delay @xmath383 in the synaptic interaction is straightforward , as long as the integration kernel remains a dirac distribution ( centered in @xmath383 ) .bounds on the values of the couplings and currents _ e.g. _ to prevent the exponential growth of negative interactions with the leaking conductance can naturally be introduced through a prior distribution . as an example , assume that the interactions @xmath13 take values in @xmath384 $ ] .then , one could maximize @xmath385 instead of the log - likelihood @xmath184 alone , where @xmath386 if @xmath387 , 0 if @xmath388 , @xmath389 if @xmath390 and @xmath391 is a large positive coefficient .we have assumed , throughout this work , that the values of @xmath9 and @xmath26 were known . in practical situations , while the orders of magnitudes are known , the precise values of these parameters should be inferred , and could depend on the neuron @xmath5 .the inference procedure could be modified to update the values of @xmath392 and @xmath393 at the same time as the synaptic couplings @xmath13 and the current @xmath18 .the number of parameters to infer ( per neuron ) would simply increase from @xmath1 to @xmath394 , and the running time should not increase too much . nowadays , multi - electrode experiments can record a few tens , or hundreds of neurons .to which extent do the interactions inferred from this sub - population coincide with the interactions one would find from the knowledge of the whole population activity ?the question does not arise in cross - correlation analysis : the correlation between the firing activities of two neurons is obviously independent of whether a third neuron is recorded or not .however the issue must be addressed as soon as a collective model for generating the activity is assumed , such as the coupled lif models studied here .a detailed analysis suggests that the interaction between a pair of neurons is not affected by the activity of other neurons distant by more than @xmath395 m in the case of spontaneous activity ( cocco , leibler and monasson , 2009 ) .the electrode array should be at least twice longer and wider than @xmath396 , and should be dense enough to capture all the neurons on the recorded area .it is estimated that about 10% of the ganglion cells are registered in the dark experiment , compared to more than 80% with the denser but smaller electrode array used in the natural movie experiment ( segev , puchalla and berry , 2005 ) .it would thus be very interesting to repeat our study on other multi - electrode recordings , with sufficiently large and dense arrays .taking into account the stimulus @xmath0 in the inference process would also be interesting .to do so , we could add a stimulus - induced current , @xmath397 , to ( [ ode ] ) . a simple expression for thiscurrent would be @xmath398 , where @xmath399 is a kernel similar to the one used in generalized linear models ( pillow et al , 2008 ) .the expression of the current - dependent term in the potential @xmath400 ( [ defv ] ) should be modified accordingly , while the noise - dependent term would remain unchanged .it is important to note that the search procedure for contacts presented in section [ sec - algo ] would remain valid .however , the expressions of the noise coefficient , the contact time and the duration of a passive contact given in appendix [ secgnon ] for the case of a constant current @xmath166 should be rederived and would depend on the precise temporal structure of the stimulus - induced current @xmath401 .* acknowledgment : * this work originates from a collaboration with s. leibler , whom we thank for numerous and fruitful discussions .we thank c. barbieri for a critical reading of the manuscript .we acknowledge the hospitality of the rockefeller university , where this work was initiated .partial funding was provided by the agence nationale de la recherche under contract 06-jcjc-051 .[ seccp ] in this appendix , we justify the prescriptions in the search for active contacts presented in section [ sec - algo ] . for the sake of simplicity we restrict to the @xmath212 case ( no membrane leakage ); the extension to non - zero @xmath9 is briefly discussed in appendix [ secgnon ] .we consider a neuron @xmath5 , and call @xmath108 the number of spikes received by this neuron during its @xmath16 inter - spike interval @xmath402 $ ] .the arrival times are @xmath403 , and the corresponding synaptic strengths are @xmath89 . to lighten notationswe hereafter omit the index @xmath5 of the neuron . to understand the key notion of contact , we first consider the simple case of a neuron receiving no spike during the inter - spike interval @xmath405 $ ] .the optimal noise is constant according to ( [ 2.8 ] ) .equation ( [ optimb ] ) then shows that the optimal potential is a linear function of the time , which is fully determined from the boundary conditions @xmath406 .we obtain @xmath407 this solution is correct since the potential remains below the threshold at all times @xmath408 .let us now assume now that the neuron receives one input from another neuron , of strength @xmath409 at time @xmath4100;t[$ ] .the effect of the input is a discontinuous jump of the potential at time @xmath411 and of size @xmath412 , shown in fig .[ fig - sol1 ] . repeating the calculation above, we obtain the following expressions for the optimal potential and noise @xmath413 where @xmath95 is the heaviside function : @xmath414 if @xmath415 , 0 otherwise .this solution is sketched in fig .[ fig - sol1]a .it is valid when the potential @xmath416 remains below the threshold at all times .we call this situation case a. as @xmath416 is a piece - wise linear function we only need to check that @xmath417 and @xmath418 are both smaller than @xmath26 .the two conditions are fulfilled provided that @xmath419 what happens when the above condition is violated ?let us consider first @xmath420 ( referred to as case b hereafter ) .then @xmath416 exceeds the threshold @xmath26 before the input enters the neuron . to prevent the potential from crossing the threshold at time @xmath411 , the true optimal noise , @xmath421 ,should be smaller than @xmath422 .but , if @xmath423 , the potential could not reach @xmath26 when the neuron emits its spike at time @xmath33 according to the very definition of @xmath424 !the only way out is that @xmath425 takes two different values corresponding to the two sub - intervals @xmath426 and @xmath427t_1;t]$ ] , which we call , respectively , @xmath428 and @xmath429 .we expect @xmath430 .the noise can change value in @xmath431 through ( [ optimc ] ) only if the potential reaches the threshold in @xmath411 .we find that @xmath432 from the boundary conditions @xmath433 , and @xmath434 from the boundary conditions @xmath435 .this solution is drawn in fig .[ fig - sol1]b .it is important to stress that the above solution is based on the capability of the noise to abruptly change its value when the potential touches the threshold in @xmath431 . a detailed study of the behavior of the noise close to such ` contact points ' proving that this is indeed the case is postponed to appendix [ seccp - sec ] .finally , we turn to case c corresponding to @xmath436 . in this casethe input is so excitatory that the noise has to be negative to prevent the neuron from emitting a spike at a time @xmath437 . as in case b, the potential reaches the threshold in @xmath431 to allow the noise to change its value after the input has entered the neuron .we find @xmath438 according to the boundary conditions @xmath439 .right after the spike has been received , the potential has reached its threshold value , and will keep to this value until a spike is emitted at time @xmath33 , hence @xmath440 this solution is drawn in fig .[ fig - sol1]c .we now give the values of log - likelihoods @xmath184 corresponding to the cases listed above .the value of @xmath184 can be calculated from the knowledge of the optimal noise @xmath101 through ( [ deflpcc ] ) . in the case of @xmath404 spike , we find , using ( [ illm0 ] ) with @xmath441 , @xmath442 the optimal currentis then inferred by maximizing @xmath443 with the result @xmath444 , which corresponds to a vanishing value for the optimal noise , as expected .when @xmath445 spike is received by the neuron , the log - likelihood @xmath184 has three distinct expressions corresponding to the case a , b , c discussed in section [ illustration ] .the resulting expression is ( with @xmath446 ) : @xmath447 the log - likelihood @xmath184 is a continuous and convex function of its argument .the first derivatives of @xmath184 are continuous in @xmath448 , but the second derivatives are not .the noise coefficient @xmath74 in ( [ eq1 ] ) are constant over the time interval separating two active contacts .the value of @xmath74 may however change upon the crossing of an active contact .the scope of this section is to show that the noise right after the contact can take _ any value _larger than the noise immediately before the contact .this monotonicity property justifies the search for the minimal noise coefficient done in ( [ defnoise ] ) , see appendix [ seccp2 ] . to show that the noise always increases through an active contact, we consider that the synaptic integration is not instantaneous , but takes place over a finite albeit small time , @xmath369 .we thus replace the expression for the current @xmath449 in ( [ defisyn ] ) with @xmath450 where @xmath13 is the strength of the connection from neuron @xmath14 onto neuron @xmath5 , and @xmath451 is is the memory kernel of the integration of synaptic entries ( top panel in fig .[ fig - cont ] ) .we assume that @xmath451 vanishes for @xmath452 and for @xmath453 where the integration time @xmath369 is independent of the pair @xmath314 .in addition , @xmath454 is positive , and its integral over the interval @xmath455 $ ] is equal to unity .we consider the case of a single incoming spike , as in section [ illustration ] .we want to show that , in the @xmath456 limit , the only constraint linking the values @xmath457 and @xmath458 of the optimal noise , respectively , before and after a spike entering at @xmath411 , is @xmath459 , as we have found for a single incoming input in cases b and c. to do so , we assume that the time of synaptic integration @xmath365 is small but _ finite _ , and consider case b. the dynamics of @xmath150 and @xmath101 can be divided in several steps , whose numbered are reported on fig .[ fig - cont ] : 1 .prior to the input , _i.e. _ at times @xmath460 , the optimal noise @xmath457 is constant and the optimal potential @xmath150 is a linear function of the time , with slope @xmath461 , as shown in fig .[ fig - cont](left ) .2 . a strongly negative input of amplitude @xmath462 is then received by the neuron between times @xmath411 and @xmath463 .the derivative of the potential now decreases with the time until it vanishes at time @xmath102 defined through @xmath464 3 . if the value of @xmath457 is chosen so that @xmath465 , the potential tangentially reaches the threshold at @xmath102 ( contact point ) .then , the potential remains constant and equal to @xmath26 .the noise obeys eqn .( [ eq2 ] ) and , therefore , increases until it reaches the prescribed value , @xmath466 , at time @xmath467 such that @xmath468 see bottom panel in fig .[ fig - cont](left ) .4 . then the potential starts decreasing from its threshold value through eqn .( [ optimb ] ) , and reaches a minimum in @xmath469 , solution of the same equation ( [ bes ] ) as @xmath470 , see fig .[ fig - cont](top left ) .5 . at later timesthe derivative of the potential is positive from eqn .( [ optimb ] ) , and increases until time @xmath471 , coinciding with the end of the synaptic integration .6 . at times larger than @xmath471 ,the potential keeps growing with a constant slope equal to @xmath472 .in the @xmath473 limit , all times @xmath474 tend to the same value , that is , the time @xmath411 .more precisely , as the slope of @xmath454 is of the order of @xmath475 ( in absolute value ) , and @xmath476 are finite ( @xmath477 ) , then for @xmath478 , @xmath479 differ from each other by @xmath480 .hence the change in the potential @xmath150 between @xmath470 and @xmath471 equals @xmath481 .we conclude that , for @xmath482 the potential becomes a discontinuous function of time with a discontinuity @xmath412 .in addition , the noise @xmath101 can also jump abruptly from its value @xmath457 at @xmath483 to any _ larger _ value @xmath466 at time @xmath484 since the maximum of @xmath454 tends to infinity when @xmath482 .note that the drawing of fig .[ fig - cont](left ) tacitly assumes that @xmath485 . a hypothetic scenario would be that the noise exactly compensates the synaptic input for a longer time interval ( including the top of @xmath454 ) , while the potential would remain equal to @xmath26 . in this case, the peak value of the noise would be @xmath486 .the contribution to the integral ( [ deflpcc ] ) would be of the order of @xmath487 and would diverge in the @xmath482 limit .hence this possibility is precluded .the above discussion is straightforwardly extended to case c. the optimal potential and noise are sketched in fig .[ fig - cont](right ) .note that the contact interval spreads beyond @xmath488 $ ] in this case . in the generic case of more than one incoming spikes ,the contact interval is restricted to @xmath489 $ ] as in case b. the noise can also discontinuously change from its value @xmath490 before the contact to any larger value , @xmath466 , after the contact .we now consider the general case of @xmath108 input spikes .let @xmath492 be , respectively , the initial value of the potential and the index of the first input spike .we define the piece - wise linear function solution of ( [ optimb ] ) for a constant noise @xmath74 , @xmath493 we are looking for the smallest value of the noise coefficient @xmath74 capable of bringing the potential @xmath494 from its initial value @xmath495 to the threshold .the contact time , @xmath102 , coincides with an entering spike , _i.e. _ @xmath496 for some @xmath497 . if @xmath498 then the optimal potential is @xmath499 throughout the inter - spike interval @xmath500 $ ] , and the problem is solved . if @xmath501 , @xmath127 is the first active contact point of the potential . @xmath101 and @xmath502 are , respectively , the optimal noise and potential on the interval @xmath120 $ ] . the correctness of the above statement can be established using a proof by contradiction .* assume that the optimal noise , @xmath503 , is smaller than @xmath101 on some sub - interval of @xmath504 $ ] .remark that the potential @xmath168 in ( [ defvlin ] ) is an increasing function of the noise , @xmath505 for all @xmath506 . by virtue of ( [ monoto ] ) and the definition of @xmath101, @xmath507 can not touch the threshold at any time so the noise is constant throughout the interval @xmath504 $ ] .hence no active contact can take place at time @xmath102 . as @xmath101 is the minimal value of the noise which can drive the potential into contact with the threshold over @xmath500 $ ], we conclude that @xmath507 can not cross the threshold at any time @xmath508 .the neuron can therefore not spike at time @xmath96 . *conversely , suppose that the optimal noise is equal to @xmath509 on the interval @xmath510 $ ] with @xmath511 , and takes another value on the interval @xmath512 $ ] . as the change of noise can take place only through an active contact , and the change is necessarily positive ( section [ seccp - sec ] ), we have @xmath513 . applying ( [ monoto ] ) to the interval @xmath512 $ ] , we have @xmath514 adding the value of the optimal potential in @xmath515 to both members of the previous inequality , we find @xmath516 where the last inequality comes from ( [ monoto ] ) .but , by definition of @xmath101 , @xmath517 .hence , we find that the optimal potential in @xmath102 is above threshold , which can not be true. the optimal noise and potential on the remaining part @xmath518 $ ] of the inter - spike interval can be determined iteratively .we replace @xmath519 with @xmath127 and @xmath99 with @xmath26 if @xmath520 or @xmath521 if @xmath522 in ( [ defvlin ] ) , and look for the lowest noise producing a new contact point over the interval @xmath523 $ ] .the procedure is repeated until the whole interval is exhausted . this way an increasing sequence of noise values is obtained ,each corresponding to the slope of the optimal potential between two successive contact points .when the membrane leaking conductance is non zero , some change have to be brought to the above calculation of the optimal noise and potential . first , in the absence of inputs , the noise is no longer constant , but rather it is an exponentially increasing ( in absolute value ) function of the time ( [ eq1 ] ) .similarly , the potential @xmath150 itself is not a linear function of the time as in ( [ lp ] ) , but is a linear combination of @xmath524 with appropriate coefficients , see ( [ defv ] ) .the main conclusion of appendix [ seccp ] still holds : the difference between the noise values just after and before an active contact point , coinciding with a synaptic input , is always positive ( fig .[ fig - scheme]a ) .consequently , the procedure of section [ sec - algo ] , _ i.e. _ the iterative search for the active contact points and the minimal noise coefficient @xmath151 , defined through ( [ defnoise ] ) , remains unchanged .note that some care must be taken to translate the statement about the growth of the noise to the values of the noise coefficients .consider for instance two successive contact times , @xmath6 and @xmath525 , and call @xmath74 , @xmath526 the corresponding noise coefficients . that the noise is larger at time @xmath525 than at time@xmath6 implies that @xmath527 , but does not imply that @xmath526 is larger than @xmath74 .case , where the noise and the noise coefficient coincide . ] there exists , however , a major difference between the @xmath212 and @xmath528 cases .when @xmath529 , the optimal potential is not guaranteed to be a monotonous function of the time , as shown in fig .[ passc ] . for given values of @xmath530 , and of the times and the amplitudes of the synaptic inputs, the optimal potential @xmath150 may touch the threshold at an intermediate time , @xmath102 .such a situation is called passive contact .it is important to note that the value of the optimal noise during a passive contact is , according to eqn .( [ eq2 ] ) , equal to @xmath531 .as the optimal noise is a monotonous function of the time between two active contacts , see eqn ( [ eq1 ] ) , the value @xmath531 can be crossed at most once : there is at most one passive contact in between two successive active ones . to be more precise ,there are at most @xmath532 passive contacts in an inter - spike interval with @xmath533 active contacts .to decide the existence of a passive contact in an interval @xmath500 $ ] , we look for a solution of the two coupled equations expressing that the optimal potential touches the threshold without crossing it , @xmath534 the solutions of these equations give the noise coefficient @xmath111 and the contact time @xmath102 at which the optimal potential reaches the threshold value ( fig .[ passc ] ) .the solution can be calculated analytically , with the following result .let us call @xmath99 the value of the potential of the neuron at time @xmath535 . for each @xmath536we define @xmath537 where the summation runs overs the spikes entering the neuron between times @xmath519 and @xmath102 .a passive contact takes place in the interval @xmath538 $ ] if : @xmath539 : : @xmath531 and @xmath540 have the same sign ; @xmath539 : : the noise coefficient @xmath541 is smaller than the lowest noise coefficient corresponding to all the possible active contacts at times @xmath542 , with @xmath543 ; @xmath539 : : the corresponding contact time @xmath544 \ , \ ] ] where @xmath545 is given by ( [ etap ] ) , lies in the correct interval : @xmath546 ; @xmath539 : : the optimal potential can reach again the threshold at a later time , coinciding with an input spike or with the end of the inter - spike interval .we call @xmath547 the duration of the active contact such that the potential reaches @xmath26 at time @xmath548 , starting from @xmath26 at time @xmath549 , see fig .[ passc ] . the analytical expression for the duration of the passive contact allowing the potential to be in active contact at time@xmath548 is @xmath550 \right\}\ .\ ] ] where @xmath551 we must have @xmath552 for at least one value of @xmath553 .when all the conditions are fulfilled , a passive contact is present .the duration of the contact , @xmath133 , merely plays the role of a latency time after which the potential @xmath150 resumes its course ( fig [ passc ] ) .we can check that @xmath150 is an increasing function of @xmath554 .the optimal value of @xmath554 will therefore be equal to the smallest possible value of @xmath555 , for the very same reason that we had to chose the minimal noise when looking for active contacts , see example in fig .[ passc ] . to end this appendix, we give the expression for the log - likelihood @xmath184 ( [ deflpcc ] ) for an interval including a passive contact between two active contacts .gathering the contributions to the integral of the squared optimal noise coming from the three intervals @xmath504 $ ] , @xmath556 $ ] , and @xmath557 $ ] , we obtain @xmath558 - \exp \big[-2(t_c - t_0)/\tau\big ] } { 2}\right\}\ .\ ] ] differentiation of @xmath184 with respect to the current and couplings gives the expressions for the gradient and hessian matrix needed for the newton - raphson method .the expressions are easy to obtain but are lengthy , and thus we do not reproduce them .in this appendix , we analyze the eigenmodes and eigenvalues of the hessian matrix of the log - likelihood @xmath184 , and relate the eigenmodes to the fluctuations of the effective current , @xmath181 , of the current , @xmath18 , and of the couplings , @xmath13 .consider two successive spikes emitted by neuron @xmath5 and the optimal potential @xmath84 on the time interval @xmath559 $ ] .when the couplings @xmath13 vanish and passive contacts are absent , @xmath64 does not enter into contact with the threshold at times @xmath560 . by continuity , this statement remains true if the couplings @xmath13 have very small amplitudes . in this regime ,the stochastic process undergone by the potential is simply the ornstein - uhlenbeck process with a time - varying force , and the expression for @xmath184 ( [ deflpcc ] ) is exactly given by @xmath561 where @xmath562 and @xmath563 the hessian matrix of @xmath564 , attached to neuron @xmath5 , is the @xmath565 matrix ( [ fluch ] ) with elements @xmath566 @xmath225 is a positive matrix according to ( [ fluch2 ] ) . to study its spectrumlet us first consider the case of very weak leakage ( very large @xmath17 ) . in this limit ,calling @xmath567 the duration of the @xmath16 isi of neuron @xmath5 , we have @xmath568 let us define the firing rate @xmath569 of neuron @xmath570 in the @xmath16 isi of neuron @xmath5 , and @xmath571 .we obtain @xmath572 where @xmath33 is the duration of the recording .the right hand side of the above equation can be interpreted as the covariance matrix of the rates @xmath569 , where each isi of neuron @xmath5 is weighted proportionally to its duration . for vanishing couplings ,these instantaneous rates are decoupled from neuron to neuron .hence , @xmath573 fluctuates around the average firing rate @xmath180 ( number of the spikes fired by neuron @xmath14 , divided by @xmath33 ) , with a variance we denote by @xmath574 .this statement holds for @xmath575 ; in addition we define @xmath576 . neglecting terms of the order of @xmath577, we end up with the following approximation for the hessian matrix , @xmath578 which becomes exact in the limit of infinitely long recordings .the matrix @xmath225 is the sum of a rank one matrix plus a diagonal matrix . for small values of @xmath22 ,the fluctuations of the firing rates , represented by the @xmath579 s , are expected to be small compared to the product of any two average firing rates .we immediately deduce that the largest eigenvector of the matrix , @xmath580 , has components @xmath581 for all @xmath582 .the associated eigenvalue , @xmath209 , is given by @xmath583 where @xmath0 is the total number of spikes .if the neurons have quantitatively similar firing rates @xmath584 , then @xmath585 .the probability density of vector @xmath586 is @xmath587 \ .\ ] ] a fluctuation @xmath588 around the most likely values for the current and the couplings , and along vector @xmath580 , will change the log - likelihood by @xmath589 hence the effective current @xmath181 is associated to the largest eigenmode , and is the parameter requiring the least number of data to be inferred .we now look for the smallest eigenvalue , @xmath590 .numerical investigations suggest that the associated eigenvector , @xmath591 , correspond to fluctuations of the current @xmath18 only .we thus assume that the components of @xmath591 are : @xmath592 , and @xmath593 with @xmath594 .the eigensystem we need to solve is @xmath595 according to ( [ eqt2 ] ) and ( [ eqt1 ] ) , we have , for all @xmath570 , @xmath596 inserting this expression for the components @xmath597 of the eigenvector into ( [ eqt1 ] ) , we obtain @xmath598 if all neurons have quantitatively similar firing rates , @xmath599 , and variances , @xmath600 , we obtain @xmath601 . according to ( [ epsj7 ] ) , the components @xmath602 of the eigenvector are very small , @xmath603 , for all @xmath575 . hence @xmath591 is localized on its current component only . a fluctuation @xmath588 , where the @xmath604 s are chosen to be orthogonal to all the other eigenmodes of @xmath225 , modifies the log - likelihood by @xmath605 we conclude that the current @xmath18 is the hardest parameter to infer , _ i.e. _ the one requiring the largest number of data .when the membrane leaking time becomes of the order of , or smaller than the average isi duration , the above calculation has to be modified . from a qualitative point of view, the average firing rate @xmath606 must now be defined as the mean number of spikes emitted by the neuron @xmath14 in a time - window of duration @xmath17 preceding a spike of neuron @xmath5 , divided by @xmath17 , see ( [ deffjitau ] ) .the eigenvector of @xmath225 with largest eigenvalue @xmath209 is still given by @xmath607 , with @xmath608 , and @xmath609 again , these fluctuations are associated to the effective current , with the newly defined average firing rates @xmath610 .as @xmath17 gets smaller and smaller , all the rates @xmath180 with @xmath575 become smaller and smaller compared to @xmath610 , and the effective current @xmath181 gets closer and closer to the true current @xmath18 .obviously , the inference of the synaptic coupling @xmath13 is possible if the firing rate @xmath180 defined on a time - window of duration @xmath17 preceding a spike of neuron @xmath5 is much larger than @xmath611 .in this appendix , we derive formula ( [ result - fluctu ] ) for the fluctuations of the potential around its optimal value at the mid - point of the isi . a useful formulation for @xmath51 in ( [ prodfpt ] ) can be given in terms of a path integral over the potential , @xmath612\right ) \ .\nonumber\end{aligned}\ ] ] the measure @xmath613 in the path - integral ( [ pathint ] ) is restricted to the potentials @xmath4 remaining smaller than the threshold @xmath26 at all times @xmath6 .the upper bound @xmath614 means that the integral is performed over all the values of the potential smaller than @xmath26 at time @xmath615 , while @xmath616 is constrained to be zero .we introduce the dimensionless variable @xmath617 to represent the time - dependent fluctuation of the potential ( fig .[ fig - fluctuphi]c ) . according to ( [ pathint ] ) and ( [ pathint2 ] ) , the log probability density of a path - fluctuation @xmath618 on the inter - spike interval @xmath58 $ ] is , after multiplication by @xmath61 , @xmath619 \nonumber \\ & = & { \cal l}[v^*_i(t ) ; k,{\cal t},{\cal j},{\cal i } ] + \frac{v_{th}^2}{2 } \int _ { t_{i , k}}^ { t_{i , k+1 } } dt'\ ; \int _ { t_{i , k}}^ { t_{i , k+1}}dt\ ; \psi_i(t ' ) \ ; \frac{\delta ^2{\cal l}}{\delta v_i^*(t ' ) \,\delta v_i^*(t ) } \ ; \psi_i(t ) + o(\psi _i^3 ) \nonumber \\ & = & l^*({\cal t}|{\cal j},{\cal i } ) - \frac{v_{th}^2}{2 } \int _ { t_{i , k}}^ { t_{i , k+1 } } dt\ ; \psi_i(t ) \bigg [ -c^2 \frac{d^2}{dt^2 } + g^2 \bigg ] \psi_i(t ) + o(\psi _i^3 ) \ , \end{aligned}\ ] ] up to an additive term independent of @xmath620 .note that we have used the optimality condition ( [ optima ] ) to exclude terms linear in @xmath620 in ( [ pathint3 ] ) .we now want to perform the path integral over the fluctuations @xmath618 in ( [ pathint ] ) .when @xmath22 is small we may discard the cubic and higher order terms in @xmath620 . the boundary condition on @xmath620are @xmath621 : the values of the potential @xmath4 are constrained right after and before the emission of a spike , and , hence , can not fluctuate ( fig .[ fig - fluctuphi]c ) .we therefore write the fluctuations @xmath618 as the following fourier series , @xmath622 where the @xmath623 are stochastic coefficients .the integral on the last line of ( [ pathint3 ] ) can be calculated with the result @xmath624 \psi_i(t ) = \frac{\rho\,(cv_{th})^2}{4\tau}\sum _ { n\ge 1}\big[1+\left ( \frac{n\pi } \rho\right)^2\big]\ , \psi_n^2 \ , \ ] ] where @xmath625 is the duration of the isi measured in units of the membrane leaking time .hence , if we relax the constraint that the fluctuating potential should remain below threshold at all times , the @xmath623 s are independent gaussian variables with zero means and variances @xmath626 where @xmath158 is defined in ( [ defnounit ] ). we may now calculate the variance of @xmath620 at the mid - point of the isi , see fig .[ fig - fluctuphi]c , @xmath627 ^ 2\right > = \sum _ { p\ge 0 } \lambda_{2p+1 } \ .\ ] ] summing up the series over @xmath270 in ( [ pathint5 ] ) gives expression ( [ result - fluctu ] ) .in section [ sec - beyond ] we explain that the value of the moving threshold , @xmath165 , is estimated from the intersection of the tangent to the probability of survival in @xmath172 with the @xmath176 line .hence , @xmath628 the slope of @xmath171 can be expressed in terms of a series of parabolic cylinder functions ( alili , patie and perdersen , 2005 ) , @xmath629 where @xmath630 denotes the derivative of weber s function of order @xmath631 , @xmath632 , with respect to its argument @xmath633 .the normalization coefficients are @xmath634 the orders @xmath635 , @xmath636 , are the roots of the equation ( mei and lee , 1983 ) @xmath637 with @xmath638 .the gap between successive levels , @xmath639 , is larger than 1 .note that the contributions from high orders @xmath635 decay exponentially with @xmath640 in ( [ expressfpt ] ) .hence , in practice , the summation can be carried out over a finite number of terms .the moving threshold procedure was designed to take into account the effects of a moderate noise level , @xmath22 .an alternative approximate procedure consists in subtracting to the log - likelihood a cost - function preventing the current , or the effective current from getting too close to @xmath278 . for a quantitative treatment consider a single neuron in the absence of synaptic input , for which @xmath51 can be calculated under the form of a series of parabolic cylinder functions , see above .we denote by @xmath641 the approximation to @xmath51 obtained when taking into account the optimal path only .we define the cost - energy function @xmath642 \ .\ ] ] for the current @xmath166 .we show in fig .[ fig-4 ] the shape of @xmath643 for different values of @xmath9 , @xmath22 , and the inter - spike interval @xmath174 . as expected from above, this cost function is essentially flat when @xmath644 , and is repulsive when @xmath645 .the repulsion is strong when the inter - spike interval , @xmath174 , the membrane conductance , @xmath9 , and the noise standard deviation , @xmath22 , are large . in presence of synaptic inputs , we approximate the non - perturbative corrections by subtracting @xmath646 to our log - likelihood , where @xmath647 is the number of spikes of neuron @xmath5 , and @xmath181 its effective current .this simple approximation preserves the concavity of the log - likelihood and is computationally simple since @xmath643 has to be calculated only once for each step and neuron .simulations show that the performance of the inference algorithm with the cost function @xmath643 is quantitatively similar to the one obtained with the moving threshold procedure .jolivet r , lewis tj , gertsner w ( 2004 ) generalized integrate - and - fire models of neuronal activity approximate spike trains of a detailed model to a high degree of accuracy ._ j. neurophysiol . _ * 92 * : 959 .koyama s , paninski l ( 2009 ) efficient computation of the maximum a posteriori path and parameter estimation in integrate - and - fire and more general state - space models .makarov va , panetsos f , de feo o ( 2005 ) a method for determining neural connectivity and inferring the underlying network dynamics using extracellular spike recordings . _j. neurosci .methods _ * 144 * : 265 . truccolo w , eden u , fellows m , donoghue j , brown e ( 2005 ) a point process framework for relating neural spiking activity to spiking history , neural ensemble , and extrinsic covariate effects. _ j. neurophysiol . _ * 93 * : 1074 .

<S> we present two bayesian procedures to infer the interactions and external currents in an assembly of stochastic integrate - and - fire neurons from the recording of their spiking activity . </S><S> the first procedure is based on the exact calculation of the most likely time courses of the neuron membrane potentials conditioned by the recorded spikes , and is exact for a vanishing noise variance and for an instantaneous synaptic integration . </S><S> the second procedure takes into account the presence of fluctuations around the most likely time courses of the potentials , and can deal with moderate noise levels . </S><S> the running time of both procedures is proportional to the number @xmath0 of spikes multiplied by the squared number @xmath1 of neurons . </S><S> the algorithms are validated on synthetic data generated by networks with known couplings and currents . </S><S> we also reanalyze previously published recordings of the activity of the salamander retina ( including from 32 to 40 neurons , and from @xmath2 to @xmath3 spikes ) . </S><S> we study the dependence of the inferred interactions on the membrane leaking time ; the differences and similarities with the classical cross - correlation analysis are discussed . </S>

1002.3251

let @xmath0 be a set of real @xmath1 matrices . as usual , for @xmath2 ,let us denote by @xmath3 the set of all @xmath4-products of matrices from @xmath5 ; @xmath6 .for each @xmath2 , define the quantity @xmath7 where maximum is taken over all possible products of @xmath4 matrices from the set @xmath5 , and @xmath8 denotes the spectral radius of a matrix , that is the maximal magnitude of its eigenvalues .clearly , if @xmath9 then some matrices in the product @xmath10 will occur several times .the limit @xmath11 is called _ the generalized spectral radius _ of the matrix set @xmath5 @xcite . similarly , given a norm @xmath12 in @xmath13 , for each @xmath2 , define the quantity @xmath14 where @xmath15 , for a matrix @xmath16 , is the matrix norm generated by the vector norm @xmath12 in @xmath13 , that is @xmath17 .then the limit @xmath18 does not depend on the choice of the norm @xmath12 and is called _ the joint spectral radius _ of the matrix set @xmath5 @xcite . when @xmath19 this definition coincides with the famous gelfand formula for the spectral radius of a matrix @xcite as in this case @xmath20 is a singleton matrix set and @xmath21 . for matrix sets @xmath5 consisting of a finite amount of matrices , as is our case , the quantities @xmath22 and @xmath23 coincide with each other @xcite and their common value is denoted as @xmath24 while the quantities @xmath25 and @xmath26 form lower and upper bounds , respectively , for the joint / generalized spectral radius : @xmath27 this last formula may serve as a basis for a posteriori estimating the accuracy of computation of @xmath28 .the first algorithms of a kind in the context of control theory problems have been suggested in @xcite , for linear inclusions in @xcite , and for problems of wavelet theory in @xcite .later the computational efficiency of these algorithms was essentially improved in @xcite .unfortunately , the common feature of all such algorithms is that they do not provide any bounds for the number of computational steps required to get desired accuracy of approximation of @xmath28 .recently , in @xcite explicit computable estimates for the rate of convergence of the quantities @xmath29 to @xmath30 end their extension to the case of the joint spectral radius were obtained .probably , these results will help to make more constructive the problem of evaluating of @xmath28 by the generalized gelfand formula .some works suggest different formulas to compute @xmath28 .so , in @xcite it is shown that @xmath31 where , as usual , @xmath32 denotes the trace of a matrix .given a norm @xmath12 in @xmath13 , denote @xmath33 then the spectral radius of the matrix set @xmath5 can be defined by the equality @xmath34 where infimum is taken over all norms in @xmath35 @xcite , and therefore @xmath36 for any norm @xmath12 in @xmath13 .for irreducible matrix sets , is called _ irreducible _ if the matrices from @xmath5 have no common invariant subspaces except @xmath37 and @xmath13 . in @xcitesuch a matrix set was called quasi - controllable . ]the infimum in ( [ e - inf ] ) is attained , and for such matrix sets there are norms @xmath12 in @xmath13 , called _ extremal norms _ , for which @xmath38 in analysis of the joint spectral radius ideas suggested by n.e .barabanov @xcite play an important role .these ideas have got further development in a variety of publications among which we would like to distinguish @xcite .let the matrix set @xmath0 be irreducible .then the quantity @xmath39 is the joint ( generalized ) spectral radius of the set @xmath5 iff there is a norm @xmath12 in @xmath13 such that @xmath40 throughout the paper , a norm satisfying ( [ eq - mane - bar ] ) will be called a _barabanov norm _ corresponding to the matrix set @xmath5 .note that barabanov norms are not unique .similarly , as is shown in ( * ? ? ?* thm 3.3 ) and @xcite , the value of @xmath39 equals to @xmath28 if and only if for some central - symmetric convex body @xmath41 the following equality holds @xmath42 where @xmath43 denotes the convex hull of a set and @xmath44 . as is noted in @xcite , the relation ( [ e - protset ] )was proved by a.n .dranishnikov and s.v .konyagin , so it is natural to call the central - symmetric set @xmath41 the _ dranishnikov - konyagin - protasov set_. the set @xmath41 can be treated as the unit ball of some norm @xmath12 in @xmath35 ( recently this norm is usually called the _ protasov norm _ ) .note that barabanov and protasov norms are the extremal norms , that is they satisfy the inequality ( [ e - extnorm ] ) . in @xciteit is shown that barabanov and protasov norms are dual to each other .remark that formulas ( [ e - extnorm ] ) , ( [ eq - mane - bar ] ) and ( [ e - protset ] ) define the joint or generalized spectral radius for a matrix set in an apparently computationally nonconstructive manner . in spite of that ,namely such formulas underlie quite a number of theoretical constructions ( see , e.g. , @xcite ) and algorithms @xcite for computation of @xmath28 .different approaches for constructing barabanov norms to analyze properties of the joint ( generalized ) spectral radius are discussed , e.g. , in @xcite and ( * ? ? ?6.6 ) . in @xcitethe method of barabanov norms was the key instrument in disproving the lagarias - wang finiteness conjecture .the related constructions were essentially based on the study of the geometrical properties of the unit balls of some specific barabanov norms . in @xcitethe method of extremal polytope norms was the key tool in investigation of the finiteness properties of pairs of @xmath45 matrices . in this context the situation when one fails to find among current publications any detailed analysis of the geometrical properties of the unit balls of barabanov norms looks a bit paradoxical .partially this is explained by the fact that barabanov norms are defined nonconstructively , by an implicit procedure .so , even in simplest cases it is very difficult to visualize the shape of their unit balls .the present work may be treated as the first step to make up this deficiency . in the paper , an iteration procedureis considered that allows to build numerically barabanov norms for the irreducible matrix sets and simultaneously to compute the joint spectral radius of these sets .a similar iteration procedure is also discussed in @xcite .the structure of the paper is as follows . in introductionwe give basic definitions and present the motivation of the work . in section [ s - iterscheme ] the main iteration procedure is introduced and main theorem stating convergence of this procedure is formulated .the iteration procedure under consideration is called the max - relaxation procedure since in it the next approximation to the barabanov norm is constructed as the maximum of the current approximation and some auxiliary norm . in section [ mtproof ]proof main theorem is given . in section [ s - comp2d ] , to build simplest examples , the max - relaxation scheme is adapted for computations with @xmath46 matrices .results of numerical tests are illustrated by two examples . in section [ s - app ]we present the matlab code illustrating computations in example [ ex2 ] . at last , in concluding section [ s - rem ] we discuss some shortcomings of the proposed approach and formulate further unresolved problems .given @xmath47 , let @xmath0 be an irreducible set of @xmath1 real matrices . throughout the paper , a continuous function @xmath48 defined for @xmath49 and satisfying @xmath50will be called _ an averaging function_. examples for averaging functions are : @xmath51 given some averaging function @xmath52 , a norm @xmath53 in @xmath13 , and a vector @xmath54 such that @xmath55 , construct recursively the norms @xmath56 and @xmath57 , @xmath58 , in accordance with the following rules : * mr1 : * _ if the norm @xmath56 has been already defined compute the quantities _ @xmath59 * mr2 : * _ define the norms @xmath60 and _ @xmath61 : @xmath62 remark that the number of operations needed to perform one step of algorithm * mr1-mr2 * is of order @xmath63 , where @xmath64 is the number of operations needed to compute , for an arbitrary vector @xmath65 , the value of the norm @xmath66 with a relative accuracy @xmath67 . in general ,the value @xmath64 is of order @xmath68 .so , the total number of operations needed to perform @xmath4 steps of algorithm * mr1-mr2 * has the same rate of growth as @xmath69 .remark also , that the procedure ( [ e - lohibounds ] ) of calculation of @xmath70 resembles the technique of iterative approximation of the joint spectral radius suggested in @xcite . for any irreducible matrix set @xmath5 , nonzero vector @xmath54 , initial norm @xmath53 , and any averaging function @xmath48 , the sequences @xmath71 constructed by the iteration procedure * mr1 * , * mr2 * converge to @xmath72 , and the sequence of norms @xmath56 converges uniformly on each bounded set to some barabanov norm @xmath73 of the matrix set @xmath5 .moreover , the sequence @xmath74 is nondecreasing , the sequence @xmath75 is nonincreasing , and @xmath76 for all @xmath77 , which provides an a posteriori estimate for the computational error of @xmath72 .let us suppose that we managed to prove the following assertions : a1 : : _ the sequences @xmath75 and @xmath74 are convergent ; _ a2 : : _ the limits of the sequences @xmath75 and @xmath74 coincide : _@xmath78 a3 : : _ the norms @xmath57 converge pointwise to a limit @xmath73 ._ then the function @xmath73 will be a semi - norm in @xmath13 .moreover , by ( [ e - newnorm ] ) each norm @xmath57 meets the normalization condition @xmath79 , and hence @xmath80 which implies @xmath81 .note also that due to ( [ e - newnorm ] ) the norms @xmath57 differ from @xmath56 only by numerical factors .therefore , the quantities @xmath70 can be defined as @xmath82 then , passing to the limit in ( [ e - lohibounds - x ] ) , one can conclude that the semi - norm @xmath83 satisfies the _ barabanov condition _ @xmath84 but as shown in ( * ? ? ?3 ) , any semi - norm @xmath81 satisfying the barabanov condition for an irreducible matrix set is a barabanov norm .thus , under assumptions * a1 * , * a2 * and * a3 * , the iteration procedure ( [ e - lohibounds])([e - newnorm ] ) allows to build a barabanov norm and to find the joint spectral radius of the matrix set @xmath5 .so , to complete the proof of main theorem we need to justify assertions * a1 * , * a2 * and * a3 * which will be done in sections [ s - convnorm][s - fina2 ] . in section [ s - convnorm ]we establish convergence of the sequence of norms @xmath85 to some limit which allows to prove in lemma [ l - a3true ] that assertion * a3 * is a corollary of assertions * a1 * and * a2*. section [ s - rhorel ] demonstrates that the quantities @xmath74 form the family of lower bounds for the joint spectral radius @xmath39 of the matrix set @xmath5 , while the quantities @xmath75 form the family of upper bounds for @xmath39 . in section [ s - convrho ] we prove that the sequences @xmath71 are bounded and monotone which implies the existence of the limits @xmath86 and @xmath87 . at last , in sections [ s - newnorms ] and [ s - omega ] we prove that @xmath88 which allows to justify in section [ s - fina2 ] the validity of assertions * a1 * and * a2 * and thus to finalize the proof of main theorem . given a pair of norms @xmath89 and @xmath90 in @xmath13 define the quantities @xmath91 since all norms in @xmath13 are equivalent to each other , the quantities @xmath92 and @xmath93 are correctly defined and @xmath94 therefore the quantity @xmath95 which is called _ the eccentricity _ of the norm @xmath89 with respect to the norm @xmath90 ( see , e.g. , @xcite ) , is also correctly defined .[ l - eccbound ] let @xmath73 be a barabanov norm for the matrix set @xmath5. then @xmath96 and the sequence of the numbers @xmath97 is nonincreasing .note first that by ( [ e - newnorm ] ) each norm @xmath57 differs from the corresponding norm @xmath56 only by a numerical factor . from this , by the definition ( [ e - eccentr ] ) , ( [ e - defeccentr ] ) of the eccentricity of one norm with respect to another , the equality ( [ e - eqecc ] ) follows .denote by @xmath39 the joint spectral radius of the matrix set @xmath5 .then , by definitions of the function @xmath98 and of the barabanov norm @xmath73 , from ( [ e - lohibounds ] ) , ( [ e - auxnorm ] ) we obtain : @xmath99 therefore @xmath100 similarly , by definitions of the function @xmath101 and of the barabanov norm @xmath73 , from ( [ e - lohibounds ] ) , ( [ e - auxnorm ] ) we obtain : @xmath102 therefore @xmath103 by dividing termwise the inequality ( [ e - eccupbound ] ) on ( [ e - ecclobound ] ) we get @xmath104 hence , the sequence @xmath105 is nonincreasing .denote by @xmath106 the topological space of norms in @xmath107 with the topology of uniform convergence on bounded subsets of @xmath107 .[ c1-l - eccbound ] the sequence of norms @xmath85 is compact in @xmath106 . for each @xmath4 and any @xmath108 , by the definition ( [ e - eccentr ] ) of the functions @xmath98 and @xmath101 the following relations hold @xmath109 and then @xmath110 from which @xmath111 since here by construction the norms @xmath85 satisfy the normalization condition @xmath112 , and by lemma [ l - eccbound ] @xmath113 , we have @xmath114 therefore the norms @xmath56 , @xmath115 , are equicontinuous and uniformly bounded on each bounded subset of @xmath107 .moreover , their values are also uniformly separated from zero on each bounded subset of @xmath107 separated from zero . from here by the arzela - ascoli theorem the statement of the corollary follows .[ c2-l - eccbound ] if at least one of subsequences of norms from @xmath85 converges in @xmath106 to some barabanov norm then the whole sequence @xmath85 also converges in @xmath106 to the same barabanov norm .let @xmath116 be a subsequence of @xmath85 which converges in @xmath106 to some barabanov norm @xmath73 .then by definition of the eccentricity of one norm with respect to another @xmath117 here by lemma [ l - eccbound ] the eccentricities @xmath118 are nonincreasing in @xmath4 , and then the following stronger relation holds @xmath119 note now that by the definition ( [ e - eccentr ] ) , ( [ e - defeccentr ] ) of the eccentricity of one norm with respect to another @xmath120 from which by ( [ e - eccconv ] ) it follows that the sequence of norms @xmath85 converges in space @xmath106 to the norm @xmath73 . [ l - a3true ] assertion * a3 * is a corollary of assertions * a1 * and * a2*. by corollary [ c1-l - eccbound ] the sequence of norms @xmath85 has a subsequence @xmath116 that converges in space @xmath106 to some norm @xmath73 . then , passing to the limit in ( [ e - lohibounds - x ] ) as @xmath121 , we get by assertions * a1 * and * a2 * : @xmath122 which means that @xmath73 is a barabanov norm for the matrix set @xmath5 .this and corollary [ c2-l - eccbound ] then imply that the sequence @xmath85 converges in space @xmath106 to the barabanov norm @xmath73 .assertion * a3 * is proved .so , in view of lemma [ l - a3true ] to prove that the iteration procedure ( [ e - lohibounds])([e - newnorm ] ) is convergent it suffices to verify only that assertions * a1 * and * a2 * hold .the following lemma provides a way to estimate the spectral radius of a matrix set .[ l - rhorel ] let @xmath123 be numbers such that in some norm @xmath12 the inequalities @xmath124 hold .then @xmath125 , where @xmath39 is the joint spectral radius of the matrix set @xmath5 .let @xmath73 be some barabanov norm for the matrix set @xmath5 .since all norms in @xmath13 are equivalent , there are constants @xmath126 and @xmath127 such that @xmath128 consider , for each @xmath129 , the functions @xmath130 then , as is easy to see , @xmath131 similarly consider , for each @xmath129 , the functions @xmath132 for these functions , by definition of barabanov norms the following identity hold @xmath133 which is stronger than ( [ e - dbounds ] ) .now , note that ( [ e - sigmaest ] ) and the definition of the functions @xmath134 and @xmath135 imply @xmath136 then , by ( [ e - dbounds ] ) , ( [ e - dstar ] ) , @xmath137 from which the required estimates @xmath125 follow .so , lemma [ l - rhorel ] and the definition ( [ e - lohibounds ] ) of @xmath70 imply that the quantities @xmath74 form the family of lower bounds for the joint spectral radius @xmath39 of the matrix set @xmath5 , while the quantities @xmath75 form the family of upper bounds for @xmath39 .this allows to estimate a posteriori errors of computation of the joint spectral radius with the help of the iteration procedure ( [ e - lohibounds])([e - newnorm ] ) .estimate the value of @xmath138 .by definition , @xmath139 here by the definition ( [ e - lohibounds ] ) of the quantities @xmath70 the right - hand part of the chain of equalities can be estimated as follows : @xmath140 therefore , by definition of the norm @xmath141 , @xmath142 from which @xmath143 and then , @xmath144 so , the following lemma holds .[ l - monrho ] the sequence @xmath74 is bounded from above by each member of the sequence @xmath75 and is nondecreasing .the sequence @xmath75 is bounded from below by each member of the sequence @xmath74 and is nonincreasing . in view of lemma [ l - monrho ]there are the limits @xmath145 where @xmath146 which means that assertion * a1 * holds .hence , to prove that the iteration procedure ( [ e - lohibounds])([e - newnorm ] ) is convergent it remains only to justify assertion * a2 * : @xmath147 . to prove that @xmath148 , below it will be supposed the contrary , which will lead us to a contradiction . to simplify further constructionswe will switch over to a new sequence of norms for which the quantities @xmath70 will be independent of @xmath4 . by corollary [ c1-l - eccbound ] the sequence of the norms @xmath57 is compact in space @xmath106 .hence , there is a subsequence of indices @xmath149 such that the the norms @xmath150 converge to some norm @xmath151 satisfying the normalization condition @xmath152 .then , passing to the limit in ( [ e - lohibounds - x ] ) , by lemma [ l - monrho ] we obtain : @xmath153 now by induction the following statement can be easily proved .[ l - seqsharp ] for each @xmath154 , the sequence of the norms @xmath155 converges to some norm @xmath156 .moreover , for each @xmath154 , we have the equalities @xmath157 and the recurrent relations @xmath158 define , for each @xmath154 , the set @xmath160 by ( [ e - lohisharp ] ) @xmath159 is the set on which the quantity @xmath161 attains its minimum .[ l - eqnonomega ] if @xmath162 then @xmath163 .the statement of the lemma is obvious for @xmath164 .so , suppose that @xmath162 , @xmath165 . in this case ( [ e - defomega ] ) andthe inequalities @xmath166 imply @xmath167 or , what is the same , @xmath168 from here by the definition ( [ e - recsharp ] ) of the norm @xmath169 we get the required equality : @xmath170 the lemma is proved .[ l - omegadec ] if @xmath171 then @xmath172 for each @xmath154 .let @xmath173 .if @xmath164 then clearly @xmath162 , so suppose that @xmath165 . by definitions of the set @xmath174 and of the norm @xmath156 the following equalities take place : @xmath175 let here @xmath176 then from ( [ e - om ] ) it follows that @xmath177 but by the conditions of the lemma @xmath171 .then @xmath178 , and the right - hand part of the above equalities is strictly less than @xmath179 . a contradiction , since the left - hand part of the same equalities is no less than @xmath179 .the above contradiction is caused by the assumption ( [ e - wrongsuppose ] ) , and therefore it is proved that the condition @xmath180 implies the strict inequality @xmath181 in this case from ( [ e - om ] ) it follows that @xmath182 let us show that the equality ( [ e - maineq ] ) implies @xmath183 indeed , supposing the contrary , by definition of the quantity @xmath184 , there should be valid the strict inequality @xmath185 .then the left - hand part of the equality ( [ e - maineq ] ) should be strictly greater than @xmath186 , that is greater than the right - hand part of the same equality , which is impossible .this last contradiction shows that the equality ( [ e - finomega ] ) holds as soon as @xmath180 , which means by ( [ e - lohisharp ] ) that @xmath162 .[ c - mainomegaprop ] if @xmath171 then @xmath187 and @xmath188 by lemma [ l - omegadec ] @xmath189 is the family of embedded closed conic is called conic if together with each its point @xmath190 it contains the ray @xmath191 . ]then the intersection @xmath192 of these sets is also a closed conic set such that @xmath193 . by definition of the set @xmath192 , if @xmath194 then @xmath162 for every integer @xmath195 .hence , by lemma [ l - eqnonomega ] @xmath163 , from which the equalities ( [ e - eqsharpnorms ] ) follow . by corollary [ c - mainomegaprop ]there is a non - zero vector @xmath196 on which all the norms @xmath156 take the same values : @xmath197 then , due to uniform boundedness of the eccentricities of the norms @xmath156 with respect to some barabanov norm @xmath73 ( this fact can be proved by verbatim repetition of the analogous proof for the norms @xmath56 ) , the norms @xmath156 form a family which is uniformly bounded and equicontinuous with respect to the barabanov norm @xmath73 : @xmath198 hence , by the arzela - ascoli theorem the family of norms @xmath199 is compact in @xmath106 . from the definition ( [ e - recsharp ] ) of the norms @xmath156 it follows also that @xmath200 then the norms @xmath156 are monotone increasing in @xmath4 and bounded ( with respect to the barabanov norm @xmath73 ) and therefore they pointwise converge to some norm @xmath201 .moreover , since the family of norms @xmath199 is equicontinuous with respect to the barabanov norm @xmath73 , the norms @xmath156 converge to the norm @xmath201 in space @xmath106 .now , passing to the limit in the relations @xmath202 which follow from ( [ e - recsharp ] ) , we obtain @xmath203 from here @xmath204 on the other hand , passing to the limit in the first relation of ( [ e - lohisharp ] ) , we obtain @xmath205 relations ( [ e - finestim ] ) and ( [ e - finestim1 ] ) imply the inequality @xmath206 which contradicts the assumption @xmath207 because by definition of the function @xmath52 the condition @xmath208 implies the inequality @xmath209 .the obtained contradiction completes the proof of the equality @xmath148 as well as of the convergence of the iteration procedure ( [ e - lohibounds])([e - newnorm ] ) .let @xmath0 be a set of real @xmath45 matrices @xmath210a^{(i)}_{21}&a^{(i)}_{22 } \end{array}\right).\ ] ] let @xmath211 be the polar coordinates in @xmath212 .then , for a vector @xmath213 with cartesian coordinates @xmath214 , we have @xmath215 and @xmath216 define , for an arbitrary norm @xmath12 , the function @xmath217 then the norm @xmath66 of the vector @xmath190 with polar coordinates @xmath211 is determined by the equality @xmath218 and the unit sphere in the norm @xmath12 is determined as the geometrical locus of the vectors @xmath190 polar coordinates of which satisfy the relations @xmath219 see fig .[ f - polcoord ] . .] now , let @xmath220 be the function defining in the polar coordinates the graph of the unit sphere @xmath221 of the norm @xmath56 determined by the iteration procedure ( [ e - lohibounds])([e - newnorm ] ) .rewrite the relations ( [ e - lohibounds])([e - newnorm ] ) in terms of the functions @xmath220 . to do itwe should express the quantities @xmath222 , @xmath223 , in terms of the functions @xmath220 . by ( [ e - rr ] )@xmath224 here by definition of the matrix @xmath225 @xmath226 where @xmath227 similarly , by definition of the matrix @xmath225 @xmath228 where @xmath229 from the obtained relations it follows that the first two equalities in ( [ e - lohibounds ] ) take the form @xmath230 or , what is the same , @xmath231 where @xmath232 the relations ( [ e - auxnorm ] ) take the form @xmath233 or , equivalently , @xmath234 and the normalization condition ( [ e - newnorm ] ) takes the form @xmath235 where @xmath236 are polar coordinates of the vector @xmath237 .taking in place of @xmath237 the vector with polar coordinates @xmath238 the normalization condition can be rewritten in the form @xmath239 so , the max - relaxation iteration scheme can be represented as follows .given an averaging function @xmath52 , set @xmath240 , and build recursively the @xmath241periodic functions @xmath220 and @xmath242 , @xmath58 , in accordance with the following rules : _ _mr1 : : regarding the function @xmath220 already known compute the numerical values @xmath243 and @xmath244 by formulas ( [ e - rlohibounds ] ) , ( [ e - rstar ] ) and set @xmath245 ; mr2 : : define @xmath246 by ( [ e - rauxnorm ] ) and @xmath247 by ( [ e - rnewnorm ] ) , and then determine the norm @xmath61 as @xmath248 , where @xmath211 are the polar coordinates of the vector @xmath190 .matrices.,title="fig:",scaledwidth=45.0% ] matrices.,title="fig:",scaledwidth=45.0% ] [ ex1]consider the family @xmath249 of @xmath45 matrices @xmath250 the functions @xmath251 were chosen to be piecewise linear with @xmath252 nodes uniformly distributed over the interval @xmath253 $ ] .it was needed @xmath254 iterations of algorithm * mr1-mr2 * with the averaging function @xmath255 implemented in matlab to compute the joint spectral radius @xmath72 with the absolute accuracy @xmath256 .the computed value of the joint spectral radius is @xmath257 .the computed unit sphere of the barabanov norm @xmath73 after the @xmath254th iteration of algorithm * mr1-mr2 * is shown on fig .[ f - barnorms ] on the left . as is seen from fig .[ f - barnorms ] , in example [ ex1 ] the sets @xmath258 and @xmath259 have exactly @xmath260 intersection points .this was theoretically proved in @xcite for the case when one of the matrices @xmath261 is lower triangle and the other is upper triangle , and their entries are nonnegative . in @xcitethis fact was one of key points in disproving the finiteness conjecture .we do not know whether this fact is valid in a general case or not , but numerical tests based on algorithm * mr1-mr2 * with several dozens pairs of matrices @xmath261 testify for this fact .[ ex2]consider the family @xmath262 of @xmath46 matrices @xmath263 here the functions @xmath251 were also chosen to be piecewise linear with @xmath252 nodes uniformly distributed over the interval @xmath253 $ ] .it was needed @xmath264 iterations of algorithm * mr1-mr2 * with the averaging function @xmath255 implemented in matlab to compute the joint spectral radius @xmath72 with the absolute accuracy @xmath256 .the computed value of the joint spectral radius is @xmath265 .the computed unit sphere of the barabanov norm @xmath73 after the @xmath264d iteration of algorithm * mr1-mr2 * is shown on fig .[ f - barnorms ] on the right .here we present the matlab code used for computations in example [ ex2 ] ..... % % initialization % closing of all graphs , clearing of all variables and of command % window close all ; clear all ; clc ; commandwindow ; % specifying the number of points for the representation of the % boundary of the barabanov norm and making it even . npoints=3000 ; npoints=2*floor(npoints/2 ) ; % specifying the maximum number of iterations and the tolerance % for computation of the j.s.r . niter=1000 ; tolerance=0.001 ; % specifying the pair of matrices for which the barabanov norm % and the j.s.r .are computed a=[1,1;0,1 ] ; b=[0.8,0.6;-0.6,0.8 ] ; c=[1,0;-0.4,1.3 ] ; % discretized angle array ( phi ) and radii array ( r ) to represent % the boundary of the barabanov norm in polar coordinates as the % graph of the function r(phi ) .pi:2*pi / npoints : pi ; sinphi = sin(phi(2)-phi(1 ) ) ; sinphi2=sin(phi(3)-phi(1 ) ) ; sinhalf = sinphi / sinphi2 ; r = ones(1,npoints+1 ) ; % initialization of auxiliary variables rap = ones(1,npoints+1 ) ; na = ones(1,npoints+1 ) ; ra = ones(1,npoints+1 ) ; ira = ones(1,npoints+1 ) ; rbp = ones(1,npoints+1 ) ; nb = ones(1,npoints+1 ) ; rb = ones(1,npoints+1 ) ; irb = ones(1,npoints+1 ) ; rcp = ones(1,npoints+1 ) ; nc = ones(1,npoints+1 ) ; rc = ones(1,npoints+1 ) ; irc = ones(1,npoints+1 ) ; rabx = ones(1,npoints+1 ) ; rabc = ones(1,npoints+1 ) ; ir = ones(1,npoints+1 ) ; % % transforms in polar coordinates phia = atan2(a(2,1)*cos(phi)+a(2,2)*sin(phi),a(1,1)*cos(phi)+ ...a(1,2)*sin(phi ) ) ; ra = sqrt((a(1,1)*cos(phi)+a(1,2)*sin(phi)).^2+(a(2,1)*cos(phi)+ ... a(2,2)*sin(phi)).^2 ) ; phib = atan2(b(2,1)*cos(phi)+b(2,2)*sin(phi),b(1,1)*cos(phi)+ ...b(1,2)*sin(phi ) ) ; rb = sqrt((b(1,1)*cos(phi)+b(1,2)*sin(phi)).^2+(b(2,1)*cos(phi)+ ... b(2,2)*sin(phi)).^2 ) ; phic = atan2(c(2,1)*cos(phi)+c(2,2)*sin(phi),c(1,1)*cos(phi)+ ... c(1,2)*sin(phi ) ) ; rc = sqrt((c(1,1)*cos(phi)+c(1,2)*sin(phi)).^2+(c(2,1)*cos(phi)+ ...c(2,2)*sin(phi)).^2 ) ; % % angle transformation maps for m=1:1:npoints+1 fn = npoints*(pi+phia(m))/(2*pi)+1 ; na(m)=round(fn ) ; if ( na(m)<1 ) na(m)=1 ; end if ( na(m)>(npoints+1 ) ) na(m)=npoints+1 ; end end for m=1:1:npoints+1 fn = npoints*(pi+phib(m))/(2*pi)+1 ; nb(m)=round(fn ) ; if ( nb(m)<1 ) nb(m)=1 ; end if ( nb(m)>(npoints+1 ) ) nb(m)=npoints+1 ; end end for m=1:1:npoints+1 fn = npoints*(pi+phic(m))/(2*pi)+1 ; nc(m)=round(fn ) ; if ( nc(m)<1 ) nc(m)=1 ; end if ( nc(m)>(npoints+1 ) ) nc(m)=npoints+1 ; end end % % iterative evaluation of r % % computation of the next iteration for the norm i=0 ; while ( i < niter ) i = i+1 ; for m=1:1:npoints+1 rap(m)=r(na(m ) ) ; end ra = rap.*ra ; for m=1:1:npoints+1 rbp(m)=r(nb(m ) ) ; end rb = rbp.*rb ; for m=1:1:npoints+1 rcp(m)=r(nc(m ) ) ; end rc = rcp.*rc ; rabc = max(max(ra , rb),rc ) ; % % making rab locally convex in the case when % % computation errors caused its inconvexity rabx(1)=min(rabc(1),sinhalf*(rabc(2)+rabc(npoints ) ) ) ; rabx(npoints+1)=rabx(1 ) ; for m=2:1:npoints rabx(m)=min(rabc(m),sinhalf*(rabc(m-1)+rabc(m+1 ) ) ) ; end rabc = rabx ; srmax = max(rabc./r ) ; srmin = min(rabc./r ) ; sout = strcat('i=%4d , bounds for j.s.r . :% 5.3f < r < % 5.3f ' ) ; s = sprintf(sout , i , srmin , srmax ) ; disp(s ) ; sr=2/(srmax+srmin ) ; rx = max(sr*rabc , r ) ; nfact = rx(npoints/2 + 1 ) ; r = rx / nfact ; if ( i>(niter-10 ) ) hold off ; polar(phi , srmax./ra ) ; hold all ; polar(phi , srmax./rb ) ; hold all ; polar(phi , srmax./rc ) ; hold all ; polar(phi,1./r ) ; pause end if ( abs(srmax - srmin)<tolerance ) break ; end end % % drawing ir=1./r ; ira=1./ra ; irb=1./rb ; irc=1./rc ; axra = max(srmax.*ira ) ; axrb = max(srmax.*irb ) ; axrc = max(srmax.*irc ) ; axr = max(ir ) ; maxr = ceil(max(max(axra , axrb),max(axr , axrc ) ) ) ; hold off ; axis equal ; axis([-maxr maxr -maxr maxr ] ) ; hold all ;the max - relaxation algorithm suggested in the paper allows to calculate the joint spectral radius of a finite matrix family ( of arbitrary matrix size and arbitrary amount of matrices in the set ) with any required accuracy and to evaluate a posteriori the computational error .still , this algorithm gives rise to a set of open problems . while the quantities @xmath71 provide a posteriori bounds for the accuracy of approximation of @xmath28 the question about the accuracy of approximation of the barabanov norm @xmath73 by the norms @xmath57 is open .it seems , the difficulty in resolving this problem is caused by the fact that in general the barabanov norms for a matrix family are determined ambiguously .namely to overcome this difficulty we preferred to consider relaxation algorithm instead of direct one .moreover , as can be shown , both theoretically and numerically , if to set @xmath266 in ( [ e - auxnorm ] ) then the obtained direct computational analog of algorithm * mr1-mr2 * may turn out to be non - convergent .the question about the rate of convergence of the sequences @xmath75 and @xmath74 to the joint spectral radius is also open .remark also that in this paper mainly the algorithm for building of barabanov norms rather than its computational details was studied .the numerical aspects of implementation of this algorithm require additional analysis .an estimation of the computational cost of the max - relaxation algorithm is required .the dependance of the algorithm on parameters @xmath267 , @xmath268 and the choice of the averaging function @xmath48 is acute , too .this work was supported by the russian foundation for basic research , project no .10 - 01 - 00175 .n. barabanov , _ lyapunov exponent and joint spectral radius : some known and new results _ , in : proceedings of the 44th ieee conference on decision and control and european control conference 2005 , seville , spain , december 1215 , pp . 23322337 , 2005 .n. guglielmi and m. zennaro , _ polytope norms and related algorithms for the computation of the joint spectral radius _ , in : proceedings of the 44th ieee conference on decision and control and european control conference 2005 , seville , spain , december 1215 , pp .30073012 , 2005 .v. kozyakin , _ a dynamical systems construction of a counterexample to the finiteness conjecture _ , in : proceedings of the 44th ieee conference on decision and control and european control conference 2005 , seville , spain , december 1215 , pp . 23382343 , 2005 .v. kozyakin , _ on accuracy of approximation of the spectral radius by the gelfand formula _ , linear algebra appl . , * 431 * ( 2009 ) , 21342141 , http://dx.doi.org/10.1016/j.laa.2009.07.008 [ ] , http://dx.doi.org/10.1016/j.laa.2009.07.008 .v. s. kozyakin , _ structure of extremal trajectories of discrete linear systems and the finiteness conjecture _ ,remote control , * 68 * ( 2007 ) , 174209 , http://dx.doi.org/10.1134/s0005117906040171 [ ] .v. s. kozyakin , _ on the computational aspects of the theory of joint spectral radius _ , dokl .nauk , * 427 * ( 2009 ) , 160164 , http://dx.doi.org/10.1134/s1064562409040097 [ ] , in russian , translation in doklady mathematics 80 ( 2009 ) , no .1 , 487491 .v. s. kozyakin and a. v. pokrovski , _ the role of controllability - type properties in the study of the stability of desynchronized dynamical systems _ , dokl .nauk , * 324 * ( 1992 ) , 6064 , in russian , translation in soviet phys .37 ( 1992 ) , no . 5 , 213215 .v. s. kozyakin and a. v. pokrovskii , `` estimates of amplitudes of transient regimes in quasi - controllable discrete systems , '' cadsem report 96005 , deakin university , geelong , australia , 1996 , http://arxiv.org/abs/0908.4138 [ ] . v. s. kozyakin and a. v. pokrovskii , _ quasi - controllability and estimation of the amplitudes of transient regimes in discrete systems _ , izv . ,nauk , mat . mat . model .upr . , * 1 * ( 1997 ) , 128150 , in russian .p. a. parrilo and a. jadbabaie , _ approximation of the joint spectral radius using sum of squares _, linear algebra appl . , * 428 * ( 2008 ) , 23852402 , http://dx.doi.org/10.1016/j.laa.2007.12.027 [ ] , http://arxiv.org/abs/0712.2887 [ ] . e. plischke , f. wirth and n. barabanov , _ duality results for the joint spectral radius and transient behavior _ , in : proceedings of the 44th ieee conference on decision and control and european control conference 2005 , seville , spain , december 1215 , pp . 23442349 , 2005 .v. protasov , _ the geometric approach for computing the joint spectral radius _ , in : proceedings of the 44th ieee conference on decision and control and european control conference 2005 , seville , spain , december 1215 , pp . 30013006 , 2005 .j. theys , `` joint spectral radius : theory and approximations , '' ph.d .thesis , facult des sciences appliques , dpartement dingnierie mathmatique , center for systems engineering and applied mechanics , universit catholique de louvain , 2005 .f. wirth , _ on the structure of the set of extremal norms of a linear inclusion _ , in : proceedings of the 44th ieee conference on decision and control , and the european control conference 2005 seville , spain , december 1215 , pp .30193024 , 2005 .

<S> the problem of construction of barabanov norms for analysis of properties of the joint ( generalized ) spectral radius of matrix sets has been discussed in a number of publications . in @xcite </S><S> the method of barabanov norms was the key instrument in disproving the lagarias - wang finiteness conjecture . </S><S> the related constructions were essentially based on the study of the geometrical properties of the unit balls of some specific barabanov norms . in this context </S><S> the situation when one fails to find among current publications any detailed analysis of the geometrical properties of the unit balls of barabanov norms looks a bit paradoxical . </S><S> partially this is explained by the fact that barabanov norms are defined nonconstructively , by an implicit procedure . </S><S> so , even in simplest cases it is very difficult to visualize the shape of their unit balls . the present work may be treated as the first step to make up this deficiency . in the paper </S><S> an iteration procedure is considered that allows to build numerically barabanov norms for the irreducible matrix sets and simultaneously to compute the joint spectral radius of these sets . </S><S> victor kozyakin </S>

0908.2219

within the quantum theory the state vector at time @xmath1 , @xmath5 , for the physical system under consideration which initially ( at @xmath6 ) was in the state @xmath7 can be found by solving the schdinger equation @xmath8 where @xmath9 , @xmath10 is the hilbert space of states of the considered system , @xmath11 and @xmath12 denotes the total selfadjoint hamiltonian for the system . if one considers an unstable state @xmath13 of the system then using the solution @xmath14 of eq .( [ schrod ] ) for the initial condition @xmath15 one can determine the decay law , @xmath16 of this state decaying in vacuum @xmath17 where @xmath18 is probability amplitude of finding the system at the time @xmath1 in the initial state @xmath19 prepared at time @xmath20 , @xmath21 we have @xmath22 from basic principles of quantum theory it is known that the amplitude @xmath18 , and thus the decay law @xmath16 of the unstable state @xmath19 , are completely determined by the density of the energy distribution @xmath23 for the system in this state @xcite @xmath24 where @xmath25 . note that ( [ a - spec ] ) and ( [ a(0 ) ] ) mean that there must be @xmath26 from the last property and from the riemann lebesgue lemma it follows that the amplitude @xmath18 , being the fourier transform of @xmath27 ( see ( [ a - spec ] ) ) , must tend to zero as @xmath28 @xcite . in @xcite assuming that the spectrum of @xmath12 must be bounded from below , @xmath29 , and using the paley wiener theorem @xcite it was proved that in the case of unstable states there must be @xmath30 for @xmath31 . here@xmath32 and @xmath33 .this means that the decay law @xmath16 of unstable states decaying in the vacuum , ( [ p(t ) ] ) , can not be described by an exponential function of time @xmath1 if time @xmath1 is suitably long , @xmath28 , and that for these lengths of time @xmath34 tends to zero as @xmath35 more slowly than any exponential function of @xmath1 .the analysis of the models of the decay processes shows that @xmath36 , ( where @xmath37 is the decay rate of the state @xmath38 ) , to a very high accuracy for a wide time range @xmath1 : from @xmath1 suitably later than some @xmath39 but @xmath40 up to @xmath41 and smaller than @xmath42 , where @xmath43 denotes the time @xmath1 for which the nonexponential deviations of @xmath18 begin to dominate ( see eg ., @xcite , @xcite @xcite . from this analysisit follows that in the general case the decay law @xmath16 takes the inverse power like form @xmath44 , ( where @xmath45 ) , for suitably large @xmath46 .this effect is in agreement with the general result ( [ |a(t)|-as ] ) .effects of this type are sometimes called the `` khalfin effect '' @xcite .the problem of how to detect possible deviations from the exponential form of @xmath16 at the long time region has been attracting the attention of physicists since the first theoretical predictions of such an effect .@xcite many tests of the decay law performed some time ago did not indicate any deviations from the exponential form of @xmath16 at the long time region . nevertheless , conditions leading to the nonexponetial behavior of the amplitude @xmath18 at long times were studied theoretically @xcite @xcite .conclusions following from these studies were applied successfully in an experiment described in the rothe paper @xcite , where the experimental evidence of deviations from the the exponential decay law at long times was reported .this result gives rise to another problem which now becomes important : if ( and how ) deviations from the exponential decay law at long times affect the energy of the unstable state and its decay rate at this time region .note that in fact the amplitude @xmath18 contains information about the decay law @xmath16 of the state @xmath19 , that is about the decay rate @xmath37 of this state , as well as the energy @xmath47 of the system in this state .this information can be extracted from @xmath18 . indeed , if @xmath19 is an unstable ( a quasi stationary ) state then @xmath48 so , there is @xmath49 in the case of quasi stationary states .the standard interpretation and understanding of the quantum theory and the related construction of our measuring devices are such that detecting the energy @xmath47 and decay rate @xmath37 one is sure that the amplitude @xmath18 has the form ( [ a - q - stat ] ) and thus that the relation ( [ e - ig ] ) occurs .taking the above into account one can define the `` effective hamiltonian '' , @xmath50 , for the one dimensional subspace of states @xmath51 spanned by the normalized vector @xmath19 as follows ( see , eg .@xcite ) , @xmath52 in general , @xmath50 can depend on time @xmath1 , @xmath53 .one meets this effective hamiltonian when one starts with the schrdinger equation ( [ schrod ] ) for the total state space @xmath10 and looks for the rigorous evolution equation for the distinguished subspace of states @xmath54 . in the case of one dimensional @xmath51 this rigorous schrdinger like evolution equation has the following form for the initial condition @xmath55 ( see , @xcite and references one finds therein ) , @xmath56 relations ( [ h ] ) and ( [ eq - for - h ] ) establish a direct connection between the amplitude @xmath18 for the state @xmath57 and the exact effective hamiltonian @xmath58 governing the time evolution in the one dimensional subspace @xmath59 .thus , the use of the evolution equation ( [ eq - for - h ] ) or the relation ( [ h ] ) is one of the most effective tools for the accurate analysis of the early as well as the long time properties of the energy and decay rate of a given quasi stationary state @xmath60 .so let us assume that we know the amplitude @xmath18 .then starting with this @xmath18 and using the expression ( [ h ] ) one can calculate the effective hamiltonian @xmath58 in a general case for every @xmath1 .thus , one finds the following expressions for the energy and the decay rate of the system in the state @xmath19 under considerations , to be more precise for the instantaneous energy and the instantaneous decay rate , ( for details see : @xcite ) , @xmath61 where @xmath62 and @xmath63 denote the real and imaginary parts of @xmath64 respectively .the deviations of the decay law @xmath34 from the exponential form can be described equivalently using time - dependent decay rate ( [ g(t ) ] ) . in terms of such @xmath65 the khalfin observation that @xmath16 must tend to zero as @xmath35 more slowly than any exponential function means that @xmath66 for @xmath67 and @xmath68 . using ( [ h ] ) and ( [ e(t ) ] ) , ( [ g(t ) ] ) one can find that @xmath69 the aim of this talk is to discuss the long time behaviour of @xmath70 using @xmath18 calculated for the given density @xmath23 .we show that @xmath71 as @xmath72 for the model considered and that a wide class of models has similar long time properties : @xmath73 .it seems that , in contrast to the standard khalfin effect @xcite , in the case of the quasi stationary states belonging to the same class as excited atomic levels , these long time properties of the instantaneous energy @xmath74 have a chance to be detected , eg ., by analyzing the properties of the high energy cosmic rays or the spectra of very distant stars .let us assume that + @xmath75 , ( where , @xmath76 ) , and let us choose @xmath27 as follows @xmath77 where @xmath78 is a normalization constant and @xmath79 for such @xmath80 using ( [ a - spec ] ) one has @xmath81 where @xmath82 formula ( [ a - bw ] ) leads to the result @xmath83\ , \big\ } , \label{a - e(1)}\end{aligned}\ ] ] where @xmath84 denotes the integral exponential function @xcite .in general one has @xmath85 where @xmath86 making use of the asymptotic expansion of @xmath84 @xcite @xmath87 where @xmath88 , one finds @xmath89 where @xmath90 , and @xmath91 for the considered case ( [ omega - bw ] ) of @xmath80 ( for details see @xcite ) .from ( [ h - as ] ) it follows that @xmath92 where @xmath93 , and @xmath94 the property ( [ re - h - as ] ) means that @xmath95 for different states @xmath96 , ( @xmath97 ) one has @xmath98 whereas in general @xmath99 .note that from ( [ a(t)-as ] ) one obtains @xmath100\,\,\,\,\ , e^{\textstyle - \frac{1}{2}\,\frac{\gamma_{\phi}^{0}}{\hbar}\,\,t}\;\ ; \frac{\gamma_{\phi}^{0}}{|h_{\phi}^{0}\,-\,{\cal e}_{min}\,|^{\,2 } } \ ; \frac{\hbar}{t } \nonumber \\ & & + \frac{n^{2}}{4 \pi^{2}}\;\frac{(\gamma_{\phi}^{0})^{2}}{|h_{\phi}^{0}\,-\,{\cal e}_{min}\,|^{\,4 } } \ ; \frac{\hbar^{2}}{t^{2}}\ ; + \;\ldots\;\ ; .\label{t - as-1}\end{aligned}\ ] ] relations ( [ a(t)-as ] ) ( [ im - h1-h2-as ] ) become important for times @xmath101 , where @xmath43 denotes the time @xmath1 at which contributions to @xmath102 from the first exponential component in ( [ t - as-1 ] ) and from the third component proportional to @xmath103 are comparable , that is ( see ( [ a - exp+a - non ] ) ) , @xmath104 for @xmath105 .so @xmath43 can be be found by considering the following relation @xmath106 assuming that the right hand side is equal to the left hand side in the above relation one gets a transcendental equation .exact solutions of such an equation can be expressed by means of the lambert @xmath107 function @xcite .an asymptotic solution of the equation obtained from the relation ( [ t - as-2 ] ) is relatively easy to find @xcite .the very approximate asymptotic solution , @xmath43 , of this equation for @xmath108 ( in general for @xmath109 ) has the form @xmath110\,+\ , \ldots \;\;. \label{t - as-3}\end{aligned}\ ] ]to complete the analysis performed in the previous section let us consider a more general case of @xmath27 and @xmath18 .for a start , let us consider a relatively simple case when @xmath111 and @xmath112 .let derivatives @xmath113 , ( @xmath114 ) , be continuous in @xmath115 , ( that is let for @xmath116 all @xmath113 be continuous and all the limits @xmath117 exist ) and let all these @xmath113 be absolutely integrable functions then ( see @xcite ) , @xmath118 where @xmath119 .let us now consider a more complicated form of the density @xmath27 .namely let @xmath27 be of the form @xmath120 where @xmath121 and it is assumed that @xmath122 , ( @xmath123 ) , exist and they are continuous in @xmath124 , and limits @xmath125 exist , @xmath126 for all above mentioned @xmath127 and @xmath112 , then @xmath128 , \label{a - eta}\end{aligned}\ ] ] where ( compare @xcite ) @xmath129 @xmath130 is the gamma function and @xmath131 , @xmath132 and @xmath133 .the asymptotic form of @xmath134 for @xmath35 for the @xmath18 given by the relation ( [ a - omega ] ) looks as follows @xmath135 in the more general case of @xmath18 ( see , e.g. ( [ a - eta ] ) ) after some algebra the asymptotic approximation of @xmath18 can be written as follows @xmath136 where @xmath137 and @xmath138 are complex numbers . from the relation ( [ a(t)-as - w ] ) one concludes that @xmath139 now let us take into account the relation ( [ eq - for - h ] ) . from this relation and relations ( [ a(t)-as - w ] ) , ( [ da(t)-as - w ] ) it follows that @xmath140 where @xmath141 are complex numbers with negative or positive real and imaginary parts .this means that in the case of the asymptotic approximation to @xmath18 of the form ( [ a(t)-as - w ] ) the following property holds , @xmath142 it seems to be important that results ( [ h - sim - w ] ) and ( [ lim - h ] ) coincide with the results ( [ h - as ] ) ( [ im - h1-h2-as ] ) obtained for the density @xmath80 given by the formula ( [ omega - bw ] ) .this means that general conclusion obtained for the other @xmath27 defining unstable states should be similar to those following from ( [ h - as ] ) ( [ im - h1-h2-as ] ) .long time properties of the survival probability @xmath143 and instantaneous energy @xmath74 are relatively easy to find analytically for times @xmath67 even in the general case as it was shown in previous section and @xcite .it is much more difficult to analyze these properties analytically in the transition time region where @xmath144 .it can be done numerically for given models . the model considered in sec .2 and defined by the density @xmath80 , ( [ omega - bw ] ) , allows one to find numerically the decay curves and the instantaneous energy @xmath145 as a function of time @xmath1 .the results presented in this section have been obtained assuming for simplicity that the minimal energy @xmath3 appearing in the formula ( [ omega - bw ] ) is equal to zero , @xmath146 .so , all numerical calculations were performed for the density @xmath147 given by the following formula @xmath148 for some chosen @xmath149 . performing calculations particular attentionwas paid to the form of the probability @xmath143 , i. e. of the decay curve , and of the instantaneous energy @xmath145 for times @xmath1 belonging to the most interesting transition time - region between exponential and nonexponential parts of @xmath143 , where the following relation corresponding with ( [ a - exp = a - non ] ) and ( [ t - as-2 ] ) takes place , @xmath150 where @xmath151 are defined by ( [ a - exp+a - non ] ) .results are presented graphically below . a similar form of a decay curves one meets for a very large class of models defined by energy densities @xmath154 of the following type ( see @xcite ) , @xmath155 where @xmath156, @xmath157 is a form factor it is a smooth function going to zero as @xmath158 and it has no threshold and no pole . the asymptotical large time behavior of @xmath18 is due to the term @xmath159 and the choice of @xmath160 ( see sec . 3 ) .the density @xmath154 defined by the relation ( [ omega - gen ] ) fulfills all physical requirements and it leads to the decay curves having a very similar form at transition times region to the decay curves presented above .the characteristic feature of all these decay curves is the presence of sharp and frequent oscillations at the transition times region ( see figs ( [ a10 - 1 ] ) , ( [ a10 - 2 ] ) , ( [ a100 - 1 ] ) , ( [ a100 - 2 ] ) , ( [ a100 - 3 ] ) ) .this means that derivatives of the amplitude @xmath18 may reach extremely large values for some times from the transition time region and the modulus of these derivatives is much larger than the modulus of @xmath18 , which is very small for these times .this explains why in this time region the real and imaginary parts of @xmath161 , which can be expressed by the relation ( [ h ] ) , ie . by a large derivative of @xmath18 divided by a very small @xmath18 ,reach values much larger than the energy @xmath47 of the the unstable state measured at times for which the decay curve has the exponential form .for the model considered we found that , eg . for @xmath152 and @xmath162 the maximal value of the instantaneous energy equals @xmath163 and @xmath74 reaches this value for @xmath164 and then the survival probability @xmath16 is of order @xmath165 .the question is whether and where this effect can manifest itself .there are two possibilities to observe the above long time properties of unstable states : the first one is that one should analyze properties of unstable states having not too long values of @xmath43 .the second one is finding a possibility to observe a suitably large number of events , i.e. unstable particles , created by the same source .the problem with understanding the properties of broad resonances in the scalar sector ( @xmath4 meson problem @xcite ) discussed in @xcite , where the hypothesis was formulated that this problem could be connected with properties of the decay amplitude in the transition time region , seems to be possible manifestations of this effect and this problem refers to the first possibility mentioned above .there is the problem with determining the mass of broad resonances .the measured range of possible mass of @xmath4 meson is very wide , 400 1200 mev .so one can not exclude the possibility that the masses of some @xmath4 mesons are measured for times of order their lifetime and some of them for times where their instantaneous energy @xmath166 is much larger .this is exactly the case presented in fig .( [ h10 ] ) and fig .( [ h100 ] ) . for broad mesonsthe ratio @xmath167 is relatively small and thus the time @xmath43 when the above discussed effect occurs appears to be not too long .astrophysical and cosmological processes in which extremely huge numbers of unstable particles are created seem to be another possibility for the above discussed effect to become manifest . the probability @xmath168 that an unstable particle , say @xmath169 , survives up to time @xmath144 is extremely small .let @xmath16 be @xmath170 where @xmath171 , then there is a chance to observe some of particles @xmath169 survived at @xmath144 only if there is a source creating these particles in @xmath172 number such that @xmath173 so if a source exists that creates a flux containing @xmath174 unstable particles and @xmath175 then the probability theory states that the number @xmath176 unstable particles @xmath177 has to survive up to time @xmath144 .sources creating such numbers of unstable particles are known from cosmology and astrophysics .the big bang is the obvious example of such a source .some other examples include processes taking place in galactic nuclei ( galactic cores ) and inside stars , etc .so let us assume that we have an astrophysical source creating a sufficiently large number of unstable particles in unit of time and emitting a flux of these particles and that this flux is constant or slowly varying in time .consider as an example a flux of neutrons . from ( [ t - as-3 ] )it follows that for the neutron @xmath178 , where @xmath179 [ s ] .if the energies of these neutrons are of order @xmath180 [ ev ] then during time @xmath181 they can reach a distance @xmath182 [ ly ] , that is the distance of about a half of the milky way radius . now if in a unit ot time a suitably large number of neutrons @xmath183 of the energies mentioned is created by this source then in the distance @xmath184 from the source a number of spherically symmetric space areas ( halos ) surrounding the source , where neutron instantaneous energies @xmath185 are much larger than their rest energy @xmath186 , ( @xmath187 is the neutron rest mass and @xmath188 denotes its velocity ) have to appear ( see fig .( [ circles - a ] ) ) .of course this conclusion holds also for other unstable particles @xmath189 produced by this source .every kind of particles @xmath189 has its own halos located at distances @xmath190 , @xmath191 from the source .radiuses @xmath190 of these halos are determined by the particles velocities @xmath192 and by times @xmath193 when instantaneous energies @xmath194 have local maxima .unstable particles @xmath189 forming these halos and having instantaneous energies @xmath195 have to interact gravitationally with objects outside of these halos as particles of masses @xmath196 .the possible observable effects depend on the astrophysical source of these particles considered .if the halos are formed by unstable particles emitted as a result of internal star processes then in the case of very young stars cosmic dust and gases should be attracted by these halos as a result of a gravity attraction .so , the halos should be a places where the dust and gases condensate .thus in the case of very young stars one may consider the halos as the places where planets are born .on the other hand in the case of much older stars a presence of halos should manifest itself in tiny changes of velocities and accelerations of object moving in the considered planetary star system relating to those calculated without taking into account of the halos presence .if the halos are formed by unstable particles emitted by a galaxy core and these particles are such that the ratio @xmath197 is suitably large inside the halos , then rotational velocities of stars rounding the galaxy center outside the halos should differ from those calculated without taking into account the halos . thus the halos may affect the form of rotation curves of galaxies .( of course , we do do not assume that the sole factor affecting the form of the rotation curves are these halos ) .another possible effect is that the velocities of particles crossing these galactic halos should slightly vary in time due to gravitational interactions , i. e. they should gain some acceleration .this should cause charged particles to emit electromagnetic radiation when they cross the halo .note that the above mentioned effects seems to be possible to examine .all these effect are the simple consequence of the fact that the instantaneous energy @xmath194 of unstable particles becomes large compared with @xmath198 and for some times even extremely large . on the other hand this property of @xmath194 results from the rigorous analysis of properties of the quantum mechanical survival probability @xmath18 ( see ( [ a(t ) ] ) ) and from the assumption that the energy spectrum is bounded from below .10 s. krylov , v. a. fock , zh .* 17 * , ( 1947 ) , 93 .l. a. khalfin , zh .33 * , ( 1957 ) , 1371 [ sov . phys . jetp * 6 * , ( 1958 ) , 1053 ] .r. e. a. c. paley , n. wiener , _ fourier transforms in the complex domain _ , american mathematical society , new york , 1934 .m. l. goldberger , k. m. watson , _ collision theory _ , willey , new york 1964 .r. g. newton , _ scattering theory of waves and particles _ , springer , new york 1982 .l. fonda , g. c. ghirardii and a. rimini , rep .on prog . in phys .* 41 * , ( 1978 ) , 587 .a. peres , ann. phys . * 129 * , ( 1980 ) , 33 . p.t. greenland , nature * 335 * , ( 1988 ) , 298 .d. g. arbo , m. a. castagnino , f. h. gaioli and s. iguri , physica * a 227 * , ( 2000 ) , 469 . j. m. wessner , d. k. andreson and r. t. robiscoe , phys . rev . lett . * 29 * , ( 1972 ) , 1126 .e. b. norman , s. b. gazes , s. c. crane and d. a. bennet , phys .bf 60 , ( 1988 ) , 2246 .e. b. norman , b. sur , k. t. lesko , r .-larimer , phys . lett .b 357 * , ( 1995 ) , 521 . j. seke , w. n. herfort , phys . rev . * a 38 * , ( 1998 ) , 833 .r. e. parrot , j. lawrence , europhys .* 57 * , ( 2002 ) , 632 .j. lawrence , journ .b : quant . semiclass . opt . * 4 * , ( 2002 ) , s446 .i. joichi , sh . matsumoto , m. yoshimura , phys . rev .* d 58 * , ( 1998 ) , 045004 .t. jiitoh , s. matsumoto , j. sato , y. sato , k. takeda , phys rev . * a 71 * , ( 2005 ) , 012109 .c. rothe , s. i. hintschich and a. p. monkman , phys .96 * , ( 2006 ) , 163601 .k. urbanowski , phys . rev . * a 50 * , ( 1994 ) , 2847 . k. urbanowski , cent . eur . j. phys . * 7 * , ( 2009 ) , doi:10.2478/s1153400900535 .k. m. sluis , e. a. gislason , phys . rev . * a 43 * , ( 1991 ) , 4581 . ,no 55 , eds .m. abramowitz nad i. a. stegun ( u.s .gpo , washington , d.c . , 1964 ) .r. m. corless , g. h. gonet , d. e. g. hare , d. j. jeffrey and d. e. khnut , adv .comput . math . * 5 * , ( 1996 ) , 329 .f. w. j. olver , _ asymptotics and special functions _ , academic press , new york , 1974 . w. heitler , _ the quantum theory of radiation _ , oxford university press , london 1954 and dover publications , new york 1984 .k. urbanowski , eur .j. * d 54 * , ( 2009 ) , 25 .a. erdelyi , _ asymtotic expansions _ , dover pub ., new york 1956 .e. t. copson , _ asymtotic expansions _ , cambridge university press , 1965 .n. g. kelkar , m. nowakowski , k. p.khemchadani , phys . rev .* c 70 * , ( 2004 ) , 024601 . c. amsler _ at al _, phys . lett .b 667 * , ( 2008 ) , 1 . m. nowakowski , n. g. kelkar , _ nishiharima 2004 , penataquark proceedings of international workshop on penataquark 04 , spring 8 , hyogo , japan , 23 24 july 2004 _ , pp .182 189 ; arxiv : hep ph/0411317 .m. nowakowski , n. g. kelkar , _ aip conf .* 1030 * , ( 2008 ) , 250 255 ; arxiv:0807.5103 .

<S> an effect generated by the nonexponential behavior of the survival amplitude of an unstable state in the long time region is considered . in 1957 </S><S> khalfin proved that this amplitude tends to zero as @xmath0 more slowly than any exponential function of @xmath1 . for a time - dependent decay rate @xmath2 khalfin s result </S><S> means that this @xmath2 is not a constant for large @xmath1 but that it tends to zero as @xmath0 . </S><S> we find that a similar conclusion can be drawn for the instantaneous energy of the unstable state for a large class of models of unstable states : this energy tends to the minimal energy of the system @xmath3 as @xmath0 which is much smaller than the energy of this state for @xmath1 of the order of the lifetime of the considered state . analyzing the transition time region between exponential and non - exponential form of the survival amplitude we find that the instantaneous energy of a considered unstable state can take large values , much larger than the energy of this state for @xmath1 from the exponential time region . </S><S> taking into account results obtained for a model considered , it is hypothesized that this purely quantum mechanical effect may be responsible for the properties of broad resonances such as @xmath4 meson as well as having astrophysical and cosmological consequences . </S>

1212.1676

an optical coupler with one arm having losses and another one having gain , balanced against each other , recently became a test - bed for many phenomena originating from the interplay of the parity - time ( @xmath0 ) symmetry and nonlinearity .unidirectional dynamics @xcite , unversality of the dynamics @xcite , symmetry breaking properties @xcite , switching of the beams @xcite and of solitons @xcite , formation of symmetric and asymmetric bright solitary waves @xcite , breathers @xcite , and their stability @xcite , dark solitons @xcite , as well as the emergence of ghost states @xcite and large - scale temporal @xmath0-symmetric lattices @xcite are some among the many topics that have been touched upon in very intense recent theoretical and experimental work . as a direct extension of the previous activity ,a large chunk of which has focused on the prototypical setting of the @xmath0-symmetric dimer , there emerges a problem of effect of @xmath0 symmetry and nonlinearity on the polarization of the electric field . in that regard ,the previously proposed settings , to the best of our knowledge , were chiefly focused on effectively scalar models . on the other hand ,the `` vector '' type of problems is natural for experimental settings where the exploited fibers obey birefringence , since the two orthogonal polarizations are to be taken into account @xcite . in this workwe consider a @xmath0-symmetric coupler whose arms are birefringent waveguides .assuming that the first waveguide is active and the second one is absorbing , we address the problem of a @xmath0-preserving ( in the linear limit ) configuration . while being an interesting model from a physical point of view , this setting also offers a different ( in comparison to what was studied before ) mathematical situation where the nonlinear modes bifurcate from doubly degenerated eigenvalues of the linear problem .this requires the generalization of earlier developed approaches ( e.g. like the one reported in @xcite ) for the bifurcation of the nonlinear modes from the linear spectrum .in addition , it presents a rich playground for dynamical systems analysis , due to the emergence of a variety of saddle - center bifurcations ( nonlinear analogs of the linear @xmath0-phase transition ) , as well as symmetry - breaking ( pitchfork ) ones .it is these nonlinear states , their emergence , stability , dynamics and the asymptotics of the system that we will focus on hereafter .the organization of the paper is as follows . in section [ sec : model ] , we present the model in its evolution as well as in its stationary form . in section [ sec : lin ] , we focus on its linear properties .then , in section [ sec : nonlin ] , we examine the nonlinear modes ( and bifurcations ) .finally , in section [ sec : stab ] , we briefly touch upon the dynamical implications of our findings and in section [ sec : concl ] , we present some conclusions , as well as some potential directions for future work .we specify the problem by imposing that the principal optical axes of the two kerr - type waveguides are @xmath1-rotated with respect to each other , as it is schematically represented in fig .[ fig : coupler ] . in each arm , labeled by @xmath2, there are two orthogonal field components of the electric fields which we write down in the form @xcite ( @xmath3 ) : @xmath4\sqrt{\frac{2}{\chi}}e^{i\omega t}+c.c.\end{aligned}\ ] ] here @xmath5 are the field envelopes depending on the propagation distance @xmath6 , i.e. we consider the stationary in time problem , assuming that the carrier wavelength @xmath7 is in the region of the normal group velocity dispersion , thus ruling out a possibility of modulational instability ; @xmath8 is a transverse radius vector , and @xmath9 are the positions of the centers of the cores of the coupler .the real parameters @xmath10 are the propagation constants of each of the field components , and @xmath11 are the polarization vectors , which are mutually orthogonal in each arm of the coupler , i.e. @xmath12 . the real functions @xmath13 and @xmath14 describe the transverse distributions of the fields in each waveguide and the normalization coefficient @xmath15 , where @xmath16 is the kerr coefficient , is introduced for convenience . for @xmath17 and @xmath18 the functions @xmath19 are centered in different points @xmath20 .also , for the sake of simplicity , we consider @xmath21 ( for @xmath3 ) , such that the integral @xmath22 ( the integration is performed over the transverse plane ) describes the linear coupling between the respective modes . since in the configuration shown in fig .[ fig : coupler ] @xmath23 we use the single linear coupling coefficient @xmath24 ( see also @xcite ) .-symmetric coupler based on birefringent fibers .( b ) equivalent graph ( plaquette ) representation illustrating the @xmath0-symmetry . here@xmath25 and @xmath26 stand for lossy and active waveguides , respectively.,title="fig : " ] -symmetric coupler based on birefringent fibers .( b ) equivalent graph ( plaquette ) representation illustrating the @xmath0-symmetry . here@xmath25 and @xmath26 stand for lossy and active waveguides , respectively.,title="fig : " ] then following the analysis described in details in @xcite we end up with the system of equations : [ dynam ] @xmath27 here @xmath28 describes gain in the first waveguide and dissipation in the second waveguide , @xmath29 with @xmath30 , @xmath31 being the carrier wave frequency , is a properly normalized mismatch between the propagation constants of the orthogonal polarizations @xmath32 and @xmath5 .the asterisk stands for complex conjugation. we will be interested in the stationary solutions , in particular in their linear stability properties and ensuing nonlinear dynamics which can be found in the two prototypical limiting cases of ( i ) zero mismatches @xmath33 and ( ii ) large mismatches @xmath34 when the respective nonlinear terms can be neglected .it is convenient to introduce a parameter @xmath35 which vanishes ( @xmath36 ) in the case ( ii ) and is unity ( @xmath37 ) in the case ( i ) . using the standing wave ansatz @xmath38 , where @xmath39 are @xmath40independent, we obtain the system of algebraic equations : [ static ] @xmath41 below the spectral parameter @xmath42 will be also referred to as the propagation constant .first we address the underlying linear problem [ which corresponds to the situation when all cubic terms in eqs .( [ static ] ) are negligible ] .it can be rewritten in the matrix form @xmath43 where @xmath44 ( hereafter we use tilde in order to distinguish eigenvalues and eigenvectors of the linear problem ) .the operator @xmath45 is @xmath0 symmetric , which means that @xmath46 = h{{\cal pt}}- { { \cal pt}}h=0 $ ] , where @xmath47 is a spatial reversal linear operator @xmath48 and @xmath49 performs element - wise complex conjugation : @xmath50 .the spectrum of operator @xmath45 consists of _ two double eigenvalues _@xmath51 which are real for @xmath52 where @xmath53 will be referred to as a primary critical point : the spontaneous @xmath0 symmetry breaking occurs at @xmath54 above which the eigenvalues are all imaginary . in order to visualize the @xmath0 symmetry of the linear system , following @xciteone can represent it with a graph shown in the right panel of fig . [fig : coupler ] , reminiscent of four linearly coupled waveguides @xcite ( notice however the sign difference in the coupling constants ) or plaquettes @xcite .since the details of our analysis are the same for both eigenvalues @xmath55 and @xmath56 , we drop the subscripts @xmath26 and @xmath25 wherever this does not lead to confusion . despite having double eigenvalues in its spectrum, @xmath45 is diagonalizable below the @xmath0-symmetry breaking point .this means that double eigenvalues are semisimple , i.e. for an eigenvalue @xmath57 one can find two linearly independent eigenvectors , i.e. @xmath58 , where @xmath3 .moreover , each eigenvalue @xmath57 possesses an invariant subspace spanned by @xmath59 and @xmath60 .let us also notice the following peculiarity of the case at hand .in a situation where a @xmath0-symmetric operator @xmath45 has no multiple eigenvalues , the condition of unbroken @xmath0 symmetry ( i.e. reality of all the eigenvalues ) requires that for each eigenvalue @xmath57 the corresponding eigenvector @xmath61 can be chosen as an eigenstate of the @xmath0 operator , i.e. @xmath62 .however , in the situation at hand arbitrarily chosen linearly independent eigenvectors @xmath59 and @xmath60 may not be @xmath0 eigenstates .however , unbroken @xmath0 symmetry requires that a certain _ linear combination _ of @xmath59 and @xmath60 is an eigenstate for the @xmath0 operator .more specifically , it is easy to establish that all the eigenvectors that belong to the invariant subspace of @xmath57 and , at the same time , are the eigenstates for the @xmath0 operator , can be parametrized by a complex parameter @xmath63 as follows : @xmath64 being interested in linearly independent vectors @xmath61 , it is sufficient to consider only the vectors with @xmath65 then eq .( [ eq : pteigs ] ) yields a monoparametric set of eigenvectors @xmath66 with a real parameter @xmath67 .in particular , setting @xmath68 and @xmath69 one can choose two orthogonal ( and therefore linearly independent ) eigenvectors : @xmath70 ( hereafter we use the standard scalar product @xmath71 ) .any eigenvector @xmath61 corresponding to the eigenvalue @xmath57 can be represented as a linear combination of @xmath0 eigenstates @xmath72 : @xmath73 ( of course, this does not mean that any eigenvector @xmath61 is also a @xmath0 eigenstate ) .let us also introduce a hermitian adjoint operator @xmath74 .since the matrix @xmath45 is symmetric , one has @xmath75 .as long as @xmath0 symmetry of @xmath45 is unbroken , the spectrum of the adjoint operator @xmath74 also consists of two double eigenvalues @xmath76 which are semi - simple .any eigenvector corresponding to an eigenvalue @xmath57 of the adjoint operator @xmath74 can be represented as a linear combination of @xmath77 and @xmath78 .now we develop a perturbation theory for the eigenstates of the linear problem giving rise to monoparametric families of nonlinear modes. we will look for nonlinear modes @xmath79 that are eigenstates of the @xmath0 operator , i.e. @xmath80 . to this endwe introduce the expansions @xmath81 here @xmath82 is a small real parameter , @xmath83 and @xmath84 are the coefficients of the expansions and @xmath67 is to be determined from the symmetry of the solution ( see below ) .we notice that for the expansion to be meaningful the coefficient @xmath84 must be real . expansions ( [ expansion ] ) describe nonlinear modes that bifurcate from the _ linear limit _ corresponds to @xmath85 and is given by the eigenvector @xmath66 being a linear combination of @xmath72 such that @xmath86 . respectively , in the linear limit the parameter @xmath42 is given by the eigenvalue @xmath57 . passing from @xmath85 to @xmath87one has to compute the coefficients @xmath83 and @xmath84 . while the physical sense of the coefficient @xmath84 is clear it is a deviation of the propagation constant due to small nonlinearity , it turns out that @xmath84 also has a clear geometrical interpretation .indeed , let us consider the total energy flow through the coupler , which is defined by @xmath88 expansions ( [ expansion ] ) imply that in the vicinity of the linear limit @xmath89 .therefore , the coefficient @xmath84 governs a slope of the energy curve in the vicinity of the bifurcation point , i.e. @xmath90 for the sake of definiteness , now we concentrate on the case @xmath36. then the nonlinear problem ( [ static ] ) is conveniently written in the matrix form @xmath91 where @xmath92 is a diagonal matrix - function describing the nonlinearity : @xmath93 . substituting ( [ expansion ] ) into eq .( [ eq : matrixfoem ] ) , noticing that @xmath94 , and collecting the terms order of @xmath95 , we obtain @xmath96{\tilde{\bf w}}(\theta).\ ] ] equation ( [ eq : vep3 ] ) implies two possibilities .the first one corresponds to the case when at some @xmath67 the eigenvector @xmath66 of the operator @xmath45 is simultaneously an eigenvector for the matrix @xmath97 .then the coefficient @xmath84 can be chosen as an eigenvalue of @xmath97 corresponding to the eigenvector @xmath66 ( provided that this eigenvalue is real ) .in this situation , the right hand side of eq .( [ eq : vep3 ] ) is zero and it is sufficient to set @xmath98 .since the matrix @xmath97 is diagonal , its eigenvalues are equal to its diagonal elements and the corresponding eigenvectors are given as columns of the @xmath99 identity matrix .let us first assume that all the eigenvalues of @xmath97 are simple . in this case @xmath66 can not be an eigenvector for @xmath97 .the latter fact becomes evident if one notices that @xmath66 has no zero entries for any @xmath67 [ see the definition ( [ eq : pteigs ] ) ] .therefore , @xmath66 can be an eigenvector for @xmath97 only if @xmath97 has a multiple eigenvalue .then @xmath100 could be searched in the form of a linear combination of eigenvectors corresponding to the multiple eigenvalue . however , using the same argument , i.e. the fact that all entries of the vector @xmath66 are nonzero , one can see that even if @xmath97 has a double or an triple eigenvalue , the matrix @xmath97 still can not have @xmath66 among its eigenvectors .therefore @xmath66 can be an eigenvector of @xmath97 only if all its eigenvalues are equal . imposing this constraint on the matrix @xmath97 ,one obtains @xmath101 and @xmath102 . noticing that the form of @xmath100 implies , through eqs .( [ eq : pteigs ] ) and ( [ a ] ) , that @xmath103 we conclude that @xmath66 is an eigenvector of @xmath97 only if the moduli of all the entries of @xmath66 are equal to unity . then matrix @xmath97 is equal to the @xmath99 identity matrix multiplied by @xmath104 . requiring the moduli of all the entries of @xmath66 to be equal , one arrives at the equation for @xmath67 whose root is given as @xmath105 since now the moduli of all the entries of @xmath66 are equal to unity and therefore @xmath106 , @xmath107 and eq .( [ eq : slope ] ) readily yields that in the vicinity of @xmath57 the slope @xmath108 .notice that the found value does not depend on @xmath24 or @xmath109 , and thus these modes correspond to their counterpart in pure conservative coupler with birefringent arms with @xmath110 [ eq .( [ eq : thetasymm ] ) is valid in this case since @xmath111 .let us now consider the second possibility to fulfill eq .( [ eq : vep3 ] ) .if for some @xmath67 the corresponding @xmath66 is not an eigenvector for @xmath97 , then one must satisfy eq .( [ eq : vep3 ] ) choosing nonzero @xmath83 .then the coefficient @xmath84 is to be determined from the solvability condition which requires the right hand side of eq .( [ eq : vep3 ] ) to be orthogonal to all the eigenvectors of the invariant subspace of @xmath57 in the spectrum of the adjoint operator @xmath74 . as we have established in sec .[ sec : lin ] , any eigenvector of @xmath74 from the invariant subspace of @xmath57 can be represented as a linear combination of @xmath77 and @xmath78 . requiring the right hand side of eq .( [ eq : vep3 ] ) to be orthogonal to an arbitrary linear combination of @xmath77 and @xmath78 , we arrive at the following relations : @xmath112 in eq . ( [ eq : solv ] ) the first equality sign is an equation which is to be solved with respect to @xmath67 .once a root @xmath67 of the latter equation is found , then @xmath84 is given from the second equality sign .notice that despite the fact that the vector @xmath66 is complex , the coefficient @xmath84 will be real @xcite . substituting the expression for @xmath66 into eq .( [ eq : solv ] ) , one obtains a rather cumbersome equation , which , however can be attacked with a computer algebra program . after some transformations , eq .( [ eq : solv ] ) yields the following condition : latexmath:[\[\label{eq : thetaell } latter equation has a real root @xmath67 only if @xmath114 , i.e. when the expression under the radical is not positive .this result suggests that there exists a critical value of the gain - loss parameter which we term as the secondary critical point @xmath115 , such that for sufficiently small @xmath109 , namely , @xmath116 , there exists another family bifurcating from the eigenvalue @xmath57 of the linear spectrum .however , this family disappears for @xmath117 , in spite of the fact that the @xmath0 symmetry of the underlying linear problem remains unbroken , i.e. @xmath118 .it is important to point that the newly found family of solutions does _ not _ correspond to equal amplitude among the different nodes , and hence pertains to an elliptically ( rather than circularly ) polarized family of modes .computing the corresponding value of @xmath84 for the newly found elliptically polarized family , one finds that the coefficient @xmath84 does depend on @xmath24 and @xmath109 ( in contrast to the above considered circularly polarized family , characterized by @xmath107 for any combination of @xmath24 and @xmath109 which does not violate @xmath0 symmetry ) . in particular , when @xmath109 approaches @xmath24 , the coefficient @xmath84 of the elliptically polarized family tends to @xmath104 , which suggests that at @xmath119 the circularly polarized and elliptically polarized families globally merge .having explored nonlinear modes close to the linear limit , where the amplitudes of the modes are small and therefore they can be analyzed by means of perturbation theory , let us now consider nonlinear modes of arbitrary amplitudes ( turning again to the general case @xmath120 ) . relying on results of the previous subsectionwe firstly search for nonlinear modes which have equal intensities in all four waveguides . making the substitution @xmath121 and imposing the condition @xmath122 , @xmath123 , which is necessary for a nonlinear mode to be an eigenstate of @xmath0 thus leading to the circularly polarized light in each of the coupler arms , system ( 3 ) yields the single ( complex ) algebraic equation @xmath124 representing @xmath125, we obtain a bi - quadratic equation for @xmath126 yielding two families of the modes bifurcating from the eigenvalues @xmath127 , given by ( [ b ] ) , of the linear spectrum : @xmath128 respectively , the nonlinear modes have the following form : @xmath129 using eqs .( [ eq : rho ] ) one can easily obtain continuous families of nonlinear modes that can be identified as a function of the propagation constant @xmath42 , for given @xmath24 and @xmath109 .let us now recall that eq .( [ eq : thetaell ] ) predicts that for @xmath36 and @xmath130 , there exist families of elliptically polarized modes having different absolute values of the polarization vectors .such families were indeed found in our numerics .however , all such modes turned out to be unstable ( see figs .[ fig : fams][a0b2k1 ] and discussion below ) . in the case of zero propagationconstant mismatch , i.e. when @xmath37 , one also can find families which have different amplitudes of the polarization vectors .the explicit expressions for the families bifurcating from @xmath127 read @xmath131 remarkably , these modes , which also describe propagation of elliptically polarized light , are stable in a certain range of the parameters .the results of our analysis of the families of nonlinear modes are summarized in fig .[ fig : fams ] .the upper panels show that for @xmath132 each eigenvalue of the linearized problem gives rise to two distinct ( circularly and elliptically polarized ) families of nonlinear modes . in the case of @xmath36the slopes of the dependencies @xmath133 are close for the families of both types , and the elliptically polarized families are always unstable while circularly polarized families have both stable and unstable solutions . for @xmath37 one can find stable solutions both for the families with circular and for those with elliptical polarization . for @xmath134 and gain - loss parameters @xmath109 : @xmath135 ( the upper panels ) , @xmath136 ( the lower panels ). left and right columns correspond to @xmath36 and @xmath37 .stable and unstable modes are shown by by continuous and broken lines , respectively .the families with the circular and elliptical polarization ( if any ) are marked with labels `` _ _ ( c ) _ _ '' and `` _ _ ( e ) _ _ '' , respectively ( in the color online version families with the circular and elliptical polarization are also shown by blue and red lines , respectively ) . ] for @xmath137 , the case @xmath36 does not allow for elliptically polarized families [ see fig .[ fig : fams ] c ] , a feature which is in accordance with the perturbation approach developed above . in this caseone can only find circularly polarized modes , which are unstable . on the other hand , for @xmath37 ,stable and unstable modes of both types can be found [ see fig . [fig : fams ] d ] .summarizing at this point , we have identified 4 sets of solutions , two circularly polarized with equal amplitude at the nodes , and two elliptically polarized with unequal such amplitudes .these all degenerate into the two distinct eigenvalues @xmath127 , given by ( [ b ] ) , of the linear problem .the circularly polarized solutions are more robust , while the elliptically polarized ones are always unstable for @xmath36 and stable only for small enough amplitudes for @xmath37 . among the circularlypolarized ones , for @xmath138 the more fundamental state ( stemming from the negative eigenvalue at the linear limit ) is always the stable ground state of the system in continuations over the parameter @xmath42 , while the excited state is only stable for small enough amplitudes .an alternative and perhaps even more telling way to illustrate the above features stems from fixing some value of @xmath42 , starting from the hamiltonian limit of @xmath110 and subsequently identifying branches of the nonlinear modes by means of changing @xmath109 , as shown in fig .[ a0b2k1 ] .it is important to note that this alternative viewpoint affords us the ability to visualize bifurcations which we now explore .the relevant results for parametric continuations over @xmath109 are given in figs .[ a0b2k1]-[a1b2k1 ] ; typical examples of the corresponding linearization spectra for different values of @xmath109 can be found in fig .[ a0b2k1_s ] . here, it can be seen that a lower amplitude and a higher - amplitude intensity - symmetric ( i.e. , equal amplitude ) branch exist , for fixed @xmath42 , from the hamiltonian limit of @xmath110 and all the way up to the linear @xmath0-phase transition point @xmath139 . at that point ,the two equal amplitude branches collide and disappear in a saddle - center bifurcation which can be thought of as a nonlinear analog of the linear @xmath0-phase transition @xcite .an additional very interesting feature arises precisely at the point @xmath140 [ see ( [ eq : thetaell ] ) and the related discussion ] , where it can be seen that both branches of equal amplitude between the sites become dynamically unstable for @xmath36 .in fact , it is seen that for the larger amplitude branch ( associated with the blue circles ) , one pair of unstable eigenvalues arises , while for the smaller amplitude ( red diamond ) branch , two such pairs accompany the symmetry breaking bifurcation occurring at this critical point .a closer inspection reveals that the symmetric branch ( blue circles ) is destabilized through a subcritical pitchfork bifurcation with its `` corresponding '' asymmetric state ( i.e. , the one degenerate with it in the linear limit ) . in the case of the lower amplitude ( excited ) state for the same @xmath42 , the situation appears to be more complex . in particular, there exists once again a subcritical pitchfork with the corresponding asymmetric branch , yet this would justify one pair of unstable eigenvalues and we observe two . this is because at the _ same _ point , there also exists a supercritical pitchfork , which gives rise to the so - called ghost states , denoted by magenta plus symbols .these states are analogous to the ones to analyzed in @xcite , but remarkably are not stationary states of the original problem , yet they are pertinent to its dynamical ( instability ) evolution and for this reason they will be examined in further detail separately in the dynamics section below . in the case of @xmath37 ,only one pair of unstable eigenvalues emerges for the lower amplitude branch at the secondary critical point of @xmath140 ( while the larger amplitude branch remains stable throughout the continuation in @xmath109 ) .hence , in this case , once again a saddle - center bifurcation will mark the nonlinear @xmath0-phase transition , yet the number of unstable eigendirections of each symmetric branch ( fundamental and excited ) is decreased by one ( 0 and 1 real pairs instead of 1 and 2 , respectively , for @xmath36 ) . in this case ,in fact , both asymmetric branches persist up to the linear @xmath0-phase transition ( rather than terminate in a subcritical pitchfork as above ) , and collide and disappear with each other .interestingly all 3 branches ( the lower amplitude , excited symmetric one and the two asymmetric ones ) become unstable at the secondary critical point @xmath115 , which again points to the existence of corresponding ghost states . for the lower amplitude symmetric branch ,the bifurcating ghost states are again identified by the magenta plus symbols in fig .[ a1b2k1 ] . ,and the bottom one is when @xmath37 . in the case@xmath36 , at @xmath141 , one pair of blue circles and two pairs of red diamonds collide at 0 so that one pair of real eigenvalues arises in blue circles branch whereas two pairs of real eigenvalues arise in the red diamonds branch .the asymmetric branches only exist ( and are unstable ) for the smaller value of @xmath135 , for @xmath36 .for the same parameters ( @xmath142 , @xmath134 ) , in the case of @xmath37 , the excited symmetric and both asymmetric branches are unstable for @xmath143.,title="fig : " ] , and the bottom one is when @xmath37 . in the case@xmath36 , at @xmath141 , one pair of blue circles and two pairs of red diamonds collide at 0 so that one pair of real eigenvalues arises in blue circles branch whereas two pairs of real eigenvalues arise in the red diamonds branch .the asymmetric branches only exist ( and are unstable ) for the smaller value of @xmath135 , for @xmath36 .for the same parameters ( @xmath142 , @xmath134 ) , in the case of @xmath37 , the excited symmetric and both asymmetric branches are unstable for @xmath143.,title="fig : " ] , and the bottom one is when @xmath37 . in the case@xmath36 , at @xmath141 , one pair of blue circles and two pairs of red diamonds collide at 0 so that one pair of real eigenvalues arises in blue circles branch whereas two pairs of real eigenvalues arise in the red diamonds branch .the asymmetric branches only exist ( and are unstable ) for the smaller value of @xmath135 , for @xmath36 .for the same parameters ( @xmath142 , @xmath134 ) , in the case of @xmath37 , the excited symmetric and both asymmetric branches are unstable for @xmath143.,title="fig : " ]to examine the potentially symmetry breaking ( and more generally instability driven ) nature of the dynamical evolution past the critical points identified above , we have also performed direct numerical simulations which are illustrated in fig .[ fig : dyn1 ] ; see also fig . [ a0b2k1_ghostb ] . here, it can be seen that although the relevant parameters are below the critical point for the linear @xmath0-phase transition @xmath139 , nevertheless , symmetry breaking phenomena are observed due to the dynamical instability of the relevant states ( the ones denoted by dashed lines in fig .[ fig : fams ] ) .this dynamics may , in principle , be associated with the so - called ghost states of complex propagation constant that have recently been proposed as relevant for the dynamical evolution in @xcite .to substantiate this claim , we note that it is observed in the left panel of fig .[ fig : dyn1 ] that the relative phase of the two gain sites that lock into an equal growing amplitude , is @xmath144 , as is those of the decaying amplitude lossy sites . in light of this , we seek ghost states with precisely this phase difference and are able to explicitly identify them via the ansatz @xmath145 , setting @xmath146 for @xmath3 . for these branches ,the propagation constant is complex .this highlights the potential growth or decay of such states .importantly also , note that these states are `` ghosts '' because they may be solving the stationary problem of eqs .( [ static ] ) , but the u@xmath147 invariance of the original model does _ not _ permit them to be a solution of the dynamical eqs .( [ dynam ] ) .the algebraic conditions that this family of solutions satisfies are @xmath148 notice that the imaginary part of the propagation constant @xmath149 is proportional to the difference @xmath150 .hence , prior to the symmetry breaking , the relevant solutions bear a real propagation constant . past the bifurcation point one ( unstable )branch has @xmath151 , while the stable branch has @xmath152 .the relevant ghost state branches and their bifurcation from the equal amplitude ones are explored in fig .[ a0b2k1]-[a1b2k1 ] . given that these are only ghost solutions of the original dynamical problem , the interpretation of their linearization spectrum ( shown for completeness in fig .[ a0b2k1_s ] ) is still an open problem .these ghost states appear , in fact , to exhibit very similar evolution dynamics to regular unstable states .to illustrate this , we observed the particular behavior of the unstable modes and how it depends on the form of the initial perturbation .a typical example in which the gain sites lead to growth and the lossy sites to decay is shown in the left panel of fig .[ fig : dyn1 ] .it is interesting that the evolution appears to be very proximal to that of the ghost states identified above .this is clearly showcased in fig .[ a0b2k1_ghostb ] , through the comparison of the growth pattern in the relevant sites ( and the decay pattern in the lossy sites ) with the exact , shifted in the propagation distance to fit the onset of growth , ghost state solution for the same parameters . on the other hand , in the right panel of fig .[ fig : dyn1 ] a different scenario of evolution is illustrated . instead of the gain nodes growing and the lossy ones decaying , a breathing oscillation settles between the two pairs .these two scenarios , illustrated in fig . [ fig : dyn1 ] , are the prototypical instability evolution ones that we have obtained in this system . , @xmath135 and for @xmath36 for two different small initial perturbations .the result of the evolution of the left panel involves growth at the gain sites ( @xmath153 , red curves in the color online version ) and decay at the lossy sites ( @xmath154 , blue curves in the color online version ) .notice that intensities among the two gain sites and among the two lossy sites are approximately equal ( @xmath155 and @xmath156 ) and are not distinguishable in the scale of the plots . in the right panel only the initial stage of the found persistent periodic dynamics is shown ; the simulations were performed up to @xmath157 . ] , @xmath142 , and @xmath134 for @xmath158 .the red solid line and blue dashed line correspond to the red diamond and magenta plus branches in fig .[ a0b2k1 ] , respectively .the blue dashed line is plotted with a shift in time , i.e. delay by @xmath159 .the time axis in the plot is the actual time of the red solid line .their dynamical behaviors are essentially identical ( and can not be distinguished on the scale of the plot ) , showcasing how the unstable dynamics of the red diamond branch is similar to that of the bifurcating ghost . ]in conclusion , in the present work , we have proposed a novel , physically realistic variant of a @xmath0 symmetric dimer where the effect of birefringence has been taken into consideration . the existence of polarization of the electric field within the coupler yields two complex dynamical equations for each of the fibers , providing a physical realization of a plaquette model with _ both linear and nonlinear _ coupling between the elements .the stationary states of the model were identified and both linear and nonlinear @xmath0-phase transitions were obtained .the degenerate nature of the linear limit complicated the problem in comparison to other ones studied earlier in this context .furthermore , the emergence of symmetry breaking phenomena and associated ( subcritical or supercritical ) pitchfork bifurcations , as well as their dynamical implications in leading to indefinite growth and decay ( of the corresponding waveguide amplitudes ) were elucidated .a connection was also given to ghost states . from the physical point of view, we emphasize that the use of @xmath0 symmetry significantly changes the stability of the modes with different polarizations . for example , in the case of large propagation constant mismatches , we saw that the symmetric ( circularly polarized ) states become destabilized in the presence of gain / loss and may even cease to exist past a certain critical point of the relevant parameter . instead of them, the dynamics may lead to a breathing exchange of `` mass '' between the gain and the lossy waveguides , or most commonly an indefinite growth of the former at the expense of the latter .this is in line with what is observed also in the evolution of the emerging ghost states of the system .the instabilities and associated dynamics should be observable in suitable generalizations of existing experiments such as @xcite .such properties are relevant also to possibilities of solitonic ( waveguide - array ) generalizations of the coupler system as well as towards a possible use of @xmath0 symmetric coupler for measurement techniques based on the use of several nonlinear modes . from the mathematical point of view, we believe that these studies may pave the way for considering multi - component , as well as multi - dimensional ( generalizing the plaquettes considered here or those of @xcite ) lattice models of @xmath0-symmetric form . in generalizing to multi - plaquette configurations , it would be especially interesting to examine which of the symmetry - breaking and nonlinear @xmath0-phase transition phenomena examined herein are preserved and what new phenomena may arise as additional degrees of freedom are added .such studies will be deferred to future publications .vvk and daz acknowledge support of the fct grants ptdc / fis/112624/2009 , pest - oe / fis / ui0618/2011 , and sfrh / bpd/64835/2009 .pgk gratefully acknowledges support from the alexander von humboldt foundation , and the alexander s. onassis public benefit foundation , as well as from the us - nsf ( grants dms-0806762 , cmmi-1000337 ) and from the us - afosr .a. w. snyder and y. chen , opt* 14 * , 517 ( 1989 ) ; y. chen , a. w. snyder , and d. n. payne , ieee j. quant .* 28 * , 239 ( 1992 ) ; m. romagnoli s. trillo , s. wabnitz , opt . andquant . electron . * 24 * , s1237 ( 1992 ) .k. li , p.g .kevrekidis , b. a. malomed and u. guenther , j. phys .a : math . theor . * 45 * 444021 ( 2012 ) .v. achilleos , p.g .kevrekidis , d.j .frantzeskakis , and r. carretero - gonzlez phys . rev .a * 86 * , 013808 ( 2012 ) .

<S> in this work , we propose a parity - time ( @xmath0- ) symmetric optical coupler whose arms are birefringent waveguides as a realistic physical model which leads to a so - called quadrimer i.e. , a four complex field setting . </S><S> we seek stationary solutions of the resulting linear and nonlinear model , identifying its linear point of @xmath0 symmetry breaking and examining the corresponding nonlinear solutions that persist up to this point , as well as , so - called , ghost states that bifurcate from them . </S><S> we obtain the relevant symmetry breaking bifurcations between symmetric ( circularly polarized ) and asymmetric ( elliptically polarized ) states and numerically follow the associated dynamics which give rise to growth / decay even within the @xmath0-symmetric phase . </S><S> our symmetric stationary nonlinear solutions are found to terminate in saddle - center bifurcations which are analogous to the linear @xmath0-phase transition . </S><S> we found that the @xmath0 symmetry significantly changes the stability and dynamical properties of the modes with different polarizations . </S>

1303.2647

there have been extensive investigations of the phase diagrams of the various iron arsenide and chalcogenide structures that display high temperature superconductivity with critical temperatures up to 55 k @xcite . in common with other unconventional superconductors , such as the copper oxides , heavy fermions , and organic charge - transfer salts ,superconductivity is induced by suppressing a magnetically ordered phase , which generates a high density of magnetic fluctuations that could theoretically bind the cooper pairs . whether this is responsible for the high transition temperatureshas not been conclusively established , but it makes the origin of the magnetic interactions an important issue to be resolved @xcite . in nearly all the iron arsenides and chalcogenides , the iron atoms form a square planar net and the magnetic orderconsists of ferromagnetic stripes along one iron - iron bond direction that are antiferromagnetically aligned along the orthogonal iron - iron bond @xcite .these systems are metallic and the fermi surfaces , which are formed by the iron 3@xmath0 electrons , are nearly cylindrical with hole pockets at the centre of the brillouin zone and electron pockets at the zone boundaries , all of similar size . in such an electronic structure ,interactions between electrons near the hole and electron pockets give rise to a spin density wave ( sdw ) order at the wavevector connecting the two pockets @xcite .this itinerant picture is consistent with the wavevector of the observed antiferromagnetism , angle resolved photoemission ( arpes ) measurements of the electronic structure @xcite , and the evolution of the dynamic magnetic susceptibility with carrier concentration @xcite . however , any theory of the magnetic order also has to explain the structural transition which occurs at a temperature either just above or coincident with the sdw transition and lowers the symmetry from tetragonal ( @xmath6 ) to orthorhombic ( @xmath7 ) .this is often referred to as nematic order , and the relation between nematicity , magnetic order , and superconductivity has become one of the central questions in the iron - based superconductors @xcite . at present , there are two scenarios for the development of the nematic order and its relation to sdw order . in the first , the structural order is unrelated to magnetism and is driven by orbital ordering as the primary instability .the orbital ordering induces magnetic anisotropy and triggers the magnetic transition at a lower temperature by renormalizing the exchange constants @xcite .this scenario is largely phenomenological , but there have been recent efforts to put it into a microscopic basis @xcite . in the second scenario ,the structural order is driven by magnetic fluctuations , associated with the fact that striped sdw order can be along the @xmath8-axis ( ordered momentum is @xmath9 ) or along the @xmath10-axis ( ordered momentum is @xmath11 ) .theory predicts that the @xmath12 symmetry between the @xmath13 and @xmath14 directions can be broken above the true sdw ordering temperature that breaks @xmath15 spin symmetry , _ i.e. _ , the system distinguishes between @xmath16 and @xmath17 without breaking time - reversal symmetry @xcite .the order parameter of this ising - spin - nematic state " couples linearly to the lattice , inducing both structural and orbital order .the magnetic scenario has been developed for itinerant @xcite and localized @xcite fermions , and the phase diagrams are rather similar in the two approaches .below we use the fact that the systems we study are metals and use an itinerant approach .it is difficult to distinguish the orbital and the magnetic scenarios since many of the observable properties are identical .we have now identified both theoretically and experimentally a new magnetic phase that can distinguish the two scenarios . by extending our previous analysis of the itinerant model to higher carrier concentrations, we have discovered that the phase diagram is richer than previously anticipated , based on the ginzburg - landau analysis near @xmath18 , and that there exists , in some range of parameters , an additional phase at @xmath2 , that restores @xmath6 symmetry .the order parameter in this new phase combines @xmath16 _ and _ @xmath17 with equal weights . in the following ,we summarize the results of these calculations , which are explained in more detail in the supplementary information , and show experimental evidence for the predicted phase in a hole - doped superconductor , ba@xmath3na@xmath4fe@xmath5as@xmath5 .such a second transition is highly unlikely in the orbital scenario because the breaking of the symmetry between @xmath16 and @xmath17 is the pre - condition for a magnetic transition .we view the observation of the transition to an sdw state which does not break the symmetry between @xmath16 and @xmath17 as a strong indication that the nematic order is of magnetic origin rather than of orbital origin .the itinerant description of magnetism in iron - based superconductors is built upon the fact that the hole bands are centered around @xmath19 and the electron bands are centered at @xmath20 and @xmath21 , respectively ( fig .[ fig1](a ) ) .a spin susceptibility is logarithmically enhanced at momenta connecting hole and electron pockets , and sdw order develops even if the interaction is weak .the sdw order parameter is in general a combination of the two vector components @xmath22 and @xmath23 with momenta @xmath24 and @xmath25 , respectively . for a model of perfect fermi surface nesting ( circular hole and electron pockets of equal radii ) and only electron - hole interactions, sdw order specifies @xmath26 but not the relative magnitudes and directions of @xmath22 and @xmath23 .away from perfect nesting , the ellipticity of the electron pockets and interactions between the electron bands break the degeneracy and lower the symmetry of the sdw order . near @xmath18 ,an analysis within a ginzburg - landau expansion in powers of @xmath22 and @xmath23 shows that fourth - order terms select stripe magnetic order with either @xmath27 , @xmath28 , or @xmath29 , @xmath30 @xcite .such an order simultaneously reduces the lattice @xmath6 symmetry down to @xmath7 .the order parameter in the stripe phase is shown schematically in fig.[fig1](b ) .an issue that has not been discussed in detail until now is whether another magnetic ground state , in which both @xmath22 and @xmath23 are non - zero , may appear at a lower temperature , as a result of non - linear effects .this might happen either via a first - order transition , in which case the most likely outcome is the state in which @xmath31 ( see fig.[fig1](c ) ) , or via a second - order transition , in which case the second order parameter appears continuously and likely remains relatively small down to @xmath32 ( see fig.[fig1](d ) ) .to check for a potential second sdw transition , we extended the previous ginzburg - landau analysis to the entire temperature range by solving non - linear coupled mean - field equations for @xmath33 and @xmath34 and analyzing which solution minimizes the free energy .this has revealed new features in the phase diagram not previously identified . in particular, we find that sdw order with @xmath35 , which breaks @xmath15 spin symmetry but preserves lattice @xmath6 symmetry , does emerge at low @xmath1 , as the mismatch in hole and electron pocket sizes grows . and two electron pockets at x and y , using the unfolded brillouin zone with one fe atom per unit cell .the arrows refer to two equivalent nesting wavevectors * q*@xmath36 and @xmath37 ( lower panel ) possible magnetic ground states of the fe - lattice : ( b ) the @xmath7 antiferromagnetic stripe phase with @xmath38 and @xmath39 ;( c ) a @xmath6 magnetic state , in which @xmath40 , that is compatible with tetragonal lattice symmetry ( other possible @xmath6 magnetic structures are discussed in the supplementary information ) ;( d ) magnetic order in which @xmath41 . note that the nel transition at low temperatures , @xmath42 , from phase ( b ) to phase ( c ) is first order , while the transition from phase ( b ) to phase ( d ) is of the second order in which an additional small component of @xmath43 appears below t@xmath44 .our experiments are more compatible with scenario ( c ) . [ fig1 ] ] we obtained this result by analyzing the minimal three - band model with one hole and two electron pockets . for simplicity , we considered parabolic dispersions with @xmath45 where @xmath46 are band masses , @xmath47 is the offset energy , and @xmath48 is the chemical potential .the non - interacting hamiltonian takes the form @xmath49 where @xmath50 label the bands , the summation over repeated spin indices @xmath51 is assumed , and we shift the momenta of the fermions near the @xmath13 and @xmath14 fermi pockets by @xmath16 and @xmath17 , respectively , writing @xmath52 , @xmath53 .the interaction term in the hamiltonian @xmath54 contains all symmetry - allowed interactions between low - energy fermions , which include inter- and intra - band scattering processes @xcite .we present the explicit form of @xmath54 in the supplementary information .the mean - field equations on @xmath22 and @xmath23 are obtained by introducing @xmath55 and @xmath56 and using them to decouple four - fermion terms into anomalous quadratic terms with inter - band `` hopping '' , which depends on @xmath22 and @xmath23 .we diagonalized the quadratic form , re - expressed @xmath57 in terms of new operators and obtained a set of two coupled self - consistent equations for @xmath22 and @xmath23 .the blue points are the coincident antiferromagnetic and orthorhombic transition temperatures , @xmath18 , into the @xmath7 phase , and the red point is the observed transition temperature into the tetragonal @xmath6 phase , all measured by neutron diffraction .the green points are the superconducting transition temperatures , @xmath59 , determined from magnetization data .[ fig2 ] ] we solved the mean - field equations numerically as a function of two parameters , @xmath60 and @xmath61 ( see supplementary information for details ) .the parameter @xmath62 represents the mismatch in chemical potentials of the hole and electron pockets ( @xmath63 when the electron and hole pockets are identical ) .@xmath64 is proportional to the ellipticity of the electron pockets .we focused on the two sdw - ordered states the antiferromagnetic stripe state with @xmath27 and @xmath28 , in which @xmath6-symmetry is reduced to @xmath7 and on the sdw state with @xmath35 , in which @xmath6-symmetry is preserved .as we said , the two states are degenerate at zero ellipticity and perfect nesting , when @xmath65 .once the ellipticity becomes non - zero , the stripe state wins immediately below the neel temperature @xmath18 .the @xmath6-preserving state ( with @xmath35 ) is a local maximum and is unstable at @xmath66 . by solving the equations at lower temperature, we found that , at a finite @xmath60 , the @xmath6-preserving state also becomes locally stable below some @xmath67 , and at an even lower @xmath68 , its free energy becomes smaller than that of the stripe phase , _i.e. _ , at @xmath69 the system undergoes a first - order phase transition in which lattice @xmath6 symmetry gets restored .( see fig.[fig1](c ) ) . because @xmath18 fallsas the fermi surface mismatch @xmath60 increases , the new @xmath6-preserving phase in practice exists only in a narrow region of the phase diagram close to the suppression of sdw order .this behavior is illustrated in fig .we also analyzed a four - pocket model with two hole pockets and found another scenario for a second sdw transition .namely , the af stripe order @xmath70 initially involves only fermions from a hole pocket which has higher density of states .below some @xmath71 , fermions near the remaining hole pocket and near the electron pocket at @xmath13 , not involved in the initial stripe order , also produce sdw instability , and the system gradually develops the second order parameter @xmath72 , which distorts the stripe af order .the corresponding low @xmath1 spin configuration is shown in fig.[fig1](d ) . in this case , however , the @xmath6 symmetry remains broken at all temperatures .our experimental data taken as a function of doping are more consistent with a first - order transition and restoration of @xmath6 symmetry ( see below ) .it is possible , however , that the second scenario is realized under pressure @xcite .we are currently exploring the phase diagram of ba@xmath73na@xmath58fe@xmath5as@xmath5 using neutron and x - ray powder diffraction , following our recent investigation of the potassium - doped compounds @xcite .details of the synthesis and characterization of the polycrystalline samples and the powder diffraction measurements are given in the methods section below .in both series , the addition of the alkali metal dopes holes into iron @xmath0-bands , which is equivalent to reducing the chemical potential , @xmath48 , in the model described above . this has the effect of reducing the transition temperature into the stripe phase from 139k , in the parent compound bafe@xmath5as@xmath5 , to 0 at @xmath74 in both the k - doped and na - doped series .one unusual feature of both series is that the antiferromagnetic and orthorhombic transitions are coincident and first - order over the entire phase diagram @xcite , an observation that is quite unambiguous since both order parameters are determined from the same neutron powder diffraction measurement .( a ) a neutron diffractogram from the wish diffractometer showing the orthorhombic splitting of the ( 112 ) bragg peak at @xmath75k and its recombination at the reentrant tetragonal phase at 38k .( b ) the lattice parameters , @xmath76 ( blue points ) in the tetragonal phase of @xmath77 symmetry above @xmath18 and below the reentrant transition , and @xmath76 ( red points ) and @xmath78 ( black points ) in the orthorhombic phase of @xmath79 symmetry below @xmath18 .the tetragonal lattice parameter is multiplied by @xmath80 .[ fig3 ] ] we will publish a more detailed report on the whole phase diagram of ba@xmath73na@xmath58fe@xmath5as@xmath5 in a forthcoming publication , but to understand its significance to this article , we show in fig .2 the transition temperatures into the antiferromagnetic / orthorhombic ( af / o ) @xmath7 ( @xmath18 ) and superconducting phases ( @xmath59 ) up to @xmath81 . in ba@xmath73k@xmath58fe@xmath5as@xmath5 , it was possible to fit @xmath18 _ vs _ @xmath8 to a scaling form , _i.e. _ , to @xmath82 , with @xmath83 and @xmath84 @xcite .we do not have as many data points in the sodium - doped system , but the phase diagram is extremely similar so we have assumed that the same scaling curve applies here as well .this scaling has not been observed in the other iron - based superconductors and may reflect the reduced disorder in these hole - doped compounds , in which the doping occurs away from the iron arsenide layers .phase ( red points ) and the orthorhombic @xmath7 phase ( black points ) below the reentrant transition at 40k derived from two phase rietveld refinements .( b ) the temperature dependence of the magnetic moment in the tetragonal @xmath6 phase ( red points ) , and the magnetic moment ( black points ) and orthorhombic order parameter @xmath85 ( gold points ) in the orthorhombic @xmath7 phase both above and below the reentrant phase transition .[ fig4 ] ] the only member of the sodium series to show a significant departure from the potassium compounds is at @xmath86 close to the suppression of the af / o order , which has a superconducting transition temperature of 24k .the transition into the conventional @xmath7 phase occurs at 90k .however , at @xmath87k , there is a remarkable first - order phase transition , at which the sample phase separates into regions with two different atomic and magnetic structures .3(a ) shows the orthorhombic splitting of the ( 112 ) nuclear bragg peak at 90k and its recombination at @xmath42 marking the onset of a reentrant tetragonal phase .not all the sample transforms back to @xmath6 symmetry . from a multi - phase rietveld refinement, we estimate that approximately 40% of the sample remains in the orthorhombic phase below the transition ( fig .it is nevertheless clear that the whole sample is involved in the first - order transition even if only a fraction is transformed because , even within the orthorhombic phase , the in - plane lattice parameters relax below the transition ( fig .the in - plane lattice parameters show a reduction in the low - temperature @xmath7 phase with an increase in the @xmath88-axis lattice parameter , _i.e. _ , the high- and low - temperature @xmath7 phases are not identical .the @xmath6 phase shows the opposite trend , which means that the in - plane lattice parameter is slightly higher than above @xmath18 .the overall volume is roughly constant through the reentrant transition for both phases .both the high - temperature and low - temperature @xmath7 phases have identical magnetic order with the moment in the plane , equivalent to the conventional af stripe phase observed at other compositions . within the high - temperature phase ,the temperature dependence of the magnetic and orthorhombic order parameters are directly proportional ( fig .4(b ) ) , as we saw in the k - doped series @xcite .there is some evidence of phase competition above the reentrant transition since the moment in the low - temperature orthorhombic phase increases abruptly at the transition .the @xmath6 phase is also magnetically ordered .an unambiguous refinement of the magnetic structure will require measurements on single crystals , but possible solutions that are consistent with the structural space group ( @xmath77 ) and the lack of any in - plane magnetoelastic distortion are discussed in the supplementary information .they include the in - plane magnetic order illustrated in fig .1(c ) , as well as variants , in which the moments are aligned along the @xmath88-axis . since the theory does not include spin - orbit coupling , both solutions are theoretically allowed .it is important to distinguish these new results from previous observations of a reentrant tetragonal phase in electron - doped compounds , such as bafe@xmath89co@xmath58as@xmath5 @xcite .all those transitions were within the superconducting phase and have been convincingly shown to result from the competition between superconductivity and antiferromagnetic stripe order @xcite .the reentrant phase that we report here occurs well above the superconducting transition temperature and so requires a different explanation .we are also unaware of any model of orbital order that would predict a reentrant non - orbitally - ordered phase at lower temperature .our results are broadly consistent with the itinerant model described above and are a natural consequence of magnetic order induced by the interaction between the hole and electron pockets suggested by the fermi surface geometry .the relaxation of the lattice parameters below the reentrant transition implies that there is some phase competition between the @xmath7 and @xmath6 phases in the high - temperature phase that is relieved below @xmath42 . in principle , this could be modeled theoretically by extending the mean - field theory calculations performed so far to the more microscopic theory .such a theory would be necessary to study a @xmath7 phase above the reentrant transition in which @xmath90 but both are non - zero .nevertheless , the prediction that a @xmath6 phase can become degenerate with the @xmath7 phase only at higher doping when the hole and electron fermi surfaces are not as well - matched in size , and that the stability of the @xmath6 phase would be limited to a very narrow region close to the suppression of antiferromagnetism is borne out by the new data .our results provide strong evidence for the validity of an itinerant model of nematic order in the iron - based superconductors , in which the orbital reconstruction of the iron @xmath91 states is a consequence of magnetic interactions induced by fermi surface nesting .whether nematic order , or at least strong nematic fluctuations , is a prerequisite for superconductivity is another challenge to address in the future .we are thankful to j. knolle , r. fernandes , j. schmalian , r. moessner , and v. stanev for useful discussions .this work was supported by the materials sciences and engineering division of the office of basic energy sciences , office of science , u.s .department of energy , under contract # de - ac02 - 06ch11357 ( s.a ., o.c . , s. r. , j.a . ,, j .- p.c ., j.a.s . ,h.c . , r.o . ) and grant # de - fg02-er46900 ( a.v.c . ) .i.e. acknowledges financial support from the german academic exchange service ( daad ppp usa grant no .50750339 ) .10 url # 1`#1`urlprefix[2]#2 [ 2][]#2 ._ _ * * , ( ) . & ._ _ * * , ( ) . ._ _ * * , ( ) . & ._ _ * * , ( ) . , & ._ _ * * , ( ) . & ._ _ * * , ( ) ._ et al . _ . __ * * , ( ) . & ._ _ * * , ( ) ._ et al . _ . __ * * , ( ) ._ et al . _ ._ _ * * , ( ) ._ et al . _ . __ * * , ( ) ._ et al . _ . __ * * , ( ) ._ et al . _ . __ * * , ( ) ._ et al . _ . __ * * , ( ) ._ et al . _ . _ _ ( ) . , , & . __ * * , ( ) . , &_ _ * * , ( ) ._ et al . _ . _* * , ( ) . , & ._ _ * * , ( ) . , , , &_ _ * * , ( ) . , & ._ _ * * , ( ) . , & ._ _ * * , ( ) . , , , & ._ _ * * , ( ) . , & ._ _ * * , ( ) . , & ._ _ * * , ( ) ._ et al . _ . __ * * , ( ) ._ et al . _ . __ * * , ( ) ._ et al . _ . _ _ * * , ( ) ._ et al . _ . __ * * , ( ) ._ et al . _ . __ * * , ( ) . , & ._ _ * * , ( ) .the synthesis of homogeneous single phase ba@xmath73na@xmath58fe@xmath5as@xmath5 polycrystalline powders was performed using conditions that are similar to ba@xmath73k@xmath58fe@xmath5as@xmath5 .further details of the synthesis can be found in ref .handling of all starting materials was carried out in a n@xmath5-filled glove box .mixtures of ba , na , and feas were loaded in alumina tubes , sealed in niobium tubes under argon , and sealed again in quartz tubes under vacuum .the mixtures were heated at 850@xmath92c for one day and ground to homogeneous powders and then reheated for 3 days at 850@xmath92c .this last step was performed twice .initial characterization of the the final black powders was performed by x - ray diffraction , magnetization measurements , and inductively coupled plasma ( icp ) elemental analysis .powder diffraction experiments were performed on three time - of - flight neutron diffractometers , hrpd ( x = 0 , 0.24 ) and wish ( 0.15 , 0.24 ) at the isis pulsed neutron source ( rutherford appleton laboratory ) and hipd ( 0.1 , 0.2 , 0.3 ) at the lujan neutron scattering center ( los alamos national laboratory ) .synchrotron x - ray diffraction measurements were performed as a function of temperature for x = 0 , 0.1 , 0.15 , 0.2 at 11bm - b at the advanced photon source of argonne national laboratory .the nuclear and magnetic structures and internal parameters were determined by the rietveld technique using gsas @xcite and expgui @xcite .the magnetic phase in most parent compounds of the iron - based superconductors is the stripe spin - density wave order with momentum either @xmath93 or @xmath94 in the unfolded brillouin zone , in which there is one iron atom per unit cell @xcite ( see fig .1(b ) in the main manuscript ) .these two wavevectors connect the hole pockets at the center of the brillouin zone and the electron pockets at the zone boundary , along the two orthogonal iron - iron bond directions .the stripe magnetic ordering breaks both @xmath15 spin - rotational symmetry and @xmath6 lattice rotational symmetry ( by selecting either @xmath16 or @xmath17 ) , and is often preceded by a `` nematic '' phase in which @xmath95 symmetry is broken , but @xmath15 rotational symmetry remains unbroken .in general , the geometry of iron pnictides allows more complex orders in which both @xmath22 and @xmath23 are present .previous analysis by two of us @xcite have shown that near @xmath18 , when ginzburg - landau theory is valid , the system definitely prefers a stripe order in which only the order parameter with @xmath16 or @xmath17 is non - zero . herewe extend the previous analysis to lower t and study the magnetic order between @xmath96 and @xmath32 .our results show that the phase diagram of the system is more complex than previously thought .in particular , we find that at some range of dopings , the system undergoes a first - order transition , upon lowering @xmath1 , into a phase in which there is a two - component order parameter with equal magnitudes of the components with @xmath16 and @xmath17 , and the four - fold lattice rotational symmetry is restored ( we label this phase as a @xmath6 phase ) .the first order phase transition between the @xmath7 stripe and @xmath6 phases is consistent with the observed by neutron diffraction experiments reported in the main text .we consider the minimal three - band model with the hole pocket @xmath97 at the center of the brillouin zone and two electron pockets @xmath13 and @xmath14 at @xmath98 and @xmath99 , respectively . for simplicity , we consider parabolic dispersions with @xmath100 , @xmath101 , and @xmath102 , where @xmath46 denotes the band masses , @xmath47 is the offset energy , and @xmath48 is the chemical potential . near the fermi energy and for small ellipticity , the dispersions can be approximated by @xmath103 , @xmath104 , @xmath105 , with @xmath62 , @xmath64 , and @xmath106 @xcite .electrons with spin @xmath51 of the band @xmath107 are created by the operators @xmath108 , and free - fermion part of the hamiltonian has the form @xmath49 here the summation over repeated spin indices is assumed , and we shift the momenta of the fermions near the @xmath13 and @xmath14 fermi pockets by @xmath98 and @xmath99 , respectively , _i.e. _ , write @xmath109 , @xmath110 .to shorten presentation , we restrict the interacting part of hamiltonian to the interaction in the spin channel with momenta near @xmath98 and @xmath99 , i.e. , to @xmath111 where @xmath112 is the electronic spin operator , and @xmath113 are pauli matrices .the coupling @xmath114 is the sum of density - density and pair - hopping interactions between hole and electron states ( @xmath115 in the notation of ref .@xcite ) , where @xmath116 and the dots stand for the terms with @xmath117 , which only contribute to the cdw channel . the couplings @xmath118 and @xmath119 do depend on the angle along the electron pockets @xcite , but for our purposes this dependence may be neglected , i.e. , @xmath114 can be approximated by a constant .once @xmath114 exceeds some critical value ( which gets larger when @xmath120 and @xmath121 increase ) , the static magnetic susceptibility diverges at @xmath25 and @xmath24 , and the system develops long - range magnetic order . to understand what kind of magnetic order wins below @xmath18we introduce the two spin fields @xmath122 .we apply hubbard - stratonovich transformation , integrate out fermions , obtain the action @xmath123 in terms of @xmath124 and @xmath125 , use saddle - point approximation @xmath126 , and solve a set of coupled saddle - point equations for @xmath124 and @xmath125 a straightforward way to perform this calculation is is to introduce the @xmath127-dimensional nambu operator @xmath128 applying the hubbard - stratonovich transformation and evaluating the products of the pauli matrices , we obtain the partition function in the form @xcite : @xmath129}\label{a_z}\ ] ] with the action @xmath130=-\int_{k}\psi_{k}^{\dagger}\mathcal{g}_{k}^{-1}\psi_{k}^{}+\frac{2}{u_{\mathrm{spin}}}\int_{x}\left(\delta_{x}^{2}+\delta_{y}^{2}\right)\label{a_s}\ ] ] here @xmath131 , and the green s function @xmath132 is given by : @xmath133 where the free - fermion term is @xmath134 and the interaction term is @xmath135 here we introduced the @xmath136 matrices @xmath137 and @xmath138 , where @xmath139 is the identity matrix .the functions @xmath140 are the non - interacting single - particle green s functions for @xmath97 , y , and x fermions .it is now straightforward to integrate out the fermions , since the action is quadratic in them , and obtain the effective magnetic action : @xmath141=-\mathrm{tr}\ln\left(1-\mathcal{g}_{0,k}\mathcal{v}\right)+\frac{2}{u_{\mathrm{spin}}}\int_{x}\left(\delta_{x}^{2}+\delta_{y}^{2}\right)\label{a_aux_effective_s1}\ ] ] here @xmath142 refers to the sum over momentum , frequency and nambu indices .in contrast to the previous studies @xcite , we do not perform a series expansion in powers of @xmath143 but analyze the full non - linear saddle - point ( mean - field ) solutions @xmath144 for the two cases : @xmath145 a @xmath7 stripe phase , in which we set @xmath146 , @xmath147 and @xmath148 a @xmath6 phase in which we set @xmath149 for the case @xmath145 , the mean - field equation has the form @xmath150 while for the case @xmath148 we have @xmath151 in both cases the sum over matsubara frequencies can be evaluated exactly . for stripe magnetic order ,the mean - field equation becomes @xmath152 where @xmath153 . for the @xmath6 phase, we obtain @xmath154 \label{c4}\end{aligned}\ ] ] where the energies @xmath155 ( @xmath50 ) are the three solutions of the cubic equation @xmath156 at a finite mismatch between the electron and hole pockets , @xmath157 ( a ) and as a function of the mismatch @xmath60 for the finite ellipticity @xmath158 ( b ) .the parameter @xmath60 increases with increasing doping .the squares denote the nel temperature for the stripe phase .the triangles denote the temperature below which the @xmath6 phase wins over the stripe phase , although the two phases remain nearly degenerate in energy .the solid and dashed curves are guides to the eye .we used @xmath159ev , @xmath160ev , and @xmath161 . in fig .[ figt1]b , the energies of the @xmath6 and the stripe phases are very close for all @xmath1 , so @xmath6 phase may actually win at low @xmath1 in a more generic model ., title="fig : " ] at a finite mismatch between the electron and hole pockets , @xmath157 ( a ) and as a function of the mismatch @xmath60 for the finite ellipticity @xmath158 ( b ) .the parameter @xmath60 increases with increasing doping .the squares denote the nel temperature for the stripe phase .the triangles denote the temperature below which the @xmath6 phase wins over the stripe phase , although the two phases remain nearly degenerate in energy .the solid and dashed curves are guides to the eye .we used @xmath159ev , @xmath160ev , and @xmath161 . in fig .[ figt1]b , the energies of the @xmath6 and the stripe phases are very close for all @xmath1 , so @xmath6 phase may actually win at low @xmath1 in a more generic model ., title="fig : " ] we solved these equations numerically together with the equation for the chemical potential , for different values of the chemical potential mismatch @xmath60 and ellipticity parameter @xmath61 . like in previous analysis @xcite, we find that the actions for @xmath7 and @xmath6 phases are degenerate at zero ellipticity and for equal sizes of the electron and hole pockets ( @xmath162 ) .once ellipticity becomes non - zero , the @xmath7 wins near the nel temperature . within the ginzburg - landau expansion to order @xmath163, the lower free energy of the @xmath7 phase is the consequence of the fact that the ellipticity generates the term @xmath164 with positive coefficient @xmath165 , which increases the energy of the @xmath6 phase but does not affect @xmath7 phase . going beyond ginzburg - landau approximation , we found that for small enough ellipticity , the solution of eq .( [ c4 ] ) re - emerges below some @xmath2 and , below this @xmath1 , @xmath6 phase becomes a local minimum .furthermore , in some range of @xmath166 , at even lower @xmath167 , the free energy of the @xmath6 phase becomes slightly smaller than that of the @xmath7 phase , i.e. , the system undergoes a first - order transition from @xmath7 to @xmath6 magnetic phase , and lattice @xmath6 symmetry gets restored . at @xmath168the region of @xmath61 where @xmath6 phase wins is exponentially small .however , at a finite @xmath60 , this region widens up .we show the results for a particular @xmath60 in fig . [ figt1](a ) and for a given ellipticity as a function of @xmath60 in fig . [figt1](b ) .a non - zero @xmath60 means that the sizes of electron and hole pockets are non - equal , i.e. , that there is a finite amount of doping .the implication of this result is that , at a finite doping , as the temperature is lowered , the system first orders into a stripe @xmath7 phase in which four - fold lattice rotational symmetry is broken ( @xmath13 and @xmath14 directions become non - equivalent ) , and then , at a lower @xmath1 , it undergoes a first order transition into the @xmath6 phase in which four - fold lattice rotational symmetry is restored .this is consistent with our experiment .high resolution neutron diffraction on hrpd ( isis facility ) shows that the atomic structure of the @xmath6 phase has the same @xmath77 space group as the high - temperature paramagnetic phase , but coexists with the @xmath7 phase with the orthorhombic @xmath79 space group below @xmath87 k ( see fig . 3 of the manuscript ) .since the ordered moments are so small ( @xmath169 to @xmath170 ) , the magnetic peaks are too weak to be observed on hrpd , but are measurable on the wish diffractometer ( isis facility ) , where they are comparable in intensity to bragg peaks due to fe@xmath5as and other impurity phases .this makes a completely unambiguous refinement of the magnetic structure , and even the direction of the magnetic moments , impossible .however , the symmetry imposed by the atomic structure and the constraints imposed by the data allow us to focus on the models discussed below .most simple magnetic models can be ruled out due to their inability to fit the large increase in intensity in the @xmath171 peak that coincides with the @xmath6 phase ( the subscript denotes that it is indexed based on the @xmath172-centered @xmath173 cell ) .only two models fit the data satisfactorily ( fig .[ figs1 ] ) . in the first ( a ) ,the magnetic moment is oriented in the @xmath88 direction , with stripes in either the @xmath76 or @xmath78 direction ( which are equivalent ) . in this casethe magnetic space group is @xmath174 . in the second model ( b ) ,there are no stripes ; instead the moments lie in the @xmath175 plane , spacegroup @xmath176 ( @xmath177 with respect to the @xmath77 cell ) .moments adopt orientations in the @xmath178 directions , with the @xmath6 axis mandating that each moment is orthogonal to its nearest neighbor ( this is the model illustrated in fig .1(c ) in the main manuscript ) . in both a and b models , each layeris coupled antiferromagnetically to its neighbor , which is determined by the intensity of the magnetic @xmath179 reflections .( in the afm case @xmath180 for @xmath181 and in the fm case @xmath180 for @xmath182 .we observe the former . )the fits to the two co - existing phases observed below 39 k , _i.e. _ , the @xmath6 phase ( models a or b ) and the @xmath7 stripe phase , are compared in fig . [ figs2 ] .model a produces a slightly better overall fit than model b. however , the weak signal - to - noise ratio and the overlap with peaks from the coexisting @xmath7 magnetic phase as well as weak impurity phases preclude an unambiguous solution . the magnetic moments reported in the manuscript ( see fig . 4 in the main manuscript ) for the @xmath6 and @xmath7 phases are based on a free refinement of the moments for the @xmath6 model a and the standard @xmath7 model .[ [ relationship - of - magnetic - ordering - to - the - theoretical - electronic - model ] ] relationship of magnetic ordering to the theoretical electronic model ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ the atomic structure of the @xmath6 phase is tetragonal , with no evidence of orthorhombic strain even in the high resolution measurements on hrpd .this implies by symmetry that @xmath183 , which is consistent with model b , whose magnetic space group is also tetragonal .however , the magnetic structure of model a is still apparently orthorhombic even though the moments are along the @xmath88-axis .this can not be reconciled with an orbital order model , in which the stripes are a result of @xmath184 , but it is also not consistent with the itinerant model , in which the @xmath76 and @xmath78 axes are equivalent . however , tetragonal symmetry can be restored if we assume that , instead of separate domains of @xmath76 and @xmath78-axis stripes , the magnetic structure involves a simultaneous modulation in both directions , _i.e. _ , a double-@xmath185 structure .writing the two forms as a spin - density waves with modulations @xmath186 and @xmath187 shows that adding the waves interfere such that half of the moments double and the others sum to zero ( fig .[ figs3 ] : upper panel ) .the nodes on every other fe atom in a checkerboard fashion remove striping and recover tetragonal symmetry . in an itinerant magnetic model, a node implies that there is actually zero moment centered on the nodal atom ( unlike localized moment systems where nodes in a magnetic model imply a fluctuating spin and therefore contribute a significant entropy to the system ) .this is physically allowed but an alternative choice without nodes is to change the phasing between waves , _e.g. _ , @xmath188 , which scales the amplitude of one direction because @xmath189 where @xmath190 .this gives a situation where @xmath191 and @xmath192 .this gives a checkerboard of alternating large and small moments ( fig .[ figs3 ] : lower panel ) .this can be modeled in tetragonal space groups without a proper @xmath6 axis , such as @xmath193 and @xmath194 which allow for two inequivalent coplanar fe atoms .we modeled this arrangement ( model a2 ) in the spacegroup @xmath195 ( # 114 ) .this maintains the striping , but there is an intractable difference between models a and a2 .the @xmath196 axis preserve tetragonal symmetry by requiring that the striping alternates directions between layers , so that there is no preferred direction ( left panel , fig .[ figs4 ] ) .thus , the only situation that is fully afm coupled between layers is when @xmath197 .this makes unobserved reflections @xmath198 non - zero .however , situations where @xmath199 ( right panel , fig .[ figs4 ] ) predict intensities for these peaks that are on the same scale as the noise .otherwise the two models are identical . in short , modela is a more elegant solution to the magnetic data than a2 .however , a final determination of the magnetic structure will have to wait for single crystal measurements .

<S> a theory of superconductivity in the iron - based materials requires an understanding of the phase diagram of the normal state . in these compounds </S><S> , superconductivity emerges when stripe spin density wave ( sdw ) order is suppressed by doping , pressure or disorder , suggesting a possible role for magnetic fluctuations in the pairing mechanism . </S><S> experimentally , sdw order is often pre - empted by nematic order whose phase transition occurs at a slightly higher temperature but whose origin is yet to be resolved . </S><S> one scenario is that the nematic order is unrelated to magnetism and reflects an orbital ordering of the iron 3@xmath0-electrons . </S><S> this ordering triggers the anisotropic magnetic coupling that produces stripe sdw order at a lower @xmath1 . </S><S> another scenario is that stripe - sdw magnetism is the primary instability , but the magnetic transition occurs in two steps : first four - fold spin - rotational symmetry gets broken and the system develops spin - nematic order , and then time - reversal symmetry also gets broken , and full stripe - sdw order develops . because spin - nematic order also induces orbital order , it is difficult to distinguish between the two scenarios experimentally . in this letter , we show that the magnetic scenario predicts that , at high doping , the system should undergo a transition at @xmath2 into a new sdw phase in which four - fold rotational symmetry is restored . </S><S> we have now observed a first - order transition to such a phase in ba@xmath3na@xmath4fe@xmath5as@xmath5 , in good agreement with the spin - nematic model . </S>

0912.5444

in many different physical and mathematical problems ( see e.g. @xcite and references therein ) it is of interest to know the spectral density of sub - unitary ensemble of random matrices of the form @xmath2 where @xmath3 is @xmath1 unitary matrix distributed according to the standard haar measure and @xmath4 is a diagonal matrix of the form @xmath5 where each @xmath6 obeys @xmath7 . the exact formula for the spectral density of these matrices for finite @xmath0 and arbitrary @xmath6 has been obtained in @xcite by a supersymmetry method . in the conclusion of that paperthe authors wrote : `` an interesting problem would be to investigate the density of complex eigenvalues in the limit @xmath8 assuming that @xmath9 has a finite density @xmath10 of eigenvalues @xmath6 in an interval of the @xmath9-axis '' .first of all we would like to stress that such type of asymptotic questions is very natural within the free probability theory @xcite-@xcite .in fact , an explicit solution to this problem has been found in @xcite .the purpose of this note is to demonstrate how this answer can be derived from the exact formulas obtained in @xcite .for general @xmath6 the matrix ( [ main ] ) have complex eigenvalues @xmath11 .the question we address is to calculate the mean density of these eigenvalues in the limit of large matrix dimensions @xmath12 here @xmath13 denoted the average over the haar measure of matrices @xmath3 .assume that limiting positive moments of @xmath6 in ( [ g_i ] ) @xmath14 are finite which is equivalent to the existence of finite limiting measure for @xmath6 . according to the theorem 4.4 of @xcite the limiting measure @xmath15 ( [ nu_t ] ) of ensemble of matrices ( [ main ] )is non - zero in the annulus @xmath16 if @xmath17 diverges the lower bound is set to zero . inside this annulusthe mean number of eigenvalues of matrix ( [ main ] ) in complex plane in polar coordinates @xmath18 is @xmath19 and the integrated spectral density is determined from the equation @xmath20 the function @xmath21 here is calculated as follows .define first the generation function of integer moments @xmath22 by the formula @xmath23 then one has to invert this function , i.e. to find the function @xmath24 such that @xmath25 the knowledge of this function permits to calculate the so - called @xmath26-transform of @xmath9 which is typical in free probability theory @xcite @xmath27 finally the function @xmath21 which enters to the mean density of eigenvalues ( [ final_1 ] ) takes the form @xmath28 in other words , the integrated mean density of eigenvalues , @xmath29 , is the inverse function to @xmath21 i.e. @xmath30 . from the above equations one easily gets that @xmath31 has to be determined from the equation @xmath32 for clarity let us consider the simplest example of truncated random unitary matrix ensemble investigated in @xcite .it corresponds to the case when @xmath33 for @xmath34 with all others @xmath35 and @xmath36 when @xmath8 . in this case @xmath37 and @xmath38 .the function @xmath39 in ( [ psi ] ) is now @xmath40 from which it follows that the solution of ( [ simple ] ) is @xmath41 which corresponds to the radial spectral density equal to @xmath42 this result agrees with eq . ( 19 ) of @xcite up to the normalisation . in that paperthe authors normalized the density to the number of non - zero eigenvalues , but in the above formulas the density is normalized to the total number of eigenvalues which gives an additional factor of @xmath43 .assume that @xmath6 are ordered : @xmath44 .then according to eq .( 2.3 ) of @xcite the exact mean density can be written as follows @xmath45 here @xmath46 is the following function @xmath47 \ , .\label{formula}\ ] ] the density in ( [ sum ] ) is normalized in such a way that @xmath48 for the later use it is convenient to change @xmath49 as @xmath50 .it gives @xmath51 \label{j}\ ] ] where @xmath52 as this function is symmetric functions of @xmath53 , the sum entered to ( [ sum ] ) can be represented as the contour integral .define the function @xmath54 it is clear that @xmath55 f(q)\mathrm{d}q\ ] ] where the contour @xmath56 enclosed poles @xmath57 .changing the variable @xmath58 , one gets the convenient representation of the exact eigenvalue density ( [ sum ] ) @xmath59 \label{main_integral}\ ] ] and the contour @xmath60 encircles the images of the same poles as above but has to avoid the pole at @xmath61 with @xmath62 as indicated at figure [ fig1 ] .notice that the functions in the denominator and in the numerator are the same which reveals the supersymmetry - like structure of these integrals . ) .@xmath63 are poles of the integrand in @xmath64 .the positions of the fixed saddle point @xmath65 obeying ( [ saddle_2 ] ) and of the moving pole @xmath61 are indicated by the arrows . ]the function @xmath66 in ( [ w ] ) can be rewritten in the form @xmath67 $ ] where @xmath68 at large @xmath0 the integral over @xmath69 ( as well the integral over @xmath70 ) can be calculated by the saddle point method .the saddle point @xmath71 has to be determined from the condition @xmath72=0 \label{saddle_2}\ ] ] where we introduce the same function @xmath22 as in ( [ psi ] ) . from this expressionit follows that the saddle point obeys the same equation as ( [ simple ] ) which justify the notation @xmath73 for the saddle point .as the same saddle point appears also in the integration over @xmath70 , it will be important only when it lies between the limits of integration over @xmath70 , i.e. when it obeys the inequalities @xmath74 .it is straightforward to conclude that it will be the case only when the inequalities ( [ inequalities ] ) for @xmath75 are fulfilled . from these arguments it follows that when @xmath76 the dominant contribution to the integrals ( [ main_integral ] ) comes from a vicinity of the above saddle point @xmath77 .the difference between the integrals over @xmath70 and over @xmath69 is that in the former the integration is performed along the real axis but in the later the integration contour should be deformed to pass through the saddle point parallel to the imaginary axis .as the integrand over @xmath69 decreases quickly at large @xmath69 the necessary deformation of the integration contour is legitimate but as there exists a pole at @xmath61 such deformation will result at an additional contribution from this pole ( cf .figure [ fig1 ] ) . from ( [ main_integral ] )it is straightforward to check that this contribution takes the form @xmath78 after deforming the contour , the integration in the saddle point approximation is standard and the final answer is @xmath79 here @xmath80 is the value of the second derivative of the function ( [ phi ] ) calculated at the saddle point @xmath73 obeying ( [ simple ] ) . from ( [ saddle_2 ] ) onegets @xmath81 \label{psi_second}\ ] ] where prime indicates the differentiation over argument of the function .one can check that @xmath80 is negative as it should be for the application of the saddle point method . from ( [ simple ] )it follows that @xmath82=-\frac{y-1}{y|z|^4}\psi^{\prime}_g\left ( \frac{y-1}{y|z|^2}\right ) \ , .\label{y_prime}\ ] ] expressing @xmath83 through @xmath80 from ( [ psi_second ] ) and substituting to ( [ y_prime ] ) one gets @xmath84 taking into account that @xmath85 one finally concludes that the combination in ( [ last ] ) equals @xmath86 , or @xmath87 which exactly corresponds to the free probability result ( [ simple ] ) derived in @xcite .the author is grateful to s. nonnemacher for the discussion which stimulates his interest to this problem .he is also greatly indebted to a. comptet and p. biane for pointing out the reference @xcite .99 yi wei and yan v. fyodorov , _ on the mean density of complex eigenvalues for an ensemble of random matrices with prescribed singular values _ , j. phys .a : math . theor .* 41 * ( 2008 ) 50201 d.v .voiculescu , _ addition of certain non - commuting random variables _ , j. func .analysis * 66 * , 323 ( 1986 ) d.v .voiculescu , _ multiplication of certain non - commuting random variables _ ,. theory , * 18 * , 223 ( 1987 ) d.v .voiculescu , k. j. dykema , and a. rica , _ free random variables _ , amer . math .providence , ri ( 1992 ) u. haagerup and f. larsen , _brown s spectral distribution measure for @xmath88-diagonal elements in finite von neumann algebras _ , journal of functional analysis , * 176 * ( 2000 ) 331 k. zyczkowski and h .- j .sommers , _ truncations of random matrices _ , j. phys .theor . * 33 * ( 2000 ) 2045

<S> the large @xmath0 limit of mean spectral density for the ensemble of @xmath1 sub - unitary matrices derived by wei and fyodorov ( j. phys . </S><S> a : math . theor . </S><S> * 41 * ( 2008 ) 50201 , ref . </S><S> @xcite ) is calculated by a modification of the saddle point method . </S><S> it is shown that the result coincides with the one obtained within the free probability theory by haagerup and larsen ( j. funct . </S><S> anal . * </S><S> 176 * ( 2000 ) 331 , ref . </S><S> @xcite ) . </S>

1611.00149

in the course of the past decades , the role of entanglement has evolved into a genuine quantum resource utilized in various quantum communication and computation protocols @xcite .this evolution has been supported by the formidable progress made on the techniques of generating entanglement in practice .inevitable and inescapable noise , together with imperfections present in every real experiment , may , however , degrade the intended entangled state .being able to detect and measure the entanglement content becomes important since any amount of entanglement can be harnessed in non - classical tasks @xcite .although several theoretical measures have been developed for this purpose @xcite , realizing them in practice remains challenging in general .the reason is that typically these measures of entanglement contain rather involved , even unphysical , operations or are non - linear functions of the state .one of the most widely used measures of entanglement is the so - called concurrence @xcite , which in the case of two qubits in a pure state takes a particularly simple form . despite the mathematical simplicity , the task of quantifying the value of concurrence of an _ unknown _ two - qubit pure state using only a single measurement set - up of a fixed normalized projection - valued measure ( pvm ) is impossible @xcite .nevertheless , several different procedures circumventing this impossibility have been reported that exploit collective measurements done with simultaneous copies of the state @xcite or utilize the curious relation between concurrence and two - particle interference @xcite .furthermore , measurements of concurrence that rely on relaxing the aforementioned pvm - criterion have been developed @xcite . in this study , we propose a local tomographic strategy to quantify the concurrence of _ any _ two - qubit pure state that takes advantage of so - called weak measurements .we also consider an experimental implementation on an optical set - up that can be deployed to measure the concurrence of two polarization entangled photons using the proposed protocol .our method is , however , universal in the sense that it works for all two - qubit systems .the key tools of our proposal are weak measurements and the resulting weak values @xcite .weak measurements are ( von neumann ) standard measurements @xcite where the coupling strength @xmath0 between the measured system and the measuring pointer is minuscule .consequently , the disturbance of the weak measurement to any subsequent ( strong ) measurement , usually called post - selection , is negligible . by post - selecting on a particular pure state @xmath1 , in the vanishing interaction strength limit @xmath2, one can derive the weak value of the observable @xmath3 as @xmath4}}{{\textrm{tr}\left[\rho\ , |\varphi \rangle\langle \varphi|\right]}},\end{aligned}\ ] ] where @xmath5 is the pre - selected ( mixed ) state of the measured system @xcite . throughout this paper , we omit the pre - selection sub - index whenever it is clear from the context .weak values are intrinsically complex which has already proved useful in characterizing the mathematically observable - independent probability space @xcite , several quantum paradoxes @xcite , the quantum state @xcite , and unobservable quantities such as the geometric phase @xcite and the non - hermitian operator @xcite ; see also the review papers @xcite .we show one that may also take advantage of the complex feature of the weak values in assessing the amount of entanglement with a single measurement set - up .this result builds up on the fact first noted in ref.@xcite that , when a laguerre - gaussian beam is used as the pointer state of the weak measurement , certain weak values can be interpreted as stereographical projections of the bloch sphere onto @xmath6-plane .let us assume that two observers , alice and bob , are tasked with determining the amount of entanglement in a bipartite state @xmath7 by means of performing local operations .furthermore , assume that the source generates only pure two - qubit states , that is , @xmath8 for some @xmath9 and @xmath10 are the eigenvectors of pauli operator @xmath11 , and @xmath12 and @xmath13 ( @xmath14 ) are complex numbers satisfying the normalization @xmath15 .one of the most widely used entanglement measures in two - qubit systems is the _ concurrence _ @xmath16 . in the case of a pure state @xmath17 , the concurrence @xmath18 takes the simple form @xcite @xmath19 where @xmath20 is the reduced density matrix of alice ( bob ) , _e.g. , _ @xmath21 the concurrence has a one - to - one connection to the von neumann entropy @xcite @xmath22}=-{\textrm{tr}\left[\rho_b \log_2(\rho_b)\right]}\nonumber \\ & = & -\frac{1+\sqrt{1-c^2}}{2}\log_2\left(\frac{1+\sqrt{1-c^2}}{2}\right ) \nonumber \\ & & - \frac{1-\sqrt{1-c^2}}{2}\log_2\left(\frac{1-\sqrt{1-c^2}}{2}\right),\end{aligned}\ ] ] and via that to a plethora of other entanglement measures @xcite , which makes it a natural choice of figure of merit for our task .our main result is to reveal a mathematical relationship between the concurrence and the weak values corresponding to weak measurements of either one of the local observers .for instance , alice s weak values of the observable @xmath23 , pre - selected on her reduced state @xmath24 and post - selected on either @xmath25 or @xmath10 , read @xmath26}}{{\textrm{tr}\left[\rho_a \ , |0\rangle\langle 0|\right]}}= \frac{a_{00 } a_{10}^ * + a_{01 } a_{11}^ * } { |a_{00}|^2+|a_{01}|^2 } , \nonumber \\ { \,_{{\langle{1}|}}\langle{\sigma_x^a}\rangle_{}^{w } } & : = & \frac{{\textrm{tr}\left [ \sigma_x^a \ , \rho_a \,|1\rangle\langle 1|\right]}}{{\textrm{tr}\left[\rho_a \ , |1\rangle\langle 1|\right ] } } = \frac{a_{00}^ * a_{10 } + a_{01}^ * a_{11 } } { |a_{10}|^2+|a_{11}|^2};\end{aligned}\ ] ] see fig.[fig : weak](a ) . (a ) ( b ) a weak value may not be well - defined , if its denominator vanishes .physically this corresponds to receiving no signal on the measuring pointer whatsoever .we notice that either one of the above weak values being non - vanishing automatically implies that the other one is also non - zero .therefore , whenever @xmath27 , we may write @xmath28 and solve @xmath29 where we have used the information @xmath30 . since @xmath31 , we additionally conclude that @xmath32 on the other hand , one of the weak values in eq.([eq : wvs ] ) being zero implies that the other one either also vanishes or is not well - defined .assume for example that @xmath33 is not well - defined . then @xmath34 implying @xmath35 .similarly , @xmath35 if @xmath36 is not well - defined .these observations can also be reproduced from eq.([eq : concwv ] ) as limiting cases . therefore , except for the point @xmath37 , the concurrence , plotted in fig.[fig : weak](b ) , may be determined from eq.([eq : concwv ] ) .it is noteworthy that the protocol presented works completely locally .we note in passing that the optimal line passing through the origin @xmath38 in fig.[fig : weak](b ) is @xmath39 .namely , except for this point , concurrence @xmath18 is a continuous function of @xmath40 and @xmath41 .concurrence is not even a function of two weak values .namely , with a proper choice of @xmath42 @xmath18 can acquire any value from the interval @xmath43 $ ] , while satisfying @xmath44 . ] using this information , we can verify that @xmath45 is the only line passing through the origin on which @xmath18 attains its maximum value @xmath46 .this optimal line corresponds to the initial state @xmath24 prepared on the equatorial plane of the bloch sphere in fig.[fig : weak](a ) . on this line ,eq.([eq : concwv ] ) simplifies to @xmath47 this observation is useful in order to calibrate @xmath16 as close to unity ( or any other value from the interval @xmath43 $ ] ) as desired .because this process is completely local , the other party ( bob ) can validate the result of this `` optimization '' for instance via state tomography .determining @xmath18 of arbitrary @xmath17 directly via measurement of only a single set of orthogonal projectors @xmath48 , @xmath49 , where @xmath50 ( kronecker delta ) , is impossible @xcite .in other words , one can not quantify concurrence of all bipartite states with a single fixed measurement set - up if the measured observable is a pvm .this is due to the fact that the measured probabilities @xmath51 result in three independent real numbers , which are not in general sufficient to determine @xmath18 , a non - linear function of four complex parameters .in fact , even deciding if an unknown bipartite state is the entanglement or not requires as many resources as state tomography @xcite .the relationship between alice s weak values and the concurrence introduced in the previous section suggests that weak measurements allow one to circumvent this impossibility . to extract the real and imaginary parts of the weak value two complementary pointer observablesare usually used @xcite , that is , two separate measurements have to be set up . remarkably however, it is also possible to quantify both of these components simultaneously by using so - called laguerre - gaussian ( lg ) modes @xcite as the initial pointer state due to the initial correlations @xcite related to these states .as alluded in the previous section , the determination of entanglement fails only in problematic cases where @xmath52 .the vanishing weak values imply that alice s state is simplified to @xmath53 these cases correspond to the states on a line connecting the opposite poles @xmath25 and @xmath10 of the bloch sphere [ see fig.[fig : weak](a ) ] . in our protocolthe set of these states has only minor relevance since mathematically it has null measure ( in the relevant measurable space ) .accordingly , the impossibility of determining the concurrence of states with a single pvm strategy persist even if these problematic states were excluded .nevertheless , in these cases a local measurement of the post - selection probabilities can be used to reveal the amount of entanglement in the state @xmath17 .to this end , alice can measure the relative intensities of the post - selected states to solve the diagonal elements of @xmath24 in eq.([eq : problemstate ] ) .since this measurement may be done jointly with the weak measurement protocol described above , the whole procedure of determining the entanglement content in @xmath17 can be achieved with a single fixed measurement device . moreover, the protocol uses only a single fixed pvm as post - selected measurement ; as discussed above , without the preceding weak interaction such an entanglement - measuring strategy would be impossible .the weak values @xmath33 and @xmath36 , in addition to the intensity measurements described above , give sufficient information to determine the reduced state @xmath24 [ see fig.[fig : weak](a ) ] , in accordance with ref.@xcite .thus the protocol we presented essentially relies on local tomography of the reduced state of alice ( or bob ) .simultaneously , it also generalizes the one - qubit pure state tomography described in ref.@xcite for mixed states .# 1#1| # 1| # 1 # 1 # 1#2| # 1#2#3|| # 1 _ # 1^ in this section we describe a possible optical set - up for determining the concurrence of the polarization entangled state @xmath17 of photon pairs via a weak measurement .our proposed experimental set - up is illustrated in fig.[fig : setup ] .the reduced density matrix @xmath24 weakly interacts with the pointer state via interaction @xmath54 where @xmath55 is the momentum operator along the @xmath56-direction on the cross - sectional plane of the optical beam , @xmath0 is a small interaction strength , and @xmath57 are the eigenprojectors of the pauli operator @xmath58 .the interaction ( [ eq : interaction ] ) can be implemented using a polarization sagnac interferometer and the interaction strength @xmath0 can be changed by tilting the angle of a mirror inside the interferometer [ see the inset in fig .[ fig : setup ] ] . .the initial pointer state is prepared as the laguerre - gaussian ( lg ) mode using a mode converter .the weak interaction between eigenvectors of @xmath58 , @xmath59 , can be implemented using a polarization sagnac interferometer ( pbs : polarization beam splitter , hwp : half waveplate).,scaledwidth=45.0% ] as the initial pointer state , we choose the optical propagation mode with two - dimensional normalized amplitude distribution @xmath60 , which satisfies the paraxial wave equation @xcite .after weak interaction and post - selection onto @xmath61 @xmath62 or @xmath63 , the intensity distribution @xmath64 of the final pointer state becomes @xmath65 assuming the `` weakness '' condition , @xmath66 ) induces a translational shift of the pointer state with an amount proportional to the weak value @xmath67 along the @xmath56 direction @xcite .namely , under the weakness condition , eq . ( [ eq : intensity ] ) can be approximated as @xmath68 where @xmath69 corresponds to the total intensity of the post - selected beams .if the fundamental gaussian beam is used for the pointer state , we can extract only the real part of the weak value from the shift in the beam average position and an alternative measurement set - up with additional optical components is required to obtain the imaginary part of the weak value from the shift in the beam average momentum .a more suitable choice for the pointer state for our purpose is ( the first order ) lg beam @xmath70 $ ] , which is a cylindrically symmetric solution of the paraxial wave equation @xcite .the lg beam can be generated from a gaussian one by using a mode converter , such as a q - plate @xcite or a spatial light modulator @xcite . from eq.([eq : wm - probe ] ) , the averaged value of the position operators @xmath71 and @xmath72 on the cross - sectional plane of the final intensity distribution are calculated as @xmath73 \,,\ , \langle { \mathsf{q}}_y \rangle_f & & = \lambda \ , \text{im}[{\,_{{\langle{\varphi}|}}\langle{\sigma_x}\rangle_{}^{w } } ] .\quad\end{aligned}\ ] ] using a two - dimensional image sensor as a detector the lg pointer state therefore allows us to simultaneously visualize both the real and the imaginary part of the weak values @xmath74 and @xmath75 without additional optical components @xcite . in the case of vanishing weak values , where eq.([eq : concwv] ) can not be used , we can not obtain any information about the entanglement from the averaged shifts of the pointer state .however , measuring the total intensities @xmath76 of the two post - selected beams with @xmath77 enables us to determine the diagonal elements of the state in eq.([eq : problemstate ] ) . because this can be performed jointly with measurements of the @xmath71 and @xmath72 position operators , we are able to determine the concurrence of the quantum state @xmath17 with a single measurement set - up . although the lg mode pointer states allow us to determine the concurrence using a single fixed pvm for post - selection , there are some technical difficulties .the first problem is the mode conversion from the fundamental gaussian mode to the lg mode .the conversion efficiency is limited by the mode converter and also by the mode coupling coefficient between the incident mode of the photon pairs and the lg mode . to increase the mode - coupling coefficient , a single mode optical fiber is typically used for spatial mode cleaning of the photon pair beam . in this case, however , fiber coupling loss becomes a serious problem for photon - pair detection .one practical solution is photon - pair generation via four - wave mixing in the single mode fiber @xcite .another problem is the low detection efficiency of the typical image sensor and , concurrently , the demand for a large ensemble of states needed to extract the weak values . to obtain the high - contrast two - dimensional intensity distribution, we have to generate photon pairs with high intensity using pulsed light or a high - gain imaging sensor , such as a cascade of single photon detectors .we have shown how weak measurements and weak values can be used to quantify the concurrence of any two - qubit pure state .we demonstrated that the proposed protocol can be performed with a single measurement set - up using a local weak interaction and a laguerre - gaussian mode as the pointer state .notably , the protocol uses a single fixed pvm as for the post - selection .in contrast , without the preceding weak interaction , such a measurement of concurrence is impossible @xcite .we also considered a potential experimental realization for quantifying the concurrence of the polarization entangled state of photon pairs .although the proposed implementation has some technical difficulties , such as the detection efficiency , we believe that these problems can be solved in the future .m.t . acknowledges financial support from the university of turku graduate school ( utugs ) and the hospitality of the institute for molecular science ( ims ) , national institutes of natural sciences ( nins ) under the financial support of the ims internship program .y.s . was supported in part by perimeter institute for theoretical physics .research at perimeter institute is supported by the government of canada through industry canada by the providence of ontario through the ministry of economic development & innovation .this work was supported by a grant from the murata science foundation , the matsuo foundation , ims joint study program , nins youth collaborative project , the center for the promotion of integrated sciences ( cpis ) of sokendai , icrr joint research from the university of tokyo , and jsps kakenhi grant numbers 25790068 , 25287101 , and 16k05492 .

<S> since entanglement is not an observable _ per se _ , measuring its value in practise is a difficult task . </S><S> here we propose a protocol for quantifying a particular entanglement measure , namely concurrence , of arbitrary two - qubit pure state via a single fixed measurement set - up by exploiting so - called weak measurements and the associated weak values together with the properties of the laguerre - gaussian modes . </S><S> the virtue of our technique is that it is generally applicable for all two - qubit systems and does not involve simultaneous copies of the entangled state . </S><S> we also propose an explicit optical implementation of the protocol . </S>

0803.2253

let @xmath0 be a partition with at most @xmath1 parts .the _ shifted young diagram of shape @xmath2 _ ( or just _ @xmath2-shifted diagram _ ) _ _ is the set @xmath3 we think of @xmath4 as a collection of boxes with @xmath5 boxes in row @xmath6 , and such that the leftmost box of the @xmath7 row is also in the @xmath7 column .a _ shifted standard young tableau shape @xmath2 _ ( or just _ @xmath2-shifted tableau _ ) is a bijective map @xmath8 which is increasing along rows and down columns , i.e. @xmath9 and @xmath10 ( @xmath11 is the number of boxes in @xmath4 ) .the _ diagonal vector _ of such a tableau @xmath12 is @xmath13 .the following is a shifted standard young tableau for @xmath14 .its diagonal vector is @xmath15 .we are interested in describing the possible diagonal vectors appearing in @xmath2-shifted young tableaux .the problem was solved in the case @xmath16 ( the empty partition ) by a. postnikov , in ( * ? ? ?* section 15 ) .specifically , it was shown that diagonal vectors of the shifted triangular shape @xmath17 are in bijection with lattice points of the @xmath18-dimensional _ associahedron _ @xmath19(to be defined in section 2 ) . moreover ,a simple explicit construction was given for the extreme diagonal vectors , i.e. the ones corresponding to the vertices of @xmath19 .in this article , we aim to generalize postnikov s results to arbitrary shifted shapes .specifically , in section 2 we will prove that diagonal vectors of shifted @xmath2-tableaux are in bijection with lattice points of a certain polytope @xmath20 .this polytope is a minkowski sum of simplices in @xmath21 and its combinatorial structure only depends on the length of the partition @xmath2 .in particular , if the length is @xmath1 , @xmath20 turns out to be combinatorially equivalent to @xmath22 . in section 3we shall give an exlpicit construction of @xmath2-shifted tableaux whose diagonal vectors coorespond to the vertices of @xmath20 .for a non - negative integer vector @xmath23 , let @xmath24 be the number of standard @xmath2-shifted tableaux @xmath12 such that @xmath25 for @xmath26 , where we set @xmath27 .[ thm : main]we have the following identity : @xmath28 @xmath29 where @xmath30 denotes the schur symmetric polynomial associated to @xmath2 .consider a vector @xmath31 .define the polytope @xmath32 thus @xmath33 is the section of the order polytope of shape @xmath4 where the values along the main diagonal are @xmath34 .if @xmath35 , this polytope is known as the _ gelfand - tsetlin polytope _ , which has important connections to finite - dimensional representations of @xmath36 ( see @xcite ) ._ _ our proof strategy is to compare two different formulas for the volume of @xmath33 , one of which is more direct and the other is a summation over standard @xmath2-shifted young tableaux . by (* proposition 12 ) , @xmath37 on the other hand , there is a natural map @xmath38 from @xmath33 ( defined except on a set of measure 0 ) , to the set of standard @xmath2-shifted young tableaux , given as follows : let @xmath39 be a point with distinct coordinates . arrange the @xmath40 s in decreasing order and define the tableau @xmath41 by writing @xmath42 in box @xmath43 if @xmath40 is the @xmath44 element in the above list . by the definition of @xmath33 ,it is clear that @xmath12 is a standard @xmath2-shifted young tableau . given a standard @xmath2-shifted tableau @xmath12 with diagonal vector @xmath45 , it is easy to see that @xmath46 is isomorphic to the set @xmath47 which is a direct product of ( inflated ) simplices @xmath48 therefore , @xmath49 summing over all @xmath12 , we obtain @xmath50 comparing the last formula to ( [ eq : vol1 ] ) , and making the substitutions @xmath51 , we obtain the identity in the theorem .in this section we recall the setup from ( * ? ? ? * section 6 ) .let @xmath52 and let @xmath53 denote the standard basis of @xmath21. for a subset @xmath54 , let @xmath55 , which is an @xmath56-dimensional simplex . a large class of _ generalized permutohedra _ ( cf .* section 6 ) ) is given by subsets of @xmath21 of the form @xmath57 i.e. @xmath58 is the minkowski sum of the simplices @xmath59 scaled by @xmath60 .it s not hard to see that if @xmath61 , whenever @xmath62 , then @xmath58 is the usual permutohedron obtained by taking the convex hull of points @xmath63 such that @xmath64 is a permutation of the numbers @xmath65}=\sum_{i\subseteq[n]}y_{i},\ z_{[n-1]}=\sum_{i\subseteq[n-1]}y_{i},\dots , z_{\{1\}}=y_{\{1\}}.\ ] ] generalized permutohedra have been studied extensively in @xcite .one particular example of a generalized permutohedron , the _ associahedron _ , is defined as @xmath66}$ ] .it is also known as the _ stasheff polytope _ and it first appeared in the work of stasheff ( cf .@xcite . )[ pro : newt2]for any subsets @xmath67 $ ] , and any non - negative integers @xmath68 , the coefficient of @xmath69 in @xmath70 is non - zero if and only if @xmath71 is an integer lattice point of the polytope @xmath72 . it s easy to see that the coefficient of @xmath69 in ( [ eq : newt2 ] ) is non - zero if and only if @xmath73 can be written as a sum of vertices of the simplices @xmath74 . by ( * ? ? ?* proposition 14.12 ) , this happens if and only if @xmath73 is a lattice point of @xmath72 .[ pro : schur1]the coefficient of @xmath69 in @xmath75 is non - zero if and only if @xmath73 is a lattice point of the polytope @xmath76}+\lambda_{2}\delta_{[2,n]}+\cdots+\lambda_{n}\delta_{\{n\}}$ ] .recall that@xmath77 where the sum ranges over all _ semi - standard _ young tableaux @xmath12 of shape @xmath2 and weight @xmath78 , i.e. @xmath79 is the number of @xmath6 s appearing in @xmath12 ( see @xcite ) .let @xmath12 be a ssyt of shape @xmath2 and weight @xmath80. then @xmath81 .indeed , if we consider the boxes containing the numbers @xmath82 in @xmath12 , there can be no more than @xmath6 of them in the same column .hence the number of such boxes is at most the size of the first @xmath6 rows of @xmath2 , which is @xmath83 .it follows that any monomial @xmath69 appearing in @xmath84 also appears in @xmath85 . on the other hand , @xmath85does appear in the right side of ( [ eq : schur1 ] ) as the term corresponding to the tableau @xmath12 with 1 s in the first row , 2 s in the second row , etc .therefore , the coefficient of @xmath69 in @xmath75 is non - zero if and only if it is non - zero in @xmath85 , which by proposition [ pro : newt2 ] , is non - zero if and only if @xmath73 is a lattice point of @xmath76}+\lambda_{2}\delta_{[2,n]}+\cdots+\lambda_{n}\delta_{\{n\}}$ ] .the number of ( distinct ) diagonal vectors of @xmath2-shifted young tableaux is equal to the number of lattice points of the polytope @xmath86}+\lambda_{1}\delta_{[1,n]}+\lambda_{2}\delta_{[2,n]}+\cdots+\lambda_{n}\delta_{\{n\}}.\ ] ] by theorem [ thm : main ] , and propositions [ pro : newt2 ] , [ pro : schur1 ] it follows that @xmath87 if and only if @xmath71 is an integer lattice point of the polytope @xmath88}+\lambda_{1}\delta_{[1,n]}+\lambda_{2}\delta_{[2,n]}+\cdots+\lambda_{n}\delta_{\{n\}}.\ ] ] in particular , if @xmath2 has @xmath1 parts ( i.e. @xmath89 ) , we see that @xmath20 is combinatorially equivalent to @xmath22 .in what follows we describe the vertices @xmath20 by using techniques developed in @xcite . given a generalized permutohedron @xmath90 ,assume that its _ building set _@xmath91\vert\ y_{i}>0\}$ ] satisfies the following conditions : 1 . for any @xmath6 , @xmath100 ( @xmath101 is the set of descendants of @xmath6 in @xmath98 ) .2 . there are no @xmath102 distinct incomparable nodes @xmath103 in @xmath98 such that @xmath104 .3 . @xmath105- root of @xmath106 .[ pro : vertices ] ( * ? ? ? * proposition 7.9 ) vertices of @xmath58 are in bijection with @xmath95-forests .more precisely , the vertex @xmath107 of @xmath58 associated with a @xmath95-forest @xmath98 is given by @xmath108 , for @xmath97 $ ] .the combinatorial structure of @xmath20 clearly depends only on its building set , i.e. the number of non - zero parts of the partition @xmath2 .assume @xmath109 , so that the building set of @xmath20 is @xmath110\vert\ 1\le i\le j\le n-1\}\cup\{[i,\ n]\vert\ 1\le i\le k\}.\ ] ] we first deal with the case @xmath111 .let @xmath12 be a plane binary tree on @xmath1 nodes .for a node @xmath112 of @xmath12 , denote by @xmath113 the left and right branches at @xmath112 .there is a unique way to label the nodes of @xmath12 such that for any node @xmath112 , its label is greater than all labels in @xmath114 and smaller than all labels in @xmath115 .this labelling is called the _ binary search labelling _ of @xmath12 .let @xmath12 be a @xmath116-forest .it s easy to see that @xmath117 has form @xmath118 $ ] if and only if the path from the root to @xmath119 always goes to the right . in this case ,@xmath120 $ ] and @xmath121 is maximal when @xmath119 is the right - most node in @xmath12 , i.e. @xmath122 .it follows that @xmath123\}\subseteq b_{k}\subseteq b_{n}\leftrightarrow|l_{n}|\ge n - k$ ] .this argument together with proposition [ pro : bn forests ] implies [ pro : bk forests]the @xmath124-forests are exactly plane binary trees on @xmath1 nodes with the binary search labeling and such that @xmath125 , i.e. such that the ( left ) subtree of the right - most node in @xmath12 has size at least @xmath126 . by propositions [ pro : vertices ] and[ pro : bk forests ] , the number of vertices of @xmath20 is equal to the number of plane binary trees @xmath12 on @xmath1 nodes such that left subtree @xmath130 of the right - most node in @xmath12 has size at least @xmath126 . if @xmath131 , then there are @xmath132 ways to choose @xmath130 and @xmath133 ways to choose the tree @xmath134 . summing over @xmath135 yields the desired formula . to describe the vertices of @xmath20 ,recall that plane binary trees @xmath12 on @xmath1 nodes are in bijective correspondence with the @xmath136 subdivisions of the shifted young diagram @xmath17 into @xmath1 rectangles .this can be defined inductively as follows : let @xmath6 be the root of @xmath12 ( in the binary search labeling ) . then draw an @xmath137 rectangle .then attach the subdivisions corresponding to the binary trees @xmath138 to the left and , respectively , bottom of the rectangle .for a subdivision @xmath139 of @xmath17 into @xmath1 rectangles , the _ @xmath140 __ rectangle _ is the rectangle containing the @xmath7 diagonal box of @xmath17 . if @xmath12 is the binary tree corresponding to @xmath139 , then the @xmath7 rectangle of @xmath139 has size @xmath137 . in particular, @xmath141 is the length of the ( bottom - right ) vertical strip of the subdivision @xmath139 .[ thm : vertices_of_plambda]vertices of @xmath20 are in bijection with subdivisions of the shifted diagram @xmath17 into @xmath1 rectangles such that the bottom - right vertical strip of the subdivision has at least @xmath143 boxes .specifically , let @xmath139 be such a subdivision .then we can get a subdivision @xmath144 of @xmath145 by merging the rectangles in @xmath139 with the rows of the young diagram of @xmath146 that they border .then the corresponding vertex of @xmath20 is @xmath147 , where @xmath148 is the number of boxes in the @xmath7 region of @xmath144 .the first part of the theorem follows from proposition [ pro : bk forests ] and the discussion preceeding the theorem . to prove the second part , we use proposition [ pro : vertices ] . recall that the building set of @xmath20 is @xmath149\vert\ 1\le i\le j\le n\}\cup\{[i , n]\vert\ 1\le i\le k\}$ ] , and @xmath150\in b_{k}}y_{ij}\delta_{[i , j]}$ ] where @xmath151 if @xmath152 and @xmath153 .let @xmath12 be a @xmath124-forest , i.e. a binary tree on @xmath1 nodes with the binary search labeling such that @xmath125 ( cf .proposition [ pro : bk forests ] . )note that @xmath154 $ ] .now proposition [ pro : vertices ] implies that the correponding vertex @xmath155 of @xmath20 is given by if the @xmath7 rectangle of @xmath139 borders the right edge of @xmath17 ( i.e. @xmath157 ) , then @xmath158 .otherwise , @xmath159 . in each case , @xmath148 is the number boxes in the @xmath7 region of @xmath144 .[ exa : subdiv]let @xmath160 .the figure shows how a subdivision @xmath139 of @xmath17 yields the subdivision @xmath144 of @xmath161 .the corresponding vertex of @xmath20 is given by counting boxes in the regions of @xmath144 : @xmath162 it follows that there is a ( 4,2,1,0)-shifted young tableau @xmath12 whose diagonal vector is @xmath163 .on the other hand , one can directly construct @xmath2-shifted young tableaux with diagonal vector @xmath164 by using the subdivision @xmath144 .indeed , we know what the diagonal vector of the tableau @xmath71 should be .consider again the subdivision @xmath144 of @xmath165 .we can extend the diagram @xmath165 to @xmath4 by first adding a box to the left of each row of @xmath165 , and then , by deleting the last @xmath126 boxes in the @xmath166 column of @xmath165 .now , we start by putting @xmath68 in the diagonal boxes of @xmath4 . the remaining part of @xmath4 is divided into @xmath1 regions by @xmath144 .finaly , for each @xmath167 , put the @xmath168 numbers @xmath169 in the @xmath170 region of @xmath144 in a standard way , i.e. such that entries increase along rows and down columns ( as before , we set @xmath171 . ) in this way we obtain a @xmath2-shifted tableau @xmath12 such that @xmath172 .

<S> we study vectors formed by entries on the diagonal of standard young tableaux of shifted shapes . </S><S> such vectors are in bijection with integer lattice points of certain integral polytopes , which are minkowski sums of simplices . </S><S> we also describe vertices of these polytopes , and construct corresponding shifted young tableaux . </S>

nucl-th0308037

there is a continuous attempt to improve the accuracy and reliability of the calculated single beta ( @xmath2 ) and double beta ( @xmath6 ) decay nuclear matrix elements .their values affect our understanding of astrophysical processes and the fundamental properties of neutrino , in particular neutrino mixing and masses .the complexity of the calculation of the double beta decay matrix elements consists in the fact that it is a second order process , i.e. , in addition to the initial and final nuclear states the knowledge of the complete set of the states of the intermediate nucleus is required . in solving of this problem different models as well as different nuclear structure scenarios were applied @xcite .double beta decay can occur in different modes : the neutrinoless mode ( @xmath7-decay ) requires violation of the lepton number and probes new physics scenarios beyond the standard model ( sm ) of particle physics @xcite .the two - neutrino decay ( @xmath0-decay ) mode is process allowed in the sm , i.e. , the half - life of this process is free of unknown parameters of the particle physics side @xcite .as the half - lifes of the @xmath0-decay have been already measured for about ten nuclei , the values of the associated nuclear matrix element can be extracted directly .it provides a cross - check on the reliability of the matrix element calculations .since the nuclei undergoing double beta decay are open shell nuclei the proton - neutron quasiparticle random phase approximation ( pn - qrpa ) method has been the most employed in the evaluation of the double beta decay matrix elements @xcite .in addition , this approach has succeeded in reproducing the experimentally observed suppression of the @xmath0-decay transitions @xcite . however, a strong sensitivity of the computed matrix elements on an increase of the strength of the particle - particle residual interaction in the @xmath8 channel leads to a problem of fixing this parameter @xcite .thus various refinements of the original pn - qrpa have been advanced . in the last decadethe extensions of the pn - qrpa , which include proton - neutron pairing ( full - qrpa ) @xcite , a partial restoration of the pauli exclusion principle discarded in the pn - qrpa ( renormalized qrpa ) @xcite and higher order rpa corrections @xcite , have received much attention in the context of the nuclear structure calculations for the double beta decay . in spite of the fact that these alterations of the pn - qrpa improved reliability of the evaluated matrix elements , further progress of the nuclear approaches dealing with double beta decay is needed .the deformation degrees of freedom of nuclei undergoing the @xmath0-decay were first considered within nilsson model with pairing @xcite .bogdan , faessler , petrovici and holan calculated the @xmath7-decay in the pairing model including deformation and rotation @xcite .the first qrpa calculation of the @xmath0-decay matrix elements in a deformed nilsson - bcs basis were presented in ref .the authors did not take into account particle - particle interaction of the nuclear hamiltonian and assumed initial and final nuclei to be equally deformed .the effects of nuclear deformation were considered also within the @xmath9 scheme @xcite , which has been found successfull in describing the heavy rotational nuclei .the @xmath9 scheme is a tractable shell model theory for deformed nuclei , which requires a severe truncation of the single particle basis .this approach was used for calculation of the double beta decay half - lifes of different heavy nuclear systems @xcite and the predictions were found to be in good agreement with available experimental data for @xmath10 and @xmath11 @xcite .the effect of the deformation of the nuclear shape on the two - neutrino double beta decay matrix element has been discussed in details within a method developed by raduta , faessler and delion @xcite .the authors used angular momentum projected single particle basis having the energies close to those of nilsson levels .the gamow - teller states were generated with help of the spherical proton - neutron qrpa with a good angular momentum quantum number within the considered basis .the results were presented for the @xmath0-decay of @xmath12 .it was shown that the deformation affects significantly the @xmath0-decay matrix element .the deformation effect on the double gamow - teller matrix element of @xmath13 were investigated in the hartree fock bogoliubov ( hfb ) framework in ref .it was noticed that there is a necessity of an appropriate amount of deformation in the hfb intrinsic state to reproduce the experimental @xmath0-decay half - life .there is an interest to study the effect of deformation on the double beta decay matrix elements within the deformed qrpa @xcite , which allows unified description of the @xmath0-decay in spherical and deformed nuclei .this nuclear structure aproach has been found succeessful in description of the single @xmath2-decay transitions of medium and heavy nuclei a long time ago . in the first applications only the particle - hole terms of the gamow - teller forcewere taken into account @xcite .it was supposed that the particle - particle terms have minor effect on the gamow - teller strength function .however , from the spherical qrpa calculations we know that the particle - particle force plays an important role for describing the @xmath14- and @xmath6-processes @xcite .recently , the importance of the particle - particle interaction has been confirmed also in the deformed qrpa treatment of the gamow - teller strength distributions @xcite . a strong sensitivity of the single @xmath2-decay characteristics to the nuclear shape , rpa ground state correlations and pairing correlations were manifested .the aim of the present paper is to investigate the effect of nuclear deformation on the double gamow - teller matrix element of the @xmath0-decay within the deformed qrpa .we present here as an example numerical results for the @xmath0-decay of @xmath3 , which is a prominent transition due to ongoing and planned @xmath6 experiments for this isotope @xcite .a subject of our interest is also the @xmath15 and @xmath14 distributions in @xmath3 and @xmath16 , respectively .the problem of fixing of the nuclear structure parameters will be addressed .the paper is organized as follows .the basic elements of the deformed qrpa formalism are reviewed in section 2 . in section 3we establish our choice for deformation , pairing and force parameters .then the results for the @xmath0-decay ground state transition @xmath17 are analyzed in context of the gamow - teller strength distributions of @xmath3 and @xmath16 .the final conclusions and remarks are pointed out in section iv .the theoretical approach is based on the deformed proton - neutron quasiparticle random phase approximation with separable proton - neutron residual interaction , which is relevant for the allowed gamow - teller transitions @xcite .this approach allows for an unified description of the @xmath0-decay process in deformed and spherical nuclei .the total nuclear hamiltonian takes the form @xmath18 @xmath19 denotes the hamiltonian for the quasiparticle mean field described by a deformed axially - symmetric woods - saxon potential @xcite @xmath20 where @xmath21 are the quasiparticle energies .@xmath22 ( @xmath22 ) is the quasiparticle creation ( annihilation ) operator .the p ( n ) index denotes proton ( neutron ) quasiparticle states with projection @xmath23 ( @xmath24 ) of the full angular momentum on the nuclear symmetry axis and parity @xmath25 ( @xmath26 ) .the index @xmath27 ( @xmath28 ) represents the sign of the angular momentum projection @xmath29 .we note that intrinsic states are twofold degenerate . the states with @xmath30 and@xmath31 have the same energy as consequence of the time reversal invariance .we shall use notation such that @xmath30 is taken to be positive for states and negative for time reversed states .the method includes pairing between like nucleons in the bcs approximation with fixed gap parameters for protons , @xmath32 , and neutrons , @xmath33 .@xmath34 where @xmath35 are single particle energies , @xmath36 is the gap , @xmath37 is the fermi energy .the bogolyubov transformation , which defines the quasiparticle representation , is given by @xmath38 here , @xmath39 and @xmath40 are creation and annihilation operators for particles , respectively .@xmath41 indicates the single nucleon state obtained by the time reversal operator @xmath42 .this state is given in the appendix a @xcite .the occupation probabilities @xmath43 are written as @xmath44 and the proton and neutron number equations are @xmath45 where z and n are the numbers of protons and neutrons , respectively .the residual interaction part @xmath46 of nuclear hamiltonian in eq .( [ eq:1 ] ) contains two terms associated with particle hole ( ph ) and particle particle ( pp ) interaction : @xmath47 the operators @xmath15 and @xmath48 are ph and pp components of the spin - isospin @xmath49 , namely , @xmath50 the ph and pp forces in eq .( [ eq:6 ] ) are defined to be repulsive and attractive ( @xmath51 ) , respectively , reflecting the general feature of the nucleon - nucleon interaction in the @xmath52 channel .the explicit form of the matrix element @xmath53 is presented in the appendix a. the @xmath15 and @xmath14 transition amplitudes from @xmath94 even - even initial nuclear state to a one - phonon state in odd - odd final nucleus are expressed by @xmath95 , \nonumber \\< 1m(k),m| { \hat{\beta}}^+_m |0^+_{g.s . }> & = & \sum_i [ \bar{\sigma}(i ) x^m_{i k } + \sigma ( i ) y^m_{i k } ] .\label{eq:21}\end{aligned}\ ] ] here , @xmath96 denotes the correlated rpa ground state in the laboratory frame .the @xmath97 transition operators in eq .( [ eq:21 ] ) are related with intrinsic @xmath98 operators in eq .( [ eq:7 ] ) as follows : @xmath99 for the @xmath2 strength function we have @xmath100 from the @xmath101 amplitudes one obtains , straightforwardly , the total @xmath101 strengths : @xmath102 and for the ikeda sum rule we get @xmath103 \nonumber \\ & = & \sum_{k=0,\pm 1 } \sum_{p \rho_p n \rho_n } \label{eq:25}\end{aligned}\ ] ] the factor 3 comes from the sum over @xmath104 , i.e. , the contribution from each component k is equal to ( n - z ) . in deriving the above expression we used the closure condition for qrpa states ,assumed the completeness relation for single particle states @xmath105 and applied eq .( [ eq:6 ] ) .we note that if a truncated single particle basis is considered in respect to that the full single particle basis of the woods - saxon potential , the condition in eq .( [ eq:26 ] ) is violated and as a consequence the ikeda sum rule as well .the inverse half - life of the @xmath0-decay can be expressed as a product of an accurately known phase - space factor @xmath106 and the gamow - teller transition matrix element @xmath107 in second order : @xmath108^{-1 } = g^{2\nu } ~(g_a)^4~ | m^{2\nu}_{gt}|^2 . \label{eq:27}\ ] ] the contribution from the two successive fermi transitions is safely neglected as they come from isospin mixing effect @xcite. within the deformed qrpa approach the double gamow - teller matrix element @xmath107 for ground state to ground state @xmath0-decay transition takes the form @xmath109 the sum extends over all @xmath8 states of the intermediate nucleus .the index i( f ) indicates that the quasiparticles and the excited states of the nucleus are defined with respect to the initial ( final ) nuclear ground state @xmath110 ( @xmath111 ) .the overlap is necessary since these intermediate states are not orthogonal to each other .the two sets of intermediate nuclear states generated from the initial and final ground states are not identical within the considered approximation scheme .therefore the overlap factor of these states @xmath112 is introduced in the theory @xcite .its takes the form @xmath113~{\cal r}_{l_f l_i } ~<bcs_f|bcs_i>. \label{eq:29}\ ] ] the factor @xmath114 , which includes the overlaps of single particle wave functions of the initial and final nuclei is given explicitely in the appendix b. there , a detailed derivation of @xmath115 together with the overlap factor of the initial and final bcs vacua can be find as well . in the spherical limit the value of the bcs overlap factor is about 0.8 and it is worth mentioning that it was comonnly neglected in the double beta decay calculations @xcite .we note that we neglected the overlap matrix elements between the intermediate states generated from initial and final nuclei with different k as the corresponding factors are very small due to different structure of corresponding rpa configurations .thus the spin projection flip ( @xmath71 ) and non spin projection flip ( @xmath72 ) excitations contribute coherently to the @xmath0-decay matrix element @xmath107 .the formalism described in section ii is used for the calculation of the @xmath0-decay ground state transition @xmath1 .the results are obtained with a deformed , axially symmetric woods - saxon potential @xcite .the deformation independent woods - saxon parameters ( well depth , skin thickness , radius and spin orbit constants ) are taken from @xcite .this parameterization of the woods - saxon potential was used previously in different rpa calculations , where a good agreement was achieved with experimental data , in particular for single m1 transitions at low energy , observed in @xmath116 and @xmath117 experiments @xcite .the quadrupole - hexadecapole parameterization of the nuclear surface is considered : @xmath118 , \label{eq:31}\ ] ] where @xmath119 is the normalizing constant ( @xmath120 ) and @xmath121 ( @xmath122 ) are spherical harmonics .the spherical limit is achieved for @xmath123 .the deformation parameter @xmath124 can be deduced from nuclear electric quadrupole laboratory moment ( @xmath125 , @xmath126 and @xmath127 are laboratory and intrinsic quadrupole moments , respectively ) or extracted from values based on measured e2 probability ( @xmath128 , the sign can not be extracted ) via the intrinsic quadrupole moments @xmath127 : @xmath129 ( @xmath130 is the charge root mean square radius ) .however , the available experimental data associated with the above two approaches @xcite lead to a different quadrupole deformation both for @xmath3 and @xmath16 . in particular , the quadrupole moments measured by coulomb excitation reorientation method @xcite , which yields also the sign of the intrinsic quadrupole moment , imply @xmath131 ( @xmath3 ) and @xmath132 ( @xmath16 ) @xcite , but from the measured values of @xmath133 strengths @xcite one finds @xmath134 ( @xmath3 ) and @xmath135 ( @xmath16 ) @xcite .a lack of accurate experimental information on the deformation of @xmath3 and @xmath16 suggests that it is necessary to study the associated @xmath0-decay matrix element as a function of deformation parameters of both initial and final nuclei .we note that there are theoretical calculations of deformation parameters performed within relativistic mean field theory @xcite , which imply @xmath136 these results are in good agreement with predictions of the macroscopic - microscopic model of mller , nix , myers and swiatecki @xcite as well as with nonrelativistic hartree fock calculations @xcite . in the deformed qrpa calculationwe assume a truncated model space by considering only single particle states with maximal allowed value of the asymptotic quantum number @xmath137 equal to 5 .in contrast to many other microscopic calculations , energies of single particle levels are not shifted but taken exactly as provided by the deformed woods - saxon potential .the bcs equations are solved for protons and neutrons .the proton and neutron pairing gaps are determined phenomenologically to reproduce the odd - even mass differences through a symmetric five - term formula @xcite . for both spherical and deformed shapes of the studied a=76 nuclei we consider the same values of the gap parameters , which are given by @xmath138 in fig .[ fig.1 ] the calculated proton and neutron occupation probabilities close to the fermi level are presented for @xmath16 .we see that both for protons and neutrons the bcs solutions associated with spherical and deformed nuclear shapes differ significantly each from other .this deformation dependence of the bcs results is expected to have strong impact on the qrpa solution .the calculation of the qrpa energies and wave functions requires the knowledge of the particle - hole @xmath139 and particle particle @xmath140 strengths of the residual interaction .the optimal value of @xmath139 is determined by reproducing the systematics of the empirical position of the gamow - teller giant resonance in odd - odd intermediate nucleus as obtained from the @xmath141 reactions @xcite .the parameter @xmath140 can be determined by exploiting the systematics of single @xmath2-decay feeding the initial and final nuclei . in ref .@xcite the strengths of the particle hole and particle particle terms of the separable gamow - teller force were fixed as smooth functions of mass number a by reproducing the @xmath2-decay properties of nuclei up to a=150 within the spherical pn - qrpa model .the recommended values of @xmath139 and @xmath140 of ref .@xcite are as follows : @xmath142 in the presented @xmath0-decay study of the ground state transition @xmath143 we use the recommended value for @xmath139 ( @xmath144 ) and @xmath140 is considered as free variable .the qrpa calculations for the @xmath145 states are performed by following the procedure described in section ii . the number of zeros of the non - linear rpa secular equation in eq .( [ eq:16 ] ) is equal to the number of the two - quasiparticle configurations included in the microscopic sums ( [ eq:17 ] ) .in the case of @xmath72 ( @xmath71 ) the dimension of the configuration space is of the order of 450 ( 900 ) quasiparticle pairs excitations .to each zero of the rpa secular equation corresponds one rpa energy @xmath146 .it is worthwhile to notice that by introducing the deformation degrees of freedom the zeros are not found in each subinterval of energy @xmath147 determined by the two subsequent two - quasiparticle excitation energies and that in some subintervals two or more zeros are found .this situation is illustrated in fig .[ fig.2 ] .the vertical axis is the left - hand side ( l.f.s . ) of eq.([eq:16 ] ) and the horizontal one is the energy @xmath148 in the range @xmath149 .the upper ( a ) and lower ( b ) subfigures show results for a=76 obtained within the spherical and deformed qrpa , respectively .we note that a similar phenomenon was found also in ref .@xcite in different context .the main drawback of the qrpa is the overestimation of the ground state correlations leading to a collapse of the qrpa ground state , which may lead to ambiguous determination of the @xmath2 and @xmath6-decay matrix elements .the qrpa collapses , because it generates too many correlations with increasing strength of the attractive proton - neutron interaction as a result of the quasiboson approximation used .an interesting issue is the dependence of the position of the collapse of the qrpa solution on the deformation degrees of freedom . in fig .[ fig.3 ] the energies of the lowest @xmath72 and @xmath150 states in @xmath151 calculated within the deformed qrpa in respect to the @xmath16 ground state are plotted as a function of particle - particle interaction strength @xmath140 .the calculations were performed for different values of quadrupole deformation parameter @xmath124 , namely @xmath152 , @xmath153 and @xmath154 representing oblate , spherical and prolate shapes of @xmath16 , respectively .we note that in the spherical limit the energies of @xmath155 and @xmath150 states of the intermediate nucleus coincide to each other .[ fig.3 ] shows that for a deformed nucleus the collapse of the qrpa solution appears for a slightly smaller value of @xmath140 . in the case of strongly deformed @xmath16 this effectis more pronounced in comparison with the less deformed one .thus the deformation degrees of freedom do not improve the stability of the qrpa solution .the violation of the pauli exclusion principle affects the qrpa results strongly in the range of particle - particle interaction @xmath140 above @xmath156 mev .the half - life of the @xmath0-decay of @xmath3 is known with high accuracy from the heidelberg - moscow experiment , in particular @xmath157 \times10^{21}$ ] years @xcite .all existing positive results on the @xmath0-decay were analyzed by barabash @xcite , who suggested to consider for the ground state transition @xmath1 the average value @xmath158 years . by using of eq .( [ eq:27 ] ) and the knowledge of the kinematical factor @xmath106 [ @xmath159 from the @xmath0-decay half - life of @xmath3 one can deduce the absolute value of the nuclear matrix element @xmath160 equal to @xmath161 by assuming @xmath162 . if the value of the axial coupling constant @xmath163 is considered to be unity , the value of @xmath160 deduced from the average half - life for a=76 is larger , @xmath164 . in fig .[ fig.4 ] the effect of the deformation on the @xmath0-decay matrix element @xmath165 is analyzed .the results for the @xmath0-decay matrix element are displayed as function of the particle particle strength @xmath140 .the curves drawn in subfigures a ) , b ) and c ) [ subfigures d ) , e ) and f ) ] correspond to the case both initial and final nuclei are oblate [ prolate ] .the two horizontal lines represent @xmath166 and @xmath167 . by glancing fig .[ fig.4 ] we note that within the whole range of @xmath140 there is only a minimal difference between values of @xmath165 corresponding to the same value of @xmath168 , which is defined as @xmath169 in addition , one finds that by increasing the value of @xmath168 the suppression of @xmath165 becomes stronger within the range @xmath170 .this a new suppression mechanism of the @xmath0-decay matrix element namely , @xmath165 depends strongly on difference in deformations of parent and daugter nuclei .one might ask what is the origin of this suppression . in fig .[ fig.5 ] this point is clarified by presenting the overlap factor of two bcs vacua for different values of @xmath171 as function of the quadrupole deformation of @xmath16 .we see that for a given @xmath171 the curve has a maximum for @xmath172 and with increasing difference in deformations of initial and final nuclei , i.e. , @xmath168 , the value of the bcs overlap factor decreases rapidly . from the fig .[ fig.5 ] it follows that oblate - prolate ( or prolate - oblate ) @xmath0-decay transitions are disfavoured in comparison with prolate - prolate or oblate - oblate ones for the same absolute values of associated quadrupole deformation parameters . in particular , by assuming the @xmath124 and @xmath173 parameters of ref .@xcite we obtain @xmath174 .this suppression is too strong to achieve agreement with experimental data for the calculated @xmath165 .however , by changing the deformation of @xmath16 from oblate to prolate ( @xmath175 ) we end up with significantly larger value of the bcs overlap factor : @xmath176 .the presented suppression mechanism is expecting to work also for other @xmath0-decay transitions . in fig .[ fig.6 ] we present bcs overlap factor for the @xmath0-decay of @xmath3 , @xmath13 , @xmath177 and @xmath178 .the quadrupole deformation of the initial nucleus was taken to be @xmath179 and @xmath124 related with the final nucleus was considered to be a free parameter within the range @xmath180 .the pairing gaps , which entered the bcs equations , are given in eq .( [ eq:33 ] ) for a=76 and for a=100 , 130 and 136 are given by @xmath181 we see that the behaviour of the overlap factor for all considered nuclear systems is qualitatively the same .the maximum of this factor appears for equal quadrupole deformation of the initial and final nuclei . with increasing value of @xmath182the value of the overlap factor strongly decreases .this fact implies that the deformation of nuclei plays an important role in the calculation of the @xmath0-decay transitions and should be known with high reliability .an interesting issue is whether it is possible to obtain information about the deformations of @xmath3 and @xmath16 by studying the @xmath15 and @xmath14 strength distributions in these nuclei , respectively . for different values of quadrupole parameter @xmath124they are presented in figs .[ fig.7 ] and [ fig.8 ] . in order to facilitate comparison among various calculationsthe gamow - teller distributions are smoothed with a gaussian of width @xmath183 .the assumed strengths of the particle - hole and particle - particle interaction strength were @xmath144 and @xmath184 @xcite . by glancing at fig .[ fig.7 ] we see that the position of the gamow - teller resonance is not sensitive to effect of deformation and that only a little strength is concentrated above the region of the gamow - teller resonance . however , there are significant differences between the spherical and deformed strength distributions in general @xcite especially in the case of prolate shape of @xmath3 ( see fig .[ fig.7]a ) , which is favoured by experimental data @xcite . the thick solid line in fig .[ fig.7 ] corresponds to data obtained from charge exchange @xmath185 reaction at @xmath186 @xcite , which were folded with 1 mev width gaussians as it has been done for the theoretical results .we see that the presented calculations are in a very good agreement with data for the position of gamow - teller resonance .it confirms the correct choice of the particle hole strength of the nuclear hamiltonian ( @xmath144 ) in our calculation @xcite . from fig .[ fig.7 ] it also follows that the strength in the peak of the experimental @xmath187 distribution is better reproduced by the qrpa calculations with only sligtly deformed mean field . the calculated @xmath14 strength distributions in @xmath16are displayed in fig .[ fig.8 ] .they have been folded with @xmath188 width gaussians .we see that the effect of deformation on the @xmath14 distribution strength for gamow - teller transition operator is more apparent as it is in the case @xmath15 strength distribution in @xmath3 . in fig .[ fig.8 ] we present also the experimental data , which are represented by solid points . they were obtained from charge exchange @xmath189 reaction @xcite .we note that the experimentally extracted strength distribution with the help of two plausible methods differ from each other by more than a factor of 2 @xcite .hence the experimental data must be considered to be of a qualitative nature only .there are two suppression mechanisms of the @xmath0-decay matrix element .already long ago vogel and zirnbauer showed that @xmath107 is strongly suppressed when a reasonable amount of particle - particle interaction is taken into account @xcite , what is achieved close to a collapse of the qrpa solution . in this paperwe show now that the @xmath0-decay matrix element @xmath107 depends strongly on the difference in deformations of the initial and final nuclei . a possible criterium to decide, which deformation for the initial and the final nucleus in the double beta decay should be used could be the requirement for a common description of both single @xmath2 and @xmath0-decay within the same nuclear hamiltonian . in order to clarify this point three different casesare considered : + i ) spherical shapes of both initial and final nuclei , i.e. , @xmath190 and @xmath191 .+ ii ) deformations of parent and daughter nuclei : @xmath192 and @xmath193 .+ iii ) quadrupole deformations of the ground state of @xmath3 and @xmath16 : @xmath192 and @xmath194 .+ the choice of the deformation parameters in cases i ) and ii ) do not contradict the data @xcite . in fig .[ fig.9 ] the @xmath0-decay of @xmath3 is presented as function of particle - particle interaction strength @xmath140 for the above three cases . in the calculationdetermined by casei ) ( spherical nuclei ) @xmath195 equal to @xmath161 ( @xmath162 ) is reproduced for @xmath196 .the corresponding set of nuclear structure input parameters in this calculation is denoted by capital letter a ( see table [ table.1 ] ) .this value of @xmath140 is significantly larger as the average value @xmath197 deduced from the systematic study of the single @xmath14-decays @xcite .one can hardly suppose that it is because @xmath3 and @xmath16 are specific nuclear systems .they are no indications in favor of it .it seems to be a general problem within the spherical qrpa : the calculated nuclear matrix elements reproduce the experimental @xmath0-decay half - lifes as a rule for particle - particle strength close to its critical value given by the qrpa collapse . in the casethe calculation determined by case ii ) the situation is different .@xmath195 is reproduced for considerable smaller value of particle - particle strength @xmath184 , which corresponds to that of homa et al . @xcite . the letter b donotes the coresponding set of nuclear parameters , which is listed in table [ table.1 ] .if we consider the effective axial coupling constant @xmath198 , used often in the nuclei , @xmath195 is larger , namely @xmath164 .then , a smaller difference between deformations of nuclei entering @xmath0-decay process is needed to reproduce this value for @xmath199 .this is achieved with calculation of @xmath200 performed within the case iii ) for @xmath184 ( nuclear parameter set c ) .nevertheless , one should keep in mind that the coupling strength of @xcite has been adjusted by using of a deformed nilsson single - particle model and @xmath201 .thus it is not necessarily the best possible choice .it is supposed that this prediction for @xmath140 is more accurate for nuclei with shorter @xmath2-decay half - lifes .we note that that the curves corresponding to cases ii ) and iii ) do change only weakly in the range @xmath202 .it means that the role of the particle - particle interaction within this interval is negligible .it is worthwhile to notice that for a large value of @xmath140 of 0.07 mev to 0.075 the agreement with @xmath203 deduced from the @xmath0-decay half - life is achieved as well and that in this case the sign of @xmath165 is negative .however , for negative value of @xmath165 the correspondance with @xmath204 from systematic studies of the single beta decay @xcite is not achieved .thus negative values of @xmath165 are disfavored due to a complete disagreement with single @xmath2-decay study of homma et al .@xcite . in what followswe shall discuss the single @xmath2 decay characteristics corresponding to three different results in fig .[ fig.4 ] , denoted by letters a , b and c , for which the experimental half - life of the @xmath0-decay of @xmath3 is reproduced .the corresponding mean field and residual interaction parameters are given in table [ table.1 ] .the @xmath15 and @xmath14 amplitudes calculated from the ground state of @xmath3 and @xmath16 , respectively , associated with the @xmath72 ( @xmath71 ) intermediate states of @xmath151 are drawn in fig .[ fig.10 ] ( fig . [ fig.11 ] ) . by glancing these figureswe see that deformation ( b and c sets of parameters ) contribute to a fragmentation of the @xmath2-decay amplitudes and that values of @xmath14 amplitudes associated with higher lying states of the intermediate nucleus ( up to 20 mev ) are significant .it is worthwhile to note that the absolute values of @xmath72 and @xmath71 amplitudes are equal each to other in case of the qrpa calculation with spherical mean field .the phases of these amplitudes are not of physical meaning .however , the relative phases of corresponding @xmath14 and @xmath15 amplitudes plays a crucial role due to partial destructive character of the contributions from different intermediate states to @xmath107 . in the calculationthese phases are fixed in such way that diagonal elements of the overlap matrix in eq .( [ eq:29 ] ) are taken to be positive .we note that the @xmath72 and @xmath71 contributions to @xmath107 differ only slightly . it is worthwhile to notice that a lot of @xmath14 strength in @xmath16 is concentrated in the high energy region above @xmath205 ( see fig . [ fig.8 ] ) , which can not be described within the shell model .there is a longstanding question whether in the calculation of the @xmath0-decay matrix element @xmath107 the transitions to higher lying states of the intermediate nucleus play an important role .there is so called `` single state dominance hypothesis '' ( ssdh ) , which assumes that only the lowest @xmath8 state of the intermediate nucleus is of major importance in the evaluation of @xmath107. the ssdh can be realized , e.g. , through cancellation among the higher lying @xmath8 states of the intermediate nucleus . in ref .@xcite a possibility to decide experimentally about validity of the ssdh by measuring the single electron spectra and the angular distribution of the emitted electrons was discussed .this treatment of the ssdh is suitable for the @xmath0-decay transitions with low - lying @xmath8 ground state of the intermediate nucleus .but this is not the case of the @xmath0-decay of @xmath3 . in fig .[ fig.12 ] the qrpa model dependent study of this problem is presented by drawing the running sum of matrix element @xmath206 as a function of the @xmath8 excitation energy @xmath207 of the intermediate nucleus .we see that in the case of the spherical calculation the main contribution to @xmath107 comes from the states of the intermediate nucleus lying below 5 mev scale and that there is a partial cancellation among contributions to @xmath107 from higher lying states . in the case of deformed qrpa calculationsthere is a different situation .we see that important contributions to @xmath206 arise even from relatively high energies of the intermediate states around the position of gamow - teller resonance ( about 12 mev ) , which can not be described for a given nuclear system in the framework of the shell model .these results contradict the ssdh and indicate that a strong truncation in consideration of the complete set of the states of the intermediate nucleus is not appropriate .for the @xmath0-decay ground state transition @xmath1 a detailed study of nuclear matrix elements within the deformed qrpa with separable spin isospin interactions in the particle hole and particle particle channels was performed .a new mechanism of the suppression of the double beta decay nuclear matrix elements was found .we showed that the effect of deformation on the @xmath0-decay matrix element is large for a significant difference in deformations of the parent and daughter nuclei and is not related with the increasing amount of the ground state correlations close to a collapse of the qrpa solution .the origin of this new mechanism is a strong sensitivity of the overlap of the initial and final bcs vacua to the deformations of the initial and final nuclei .this enters directly into the the overlap of the intermediate nuclear states generated from the initial and final nuclei via qrpa diagonalization . a study of other double beta decay transitions as a function of this deformation dependent overlap indicates the general importance of the new suppression mechanism for the @xmath0-decay and @xmath7-decay matrix elements .next , we showed that by assuming spherical nuclei there is a strong disagreement between the particle - particle strength @xmath140 needed to reproduce the @xmath0-decay half - life of @xmath3 and the @xmath140 deduced in ref .@xcite from a systematic study of the single @xmath2-decay transitions . by the new suppression mechanism of the @xmath0-decay matrix element we showed that for deformations of the parent and daughter nuclei , which do not contradict the experimental data , a simultaneous description of both single @xmath2 and @xmath0-decay is possible .we also discussed the role of the axial coupling constant @xmath163 .we found that a significant part of the total @xmath14 strength is concentrated in the energy region above @xmath208 .a detail study of contributions to @xmath107 from different @xmath8 intermediate nuclear states showed that for the a=76 nuclei it is necessary to consider all @xmath8 states of @xmath151 up to energy @xmath209 mev .this fact disfavors theoretical studies of the @xmath0-decay of medium and heavy nuclei in models with a strongly truncated basis , like the shell model .the gamow - teller @xmath14 strengths and double beta decay matrix elements in medium and heavy nuclei continue to be a challenge for nuclear structure models .further theoretical studies are needed to clarify the role of deformation in calculation of other double beta decay transitions , especially those including heavier nuclear systems , which are known to be strongly deformed .future works should investigate the @xmath0-decay matrix elements also within different possible extensions of the deformed qrpa , e.g. , those including proton - neutron pairing @xcite . in this wayone expects to develop a reliable many - body approach with well defined nuclear structure parameters for calculation of the @xmath7-decay matrix elements .their accurate values are highly required to determine the neutrino mixing pattern and to answer the question @xcite about the dominant mechanism of the @xmath7-decay @xcite .the single - particle states are calculated by solving the schrdinger equation with the deformed axially symmetric woods - saxon potential , which parameterization is given in ref .they are characterized by their energy @xmath35 , parity @xmath210 and by the projection @xmath30 ( @xmath211 ) of the full angular momentum on the nuclear symmetry axis .we use notation @xmath212 and @xmath213 for protons and neutrons , respectively .@xmath214 represents proton ( @xmath215 ) or neutron ( @xmath216 ) state with quantum numbers @xmath30 and @xmath210 .@xmath217 is the sign of the angular momentum projection @xmath29 ( @xmath218 ) .we note that intrinsic states are twofold degenerate . the states with @xmath30 and@xmath31 have the same energy as consequence of the time reversal invariance .we shall use notation such that @xmath217 is taken to be positive for states and negative for time reversal states . in order to solve the schrdinger equation the eigenfunctions of a deformed symmetric harmonic oscillatorare used as a basis for the diagonalization of the mean - field hamiltonian @xcite .these states are completely determined by a principal set of quantum numbers @xmath219 , where @xmath220 , @xmath221 and @xmath222 .@xmath223 and @xmath224 are number of nodes of the basis functions in the @xmath225-direction and @xmath226-direction , respectively . @xmath227 and @xmath228 are the projections of the orbital and spin angular momentum on the symmetry axis z. the explicit form of single - particle harmonic oscillator wave functions in cylindrical coordinates @xmath229 can be found , e.g. , in ref .for a given shape of the nuclear surface , the shape of the deformed harmonic oscillator is automatically chosen in a way suitable for obtaining good accuracy with a smallest number of basis functions @xcite .the deformation dependent cutt - off is chosen in such way as to assure numerical stability of the results . in our calculationwe use 11 major shells . in cylindricalcoordinates the eigenfunctions of states and time - reversed states in deformed woods - saxon potential are expressed as follows : @xmath230 .@xmath41 indicates time reversal states .the overlap of the proton and neutron single particle states of the initial @xmath233 and final @xmath234 nuclei are calculated by assuming that the corresponding sets of basis wave functions do not differ significantly each from other .then we have @xmath235 , \nonumber \\ <n_f \rho_{n_f}|n_i \rho_{n_i } > & = & \delta_{\omega_{n_f } \omega_{n_i } } \sum_{n n_z } [ b^{(+)}_{n n_z \omega_{n_f } } b^{(+)}_{n n_z \omega_{n_i}}- b^{(-)}_{n n_z \omega_{n_f } } b^{(-)}_{n n_z \omega_{n_i}}].\end{aligned}\ ] ] the index i ( f ) denotes that proton and neutron single particle states are defined with respect to the initial ( final ) nucleus .as a consequence of considered many - body approximations the two sets of intermediate nuclear states generated from the initial and final ground states are not identical within the qrpa theory .thus it is necessary to introduce the overlap factor of these states , which can be expressed with the help of intrinsic phonon operators as follows : @xmath236 here , the index i ( f ) indicates that the excited states of the nucleus are defined with respect to the initial ( final ) nuclear ground state @xmath237 ( @xmath238 ) . in order to evaluate @xmath112we express the phonon creation operator @xmath239 in terms of creation and anninhilation phonon operators associated with the final nucleus .we have @xmath240 the coefficients of expansion @xmath241 and @xmath242 will be determined below . by inserting eq .( [ eq : b2 ] ) into eq .( [ eq : b1 ] ) we get @xmath243 \nonumber \\ & \approx & a_{m_i { m'}_f } < bcs_f|bcs_i>. \label{eq : b3}\end{aligned}\ ] ] here , we neglected overlap matrix element between the final ground state and the two - phonon state generated from the initial nucleus , which is considered to be small and should be not related with the studied quantity . in addition , we approximated the overlap of initial and final rpa ground state with the bcs ones .thus the overlap factor of the intermediate nuclear states generated from the initial and final ground states is proportional to the overlap of initial and final bcs vacua .we proceed with the calculation of @xmath244 and @xmath245 coefficients .the quasiparticle creation and anninhilation operators @xmath246 [ @xmath247 of the initial [ final ] nucleus are connected with the particle creation and annihilation operators @xmath248 [ @xmath249 by the bcs transformation [ see eq .( [ eq:4 ] ) ] .in addition there is a unitary transformation between particle operators associated with initial and final nuclei @xmath250 the overlap factors of the single particle wave functions of the initial and final nuclei @xmath251 is given explicitely in the appendix a. the above mentioned transformations allow us , by using the quasiboson approximation , to rewrite the boson operators of the initial nucleus with the help of the boson operators of the final nucleus : @xmath252 , \nonumber \\a^{(i ) } ( { \overline l},k ) & = & \sum_{l ' } [ { \cal r}_{ll ' } a^{(f ) } ( { \overline l}',k ) - { \cal s}_{ll ' } a^{(f ) \dagger } ( { l}',k ) ] .\label{eq : b5 } \end{aligned}\ ] ] assuming the definition of the quasiparticle pairs operator in eq .( [ eq:11 ] ) the factors @xmath253 and @xmath254 can be expressed as @xmath255 it is worthwhile to notice that in the limit initial and final states are indentical @xmath256 and @xmath257 . by inserting eq .( [ eq : b5 ] ) into the expression for the phonon operator of the initial nucleus [ see eq .( [ eq:13 ] ) ] and by exploiting the relations @xmath258,\nonumber\\ a^{(f ) } ( { \overline l } , k ) & = & \sum_{m_f } [ x^{m_f}_{i , k } { { \tilde q}^{m_f}_k } + y^{m_f}_{i , k } { { q}^{m_f \dagger}_k } ] , \label{eq : b7}\end{aligned}\ ] ] we find @xcite @xmath259 .\label{eq : b8}\end{aligned}\ ] ] by neglecting the terms proportional to @xmath254 due to their smallness we end up with the overlap factor of the intermediate nuclear states given in eq .( [ eq:29 ] ) .the overlap factor of the initial and final bcs vacua can be written as product of proton and neutron bcs overlap factors for a given anular momentum projection quantum number @xmath29 : @xmath260 where @xmath261 @xmath262 is the number of single particle states with the same value of quantum number @xmath29 .the same model space for protons and neutrons is assumed . by a direct calculation of the above matrix element onefinds @xmath263 here , @xmath264 means that index @xmath265 runs the values from 1 to @xmath262 except the values @xmath266 and @xmath267 ( @xmath268 and @xmath269 ) .@xmath270 denotes the determinant of matrix of rank @xmath226 constructed of elements of the unitary matrix of the transformation between the initial and final single particle states with row indices @xmath271 and column indices @xmath272 .it is worthwhile to notice that by replacing all determinants in eq .( [ eq : b11 ] ) with unity , i.e. the matrix of the transformation between the single particles associated with both nuclei is just unity matrix , we obtain a compact expression @xcite @xmath273 however , this approximation is not justified and can lead to a significant inaccuracy in the calculation of @xmath165 especially if there is a strong difference in deformations of the intial and final nuclei .we are thankful to a.a .raduta and p. sarriguren for stimulating comments and discussions .this work was supported in part by the deutsche forschungsgemeinschaft ( 436 slk 17/298 ) , by the `` land baden - wrtemberg '' as a `` landesforschungsschwerpunkt : low energy neutrinos '' and by the vega grant agency of the slovac republic under contract no .1/0249/03 . 99 amand faessler and f. imkovic , j. phys .g 24 ( 1998 ) 2139 .j. suhonen , o. civitarese , phys .300 ( 1998 ) 2139 .h. ejiri , phys .338 ( 2000 ) 265 .m. doi , t. kotani and e. takasugi , prog .83 ( 1985 ) 1 .klapdor - kleingrothaus , _springer tracts in modern physics _ , vol .163 ( springer - verlag , 2000 ) , pp .elliott and p. vogel , ann .52 ( 2002 ) 115 .vergados , phys . rep .361 ( 2002 ) 1 .tretyak and yu.g .zdesenko , at .data nucl .data tabl .80 ( 2002 ) 83 .barabash , czech .52 ( 2002 ) 567 .p. vogel and m. r. zirnbauer , phys .57 ( 1986 ) 3148 .o. civitarese , a. faessler and t. tomoda , phys .b 194 ( 1986 ) 11 .k. muto , e. bender and h. v. klapdor , z. phys . a 334 ( 1989 ) 177 .cheoun , a. bobyk , amand faessler f. imkovic and g. teneva , nucl .a 561 ( 1993 ) 74 ; nucl .a 564 ( 1993 ) 329 .j. toivanen , j. suhonen , phys .lett . 75 ( 1995 ) 410 .j. schwieger , f. imkovic , and a. faessler , nucl .a 600 ( 1996 ) 179 .raduta , a. faessler , s. stoica , and w.a .kaminski , phys .b 254 ( 1991 ) 7 ; nucl . phys . a 534 ( 1991 ) 149 .a. a. raduta , f. imkovic , a. faessler , j. phys .g 26 ( 2000 ) 793 .l. zamick and n. auerbach , phys .c 26 ( 1982 ) 2185 .d. bogdan , a. faessler , a. petrovici , and s. holan , phys .b 150 ( 1985 ) 29 .k. grotz and h.v .klapdor , phys .b 157 ( 1985 ) 242 . j.g .hirsch , o. castanos , p.e .hess , rev .38 suppl . 2 ( 1992 ) 66 ; o. castanos , j.g .hirsch , p.e .hess , rev .39 suppl . 2 ( 1993 ) 29 .o. castanas , j.g .hirsch , o. civitarese , and p.o .hess , nucl .phys . a 571 ( 1994 ) 276 ; j.g .hirsch , o. castanas , p.e . hess and o. civitarese , nucl .phys . a 589 ( 1995 ) 445 ; phys .c 51 ( 1995 ) 2252 .raduta , a. faessler , d.s .delion , phys .b 312 ( 1993 ) 13 , nucl . phys . a 564 ( 1993 ) 185 .dixit , p.k .rath , and p.k .raina , phys .c 65 ( 2002 ) 034311. j. krumlinde and p. mller , nucl .phys . a 417 ( 1984 ) 447 .p. sarriguren , e.moya de guerra , a. escuderos , a.c .carrizo , nucl .phys . a 635 ( 1998 ) 55 .p. sarriguren , e. moya de guerra , a. escuderos , nucl .phys . a 658 ( 1999 ) 13 ; phys . rev .c 64 ( 2001 ) 064306 ; nucl .phys . a 691 ( 2001 ) 631 .klapdor - kleingrothaus et al . , eur .j. a 12 ( 2001 ) 147 .zdesenko , rev .mod . phys .74 ( 2002 ) 663 .haxton and g.s .stephenson , prog .nucl . phys .12 ( 1984 ) 409 .f. imkovic , g. pantis , a. faessler , prog .40 ( 1998 ) 285 ; phys .61 ( 1998 ) 1218 .w.a . kaminski and a. faessler ,phys . a 529 ( 1991 ) 605 .j. suhonen , phys .b 477 ( 2000 ) 99 .a. griffiths and p. vogel , phys .c 46 ( 1992 ) 181 .j. damgaard , h.c .pauli , v.v .pashkevich , and v.m .strutinski , nucl .phys . a 135 ( 1969 ) 432 .y. tanaka , y. oda , f. petrovich , and r.k .sheline , phys .b 83 ( 1979 ) 279 .r. nojarov , z. bochnacki , and a. faessler , z. phys . a 324 ( 1986 ) 289 .r. nojarov , a. faessler , p. sarriguren , e. moya de guerra and m. grigorescu , nucl .phys . a 563 ( 1993 ) 349 .p. sarriguren , e. moya de guerra , r. nojarov and a. faessler , j. phys .g 19 ( 1993 ) 291 ; g 20 ( 1994 ) 315 .lalazissis , s. raman , p. ring , at .data nucl .data tabl . 71( 1999 ) 1 .p. sarriguren , e. moya de guerra , l. pacearescu , a. faessler , f. imkovic and a.a .raduta , phys .c 67 ( 2003 ) 044313 .p. mller , j.r .nix , w.d .myers and w.j .swiatecki , at .data nucl .data tabl .59 ( 1995 ) 185 .s. balraj , nucl .data sheets 74 ( 1995 ) 63 .p. raghavan , at .data nucl .data tabl .42 ( 1989 ) 189 ; n.j .stone , oxford university preprint , http://ie.lbl.gov/toi.html .s. raman , c.h .malarkey , w.t .milner , c.w .nestor jr . and p.h .stelson , at .data nucl .data tabl .36 ( 1987 ) 1 . g. audi and a.h .wapstra , nucl .phys . a 595 ( 1995 ) 409 ; g. audi , o. bersillon , j. blachot , and a.h .wapstra , nucl .phys . a 624 ( 1997 ) 1 .h. homma , e. bender , m. hirsch , k. muto , h.v .kleingrothaus , and t. oda , phys .c 54 ( 1996 ) 2972 .j. suhonen , j. taigel , and a. faessler , nucl .phys . a 486 ( 1988 ) 91 .raduta , v. ceausescu , a. gheorghe , and m. popa , nucl .phys . a 427 ( 1984 ) 1 .r. madey , b.s .flanders , b.d .anderson , a.r .baldwin , j.w .watson , s.m .austin , c.c .foster , h.v .klapdor and k. grotz , phys .c 40 ( 1989 ) 540 .helmer et al . , phys .c 55 ( 1997 ) 2802 .f. imkovic , p. domin , s.v .semenov , j. phys .g 27 ( 2001 ) 2233 .bilenky , c. giunti , w. grimus , b. kayser , and s.t .petcov , phys .b 465 ( 1999 ) 193 ; h.v .klapdor - kleingrothaus , h. ps , a.y .smirnov , phys .d 63 ( 2001 ) 073005 . a. faessler and f. imkovic , prog .48 ( 2001 ) 233 . f. imkovic , ch .moustakidis , l. pacearescu and a. faessler , submitted to phys .c .three sets of nuclear structure input parameters ( a , b , and c ) for which the calculated nuclear matrix elements @xmath107 reproduces the experimental @xmath0-decay half - life of @xmath3 .denotes overlap factor of the initial and final bcs vacua . [ cols="^,^,^,^,^,^,^,^,^,^,^,^,^ " , ]

<S> the effect of deformation on the two - neutrino double decay ( @xmath0-decay ) for ground state transition @xmath1 is studied in the framework of the deformed qrpa with separable gamow - teller residual interaction . a new suppression mechanism of the @xmath0-decay matrix element based on the difference in deformations of the initial and final nuclei is included . </S><S> an advantage of this suppression mechanism in comparison with that associated with ground state correlations is that it allows a simultaneous description of the single @xmath2 and the @xmath0-decay . by performing a detail calculation of the @xmath0-decay of @xmath3 </S><S> , it is found that the states of intermediate nucleus lying in the region of the gamow - teller resonance contribute significantly to the matrix element of this process . </S><S> pacs numbers:23.40.bw,23.40.hc -1.5 cm -1 cm * two neutrino double beta decay of @xmath3 within deformed qrpa : a new suppression mechanism * fedor imkovic@xmath4 , larisa pacearescu@xmath5 and amand </S><S> faessler@xmath5 _ 1 ) institute fr theoretische physik der universitt tbingen , + auf der morgenstelle 14 , d-72076 tbingen , germany _ + _ 2 ) department of nuclear physics , comenius university , + sk842 48 bratislava , slovakia _ + </S>

1607.06663

supergravity can be obtained as a low energy field theory approximation of string theory @xcite .the leading order correction to this theory depends on the string used to obtain the low energy effective field theory expansion . in the heterotic string theory ,the lowest corrections are described by a gauss - bonnet ( gb ) term @xcite .furthermore , it is known from the renormalization group flow that the constants depend on the scale at which a theory is probed @xcite .so , we expect that the different constants in general relativity and gb gravity also depend on the scale at which these theories are measured . the renormalization group flow for supergravity solutions @xcite and gb gravity @xcite has been analyzed .in fact , a renormalization group flow has been used for measuring the flow of the cosmological constant @xcite and newton constant @xcite .now , as the scale at which a theory is measured depends on the energy of the probe , it is expected that these constants will also depend on the energy .therefore , in this paper , we will use gravity s rainbow @xcite for analyzing gb black holes with the energy dependent constants .it may be noted that the initial motivation of gravity s rainbow came by analyzing the target space metric of string theory .this was done by regarding string theory as a two dimensional theory , and considering the target space metric as a matrix of coupling constants . then using renormalization group flow ,this matrix of coupling constants would flow and depend on the scale at which this theory is measured .this would in turn make them dependent on the energy that is used for probing this theory @xcite .thus , the geometry would also be energy dependent .such a modification of a geometry at high energy scale can be viewed as a uv completion .it may be noted that just like the horava - lifshitz gravity horavaprd , horavaprl , the gravity s rainbow @xcite , has also been viewed as a uv completion of general relativity .this is because both of these approaches are based on a modification of the usual energy - momentum dispersion relation in the uv limit .such a modification of the usual energy - momentum dispersion relation in the uv limit occurs in a large number of approaches to quantum gravity , such as the discrete spacetime @xcite , models based on string field theory @xcite , spacetime foam ellis , spin - network in loop quantum gravity @xcite , non - commutative geometry @xcite and ghost condensation @xcite . it may be noted that even the greisen - zatsepin - kuzmin limit ( gzk limit ) suggests that the usual energy - momentum relation could get modified in the uv limit @xcite .the gzk limit can be used for analyzing quantum gravitational effects as an upper limit on the energy of cosmic ray .the pierre auger collaboration and the high resolution fly s eye experiment have reconfirmed earlier results of the gzk cutoff @xcite .all these observations suggest that there is a strong experimental motivation for a uv modification of the usual energy - momentum dispersion relation .motivated by the horava - lifshitz gravity , the geometries occurring in the types iia @xcite and iib string theory @xcite have been also modified in the uv limit .different lifshitz scaling for space and time have also been used for analyzing certain aspects of the ads / cft correspondence @xcite .it may be noted that for a suitable choice of the rainbow functions , the horava - lifshitz gravity can be related to the gravity s rainbow garattini .this is because both of these approaches are based on a modification of general relativity in the uv limit such that the general relativity is obtained in the ir limit .basically , horava - lifshitz gravity has been used to study the uv completion of various interesting geometries .motivated by the close relation between the gravity s rainbow and horava - lifshitz gravity , the uv completion of many black hole solutions has been recently obtained by using the formalism of gravity s rainbow ali341 one of the most interesting effects of the uv completion of geometries is the modification of the black hole thermodynamics at the last stage of the evaporation of the black holes .as the gravity s rainbow reduces to the general relativity in the ir limit , the black hole thermodynamics in gravity s rainbow reduces to the usual black hole thermodynamics for very large black holes .however , as the black holes evaporate and reduce in size , the black hole thermodynamics in gravity s rainbow showing deviation from the usual black hole thermodynamics .this deviation becomes significant at the end stage of the evaporation of black hole in gravity s rainbow ali104040 .it has been observed that the temperature of black holes reaches a maximum value , and then it starts to reduce beyond this maximum value . at a critical valuethe temperature of black holes becomes zero , and the black holes do not radiate hawking radiation at that stage .so , a black hole remnant forms in gravity s rainbow .this formation of a black hole remnant has important phenomenological consequences for the detection of mini black holes at the lhc @xcite .it is known that the usual uncertainty principle has to be modified to the generalized uncertainty principle to have an upper bound on the energy of a particle . as thisparticle acts as a probe for the geometry of the black hole , it also fixes the energy scale of the gravity s rainbow .thus , this bound on the energy of a particle emitted in the hawking radiation can be used as energy scale in the rainbow functions .this modifies the thermodynamics of black holes ali104040,hpemrainbow .such a modification in the black hole thermodynamics has also been observed for black rings @xcite .the temperature of black rings also reaches a maximum value and reduces beyond that value to form a remnant .it has been argued that a remnant will form for all black objects @xcite .this has been explicitly demonstrated for kerr black holes , kerr - newman black holes in de sitter space , charged anti - de sitter ( ads ) black holes , higher dimensional kerr - ads black holes and black saturn @xcite .the geometric and thermodynamic properties of the charged dilatonic black holes in gravity s rainbow have also been investigated @xcite .the effect of gravity s rainbow on the thermodynamics of black holes in gb gravity coupled to maxwell s theory has been studied @xcite . in this analysis , it was observed that even though the thermodynamics of the black holes get modified in the gb gravity s rainbow , the first law of thermodynamics still holds for this modified thermodynamics .however , in this analysis , different constants were not made to depend on the energy of the probe .as we expect the constant to depend on the energy of the probe , in this paper , we will generalize this analysis and make the constants energy dependant .we will also investigate the critical behavior of black hole solutions in this theory .it may be noted that the pressure - volume ( pv ) criticality has been studied for black holes in gb gravity using extended phase space thermodynamics @xcite . in the extended phase space thermodynamics, the cosmological constant is viewed as a thermodynamic pressure , and so it is possible to define a thermodynamic volume conjugate to this thermodynamic pressure @xcite . it has been observed that along with the usual black hole phase transition , a new phenomenon of reentrant phase transitions occurs for rotating ads black holes in extended phase space @xcite . in these reentrant phase transitions a monotonic variation of the temperature yields two phase transitions , and this situation is similar to that which is seen in multi - component liquids . the pv criticality has also been studied in quasi - topological gravity @xcite . in this paper , we study the effect of the uv completion of general relativity on the pv criticality in extended phase space .also , we analyze the effects of rainbow functions on pv criticality in gb gravity s rainbow . in addition , we analyze the effect of energy dependent constants on thermodynamic properties of black hole solutions .we also study the effects of such modification on the critical values and van der waals like behavior of the system .as it was mentioned before , we expect that all the constants depend on the energy of the probe .such a dependency of the constants on the energy of the probe can be explicitly analyzed using gravity s rainbow .this is an advantage of using the gravity s rainbow .now , following earlier work on einstein - gb - maxwell black holes in gravity s rainbow , we will generalize solutions to the case where different constants are dependent on the energy of the probe .our first step is analyzing the effect of this energy dependency on the field equations .the @xmath0-dimensional action of gb - maxwell gravity with negative cosmological constant can be written as @xmath1 , \label{action}\]]where @xmath2 and we set @xmath3 and @xmath4 in which the last one is the so - called gb coefficient . in this paper, we will only consider positive values of gb coefficient .in addition , @xmath5 s are the first three terms of lovelock lagrangian which are corresponding to the cosmological constant , einstein and gb lagrangian , with the following explicit forms @xmath6 the @xmath7 is the lagrangian of electrodynamics in which we choose linear maxwell lagrangian , @xmath8 , where @xmath9 is the maxwell invariant. variation of the action ( [ action ] ) with respect to the metric tensor @xmath10 and the faraday tensor @xmath11 , leads to the following field equations@xmath12 @xmath13where @xmath14 following ref .@xcite , one finds the spherical symmetric metric governing the gravity s rainbow which has dependency on rainbow functions in following form @xmath15 ~. \label{metric}\ ] ] using eqs .( [ field equation ] ) - ( [ g2 ] ) with metric ( [ metric ] ) , we can find following electromagnetic field tensor and metric function @xmath16where @xmath17 .also , @xmath18 and @xmath19 are integration constants which are , respectively , related to the total electric charge and total mass of the solutions , and @xmath20 . for the simplicity and in order to find the contribution of each parameters , we consider following notations @xmath21and since we are working in natural units , we set @xmath22 . here , we would like to make some remarks regarding the properties of the solutions .calculations show that there is a curvature singularity at @xmath23 , which can be covered with an event horizon .hence , it will be interesting to see the effects of the gravity s rainbow on the singularity and asymptotical behavior of the solutions . in order to study these effects ,we use series expansion of the kretschmann scalar for small and large values of radial coordinate . by doing so, one can find following relations @xmath24@xmath25 , \label{rrinf}\]]where@xmath26 .\label{leff}\ ] ] interestingly , the obtained relations indicate that the rainbow functions affect the strength of singularity as well as asymptotical behavior of the solutions . in other words ,the asymptotical behavior of the solutions is ads with an effective cosmological constant , @xmath27 .it is clear that the gb parameter and rainbow function can modify @xmath27 .it means that @xmath27 reduces to @xmath28 for @xmath29 and @xmath30 , simultaneously .now , we are in a position to study thermodynamical quantities of the solutions . using the concept of the surface gravity ,it is a matter of calculation to obtain temperature as @xmath31 r_{+}^{2d}-2q^{2}d_{3}^{2}h_{1}^{4}(e ) h_{3}^{2}(e ) g^{2}(e ) f^{2}(e ) r_{+}^{8}}{4\pi d_{2}\left ( r_{+}^{2}+2\alpha ^{\prime } h_{4}^{2}(e ) g^{2}(e ) \right ) g(e)f(e)r_{+}^{2d_{-1/2}}}. \label{tnew}\ ] ] one of the conditions for having physical solutions is imposed by positive temperature . since we have considered positive values of gb parameter , the denominator of the temperature is always positive and only numerator has contribution to negativity of the solutions . for ads black holes , only charge term in numerator of the temperaturecontribute to negativity of it .taking closer look at the numerator , one can see that energy dependency of constants plays a crucial role in domination of different terms .therefore , the conditions for having physical solutions ( positive temperature ) is highly sensitive to energy variation of different constants .since we are working in the context of higher derivative gravity , it is not allowed to use the area law for calculating entropy .we use the wald formula with the following result @xmath32 in addition , the total electric charge of the solutions is obtained through the use of gauss s law as @xmath33 for the electric potential , we have @xmath34 the total mass of the black holes could be obtained through counter - term method or the arnowitt - deser - misner approach with the following unique form @xmath35where by evaluating metric function on outer horizon , one can find total finite mass as @xmath36 r_{+}^{d_{5}}+q^{2}d_{1}d_{3}h_{1}^{4}(e)h_{3}^{2}(e)g^{2}(e)f^{2}(e)r_{+}^{-2d_{2}}% } { 8\pi d_{1}f(e)g^{d_{1}}(e)}. \label{mnew}\ ] ] before we proceed , it is worthwhile to mention a few remarks about conserved and thermodynamical quantities .the modification of gb gravity when the constants are not energy dependent has already been analyzed @xcite .thermodynamical quantities of the solutions in @xcite differ from those obtained here which highlights the effects of dependency of constants on the probe energy .especially , the electric potential was shown to be independent of energy function in ref .@xcite , while here , it depends on the energy of the probe . strictly speaking ,these two types of black holes are phenomenologically different .recently , it was pointed out that most of constants in physics are not actually constant . in fact, a measurement of their expectation values leads to the result that they should be varying parameters . here, we have taken such consideration into account and shown that thermodynamically speaking , black holes will be modified in such consideration .now , we are in a position to check the validation of the first law of black hole s thermodynamics . using total mass of these black holes ( [ mnew ] ) with the obtained entropy ( [ snew ] ) and electric charge ( [ qnew ] ) as extensive parameters, one can define following intensive parameters @xmath37 where by evaluating these equations , one can confirm that the obtained temperature and electric potential in eq .( [ tu ] ) coincide with those extracted from eqs .( [ tnew ] ) and ( [ unew ] ) . therefore , despite the modifications in thermodynamical quantities by consideration of energy dependant constants , the first law of thermodynamics is valid for these black holes .the extended phase space expression comes from consideration of the cosmological constant not as a fixed quantity but a thermodynamical variable which is known as pressure .although thermodynamical pressure is generally proportional to the negative cosmological constant with a proportionality constant @xmath38 , there are some cases in which we should modify it @xcite . in this paper , it is observed that the metric function and asymptotical behavior of the system have been modified due to the existence of rainbow functions .therefore , it is necessary to check the possible effects of rainbow functions on thermodynamical pressure . to do so , we evaluate the energy - momentum tensor .it is straightforward to obtain the following relations for different components of energy - momentum tensor @xmath39@xmath40 in these relations , the first term is due to the existence of electromagnetic field ( which is coupled with rainbow functions ) .surprisingly , the @xmath28 term is not coupled with any rainbow functions of the metric . in other words , although rainbow functions of the metric modify @xmath28 term in the metric function and asymptotical behavior of the solutions , they do not affect thermodynamical pressure which is related to the cosmological constant .therefore , in studying the critical behavior of the system through the analogy between @xmath41 and @xmath42 , one can use following relation @xmath43 thermodynamically speaking , the conjugating thermodynamical variable corresponding to the pressure is thermodynamical volume which in the context of enthalpy is calculated by @xmath44 it is worthwhile to mention that in order to have a well - defined vacuum solution with @xmath45 , the pressure @xmath42 has to satisfy the following constraint @xcite @xmath46 in which it puts a restriction on the pressure as maximal pressure @xmath47which indicates that only for sufficiently small pressures , the solution ( [ metric function ] ) has an asymptotic ads region .now , remembering that we are working in extended phase space , the total mass of the black holes will not play the role of internal energy .instead , it is interpreted as enthalpy . with this consideration, one can find the gibbs free energy as @xmath48 as for the volume of these black holes , by using eqs .( [ mnew ] ) , ( pnew ) and ( [ v ] ) , we obtain @xmath49 which is modified in the presence of the rainbow functions .in other words , in the presence of gravity s rainbow , thermodynamical volume of the black holes is a function of the rainbow functions , and therefore , the final form of these black holes is determined by the model of rainbow functions under consideration . on the other hand , even by consideration of the dependency of different constants on the probe energy ,the volume of these black holes is same as that of probe energy independent constants .this behavior is expected , since the abstract form of metric ( [ metric ] ) is free of any constant .the equation of state and gibbs free energy of these solutions could be found by using eqs .( [ tnew ] ) , ( [ pnew ] ) and ( [ g ] ) , which result into @xmath50 g(e)f(e)}{4r_{+}^{3}}t+\frac{% d_{3}^{2}h_{1}^{4}(e)h_{3}^{2}(e)g^{2}(e)f^{2}(e)}{8\pi r_{+}^{2d_{2}}}q^{2 } \notag \\ & & -\frac{d_{2}\left [ d_{3}r_{+}^{2}+\alpha ^{\prime } d_{5}h_{4}^{2}(e)g^{2}(e)\right ] g^{2}(e)}{16\pi r_{+}^{4 } } , \label{ppnew}\end{aligned}\]]and @xmath51 there are several approaches toward calculating critical values . in this paper , we will employ the properties of the inflection points in @xmath52 diagrams . in this method, one can follow the relations for calculating critical volume which in case of these black holes , it will be proportional to the horizon radius@xmath53where by using eqs .( [ ppnew ] ) and ( [ infel ] ) , one obtains @xmath54 } { % r_{+}^{2d_{5}}}-d_{3}g(e)r_{+}^{4}+12\alpha ^{\prime } h_{4}^{2}(e ) g^{3}(e ) % \left [ r_{+}^{2}-\alpha ^{\prime } d_{5}h_{4}^{2}(e ) g^{2}(e ) \right ] = 0.\ ] ] as a special case , we consider @xmath55-dimensional solutions in the absence of electric charge to obtain critical horizon radius , temperature and pressure .so , we can write @xmath56 which lead to the following ratio @xmath57 it is notable that in the absence of rainbow functions ( @xmath58 ) , this ratio reduces to @xmath59 , and therefore , eq . ( [ pvcnew ] ) indicates that consideration of the gravity s rainbow can modify this near universal ratio . interestingly , for neutral solutions ( @xmath60 ) , the critical pressure is independent of metric s rainbow functions while the critical horizon radius and the critical temperature are functions of one of rainbow functions .next , we are going to consider the charged gb black holes in the presence of gravity s rainbow in @xmath61dimensions . the critical horizon radius in this caseis given as @xmath62 } } { 3\mathcal{b}^{1/3}}. \label{rcnew}\]]where@xmath63 it is worthwhile to emphasize that , in this paper , we consider the positive values of gb parameter ( @xmath64 ) .the negative value of @xmath65 enforces other set of conditions for having positive critical parameters .furthermore , @xmath66 will be given by @xmath67 } { \pi \left ( r_{c - q}^{2}+6\alpha ^{\prime } h_{4}^{2}(e)g^{2}(e)\right ) f(e)r_{c - q}^{3 } } , \label{tcnew}\]]and @xmath68 is @xmath69 -6\alpha ^{\prime } h_{4}^{2}(e)g^{2}(e)r_{c - q}^{4}\right\ } } { 8\pi \left ( r_{c - q}^{2}+6\alpha ^{\prime } h_{4}^{2}(e)g^{2}(e)\right ) r_{c - q}^{6}}. \label{pcnew}\ ] ] the obtained relation for @xmath66 imposes a restriction for having positive critical temperature .since denominator of @xmath66 is positive , the restriction for having positive @xmath66 comes from the numerator of this relation with the following explicit form @xmath70 it is evident that the positivity of the critical temperature depends on the energy variations of constants , the electric charge and energy function of the metric . as for the critical pressure , similarly, only its numerator may yield negative values .in other words , the numerator of the critical pressure imposes following condition for having positive values of @xmath68 @xmath71 -6\alpha ^{\prime } h_{4}^{2}(e)g^{2}(e)r_{c - q}^{4}>0.\ ] ] here , the restriction is rooted in two parts of the action ; one is related to the generalization of gb gravity and the other one is related to the electrodynamic part . in other words , by cancelling the contributions of gb gravity and electromagnetic field , the pressure will always be positive .now , we are in a position to calculate @xmath72 . using eqs .( [ rcnew ] ) , ( [ tcnew ] ) and ( [ pcnew ] ) , we obtain @xmath73 -6\alpha ^{\prime } h_{4}^{2}(e)g^{2}(e)r_{c - q}^{4}\right\ } } { 8\left ( r_{c - q}^{4}-4q^{2}h_{1}^{4}(e)h_{3}^{2}(e)f^{2}(e)\right ) r_{c - q}^{2}}. \label{pvtneww}\ ] ] first of all , contrary to the absence of charge , in this case , all critical values are functions of both of the rainbow functions of metric .this emphasizes the fact that thermodynamical structure of the charged black holes in gravity s rainbow is completely different from the neutral ones .the critical horizon radius must be real valued and positive .therefore , there is a restriction @xmath74 we should note that this restriction is due to the existence of charge . in other words ,generalization from neutral to charged solutions , put restrictions on values that rainbow functions and gb parameter can acquire .@xmath75 @xmath76 @xmath77 @xmath78 @xmath79 @xmath80 @xmath81 @xmath82 the obtained critical parameters indicate that in general , these black holes with consideration of energy dependent constants , enjoy second order phase transition in their phase space . although the presence of second order phase transition is restricted to satisfy specific conditions , in general the second order phase transitionis a part of thermodynamical properties of these black holes .now , in order to elaborate the existence of second order phase transition , we choose specific examples for free parameters and energy functions to plot three set of diagrams which are @xmath83 , @xmath52 and @xmath84 .the existences of subcritical isobars for critical pressure in @xmath83 diagram , the region of phase transition ( reflection point ) for critical temperature in @xmath52 diagram and swallow - tail for pressures smaller than critical pressure in @xmath84 diagram , indicate that a second order phase transition is taking place for the obtained critical values . taking into account the rainbow functions , we regard three known models for considering their effects . the first model is motivated from loop quantum gravity and non - commutative geometry , in which rainbow functions are lqg @xmath85 the second model comes from the hard spectra of gamma - rays motivation with the following form @xcite @xmath86taking the constancy of the velocity of the light into account, one can find following relations for the rainbow functions as third model @xmath87 now , using eqs .( [ tnew ] ) , ( [ ppnew ] ) and ( [ gnew ] ) with the obtained critical values ( eqs .( [ rcnew ] ) - ( [ pcnew ] ) ) , one can plot the following @xmath83 , @xmath52 and @xmath84 diagrams for different models of rainbow functions in special cases ( eqs .( [ loop ] ) - ( [ light ] ) ) ( figs .[ fig1 ] - [ fig3 ] and [ fig4 ] - [ fig6 ] ) .it is evident from plotted graphs that the obtained critical values are representing the second order phase transition points ( due to the characteristic behaviors of the different phase diagrams ) . in some classes of rainbow functions , for neutral case ,variations of the parameters of rainbow functions have no effect ( see right and left panels of figs . fig1 and [ fig2 ] , respectively ) .this specific behavior was not observed in the charged solutions . by excluding right panel of fig .[ fig1 ] , in other cases , the critical temperature , the size of swallow - tail and differences between energy of different phases are decreasing functions of the parameters of rainbow functions ( @xmath88 , @xmath89 and @xmath90 ) ( see right panels of figs. fig2 , [ fig3 ] and [ fig4 ] [ fig6 ] ) .meanwhile , by excluding left panel of fig .[ fig2 ] , the critical pressure is a decreasing function of @xmath88 and @xmath89 ( see left panels of figs . [ fig1 ] , fig4 and [ fig5 ] ) and an increasing function of @xmath90 ( see left panels of figs . [ fig3 ] and [ fig6 ] ) . as for subcritical isobar , except for the neutral case of loop motivated rainbow functions ( see middle panel of fig .[ fig1 ] ) , its length is an increasing function of the parameters of rainbow functions ( see middle panels of figs .[ fig2 ] , fig3 and [ fig4 ] [ fig6 ] ) . in the absence of charge ,presence of the rainbow functions provides interesting effects . taking a closer look at the fig .[ fig1 ] shows that while we are varying rainbow function which results into modification of the @xmath52 diagram , the critical temperature remains fixed ( fig .[ fig1 ] middle and right panels ) , and interestingly , the total behavior of the gibbs free energy versus temperature is not affected by this variation either .same behavior is observed in fig .[ fig2 ] left panel .this property enables us to modify different critical values while a specific one of them remains unchanged .using the fact that the gibbs free energy , temperature , and pressure of the system are constant during the phase transition , we have plotted the coexistence line of the black holes ( figs .[ coexq0 ] and [ coexq ] ) .the small and large black holes have identical temperature and pressure along the coexistence line , and the critical points are located at the end of the coexistence line where above these points the phase transition does not occur .[ coexq0 ] ( left panel ) shows that the coexistence lines for variation of @xmath88 are completely identical , because @xmath91 has no effect on the critical pressure and temperature .on the other hand , fig .[ coexq ] ( left panel ) indicates that variation of @xmath92 does not have significant effect on the coexistence lines .according to properties of the coexistence lines , one can conclude that for these black holes the reentrant of phase transition does not take place .in this paper , the dependency of all constants on energy functions was considered , and charged gb black holes in the presence of gravity s rainbow were studied .thermodynamical and geometrical properties of these black holes were investigated and it was shown that the power of the singularity , asymptotical behavior and thermodynamical quantities of these black holes were modified due to the dependency of constants on energy . first of all , we found that in the presence of gravity s rainbow , thermodynamical volume of the black holes is modified and it is rainbow function dependent . in other words , the total behavior of the volumeis determined by rainbow functions as well as dimensions .next , we found that , for neutral case , the obtained critical radius and critical temperature were functions of rainbow functions of the metric , whereas the critical pressure was independent of them and was only dependent on energy variation of the constants , which in return resulted in specific behaviors in phase diagrams in special cases .interestingly , in the presence of the electric charge , a limitation was found for having real positive critical horizon radius .in other words , the presence of charge puts restriction on values that different parameters can acquire .contrary to the neutral case , in the case of charged solutions , all critical values ( the critical temperature , the horizon radius and the pressure ) were dependent on rainbow functions of the metric .in other words , the critical behavior of the system was modified due to the presence of gravity s rainbow .it is worth mentioning that the presence of rainbow functions was observed in ratio of @xmath93 for both charged and neutral cases , which is a variation of a similar ratio in van der waals system of liquid / gas .in addition , the coexistence lines indicated that for this type of black holes , the reentrant of phase transition does not happen .the specific behaviors and results of the paper motivate one to analyze new interesting phenomenology for such black hole thermodynamics .it will be worthwhile to study the effects of non - linear electrodynamics on critical behavior of gb gravity s rainbow .it will be interesting to generalize the obtained static solutions to a case of dynamical ones and investigate the cosmological consequences of such solutions .the behavior of the hawking radiation near the critical point is another subject of interest .the author would like to thank anonymous referees for suggesting important improvements .we thank shiraz university research council .this work has been supported financially by the research institute for astronomy and astrophysics of maragha , iran .

<S> following an earlier study regarding gauss - bonnet - maxwell black holes in the presence of gravity s rainbow [ s. h. hendi and m. faizal , phys . rev . </S><S> d 92 , 044027 ( 2015 ) ] , in this paper , we will consider all constants as energy dependent ones . </S><S> the geometrical and thermodynamical properties of this generalization are studied and the validation of the first law of thermodynamics is examined . next , through the use of proportionality between cosmological constant and thermodynamical pressure , van der waals - like behavior of these black holes in extended phase space </S><S> is investigated . an interesting critical behavior for sets of rainbow functions in this case </S><S> is reported . </S><S> also , the critical behavior of uncharged and charged solutions is analyzed and it is shown that the generalization to a charged case puts an energy dependent restriction on values of different parameters . </S>